ICR 2026

Development of a method for synthesizing adaptive control systems for unmanned aerial vehicles

Anastasiia Yurievna Serdiukova — 2026-10-05

The paper proposes a method for the synthesis of an adaptive control system (CS) for the spatial motion of a quadrotor unmanned aerial vehicle (UAV), ensuring desired dynamic performance under conditions of parametric uncertainty and external disturbances. The proposed CS is designed to maintain stability and tracking accuracy in the presence of model inaccuracies and unknown environmental effects, including time-varying disturbances. The CS has a cascaded structure with an outer position control loop and an inner attitude control loop, implemented using nonlinear position controllers (NPC) and nonlinear attitude controllers (NAC), respectively. Parametric uncertainties and external disturbances are estimated using variable structure observers (VSOs), whose estimates are incorporated into the control laws to provide real-time compensation of uncertainties. The synthesis of the CS and VSOs is based on a mathematical model of the UAV that accounts for the center-of-mass (CM) offset and key dynamic effects. The combined influence of external disturbances and parametric uncertainties is represented in the dynamic equations as an additive disturbance term. The effectiveness of the proposed approach is demonstrated through numerical simulations in stabilization and trajectory tracking tasks.

Recursive Batch Gaussian Sum Filter for Single-Beacon Navigation Problem

alexey_isaev — 2026-10-05

A problem of determining the position of a slave autonomous unmanned under-water vehicle (AUV) using range measurements to the leader AUV is considered within the framework of the Bayesian stochastic approach. This problem can be interpreted as a single-beacon navigation problem. Its key feature is that the form of the posterior probability density function (PDF) varies from multi-extremal to single-extremal one. Previously, this problem was solved with the use of the Recurrent Iterative Batch Linearized Smoothing Filter with Multiple Linearization (RI-BMLS) proposed by the authors. The essence of that approach lies in identifying the point in time at which the posterior PDF becomes single-extremal, and then the recursive iterative Kalman filter is used to solve the problem. However, the RI-BMLS has some significant drawbacks: first, it is ineffective until the onset of aforesaid point in time; and second, the RI-BMLS can be computationally intensive when the posterior PDF remains multi-extremal over a long-time interval. This paper proposes the Recursive Iterative Batch Gaussian Sum Filter (RI-BGSF), which is a modification of the RI-BMLS. Unlike the RI-BMLS, the proposed algorithm allows for the effective solution of the single-beacon navigation problem throughout the entire problem solution, which is made possible due to the representation of the posterior PDF at its multi-extremum points as a weighted set of Gaussian densities. The RI-BGSF is significantly much simpler computationally owing to the combined use of recursive and nonrecursive schemes in its design. The advantages of the proposed algorithm are demonstrated through simulations.

Design of Heterogeneous Robotic Warehouse Systems Based on an Intelligent Zone-Based Multi-Agent Dispatching Method

Anatoly Igorevich Kalyaev — 2026-10-05

Design of Heterogeneous Robotic Warehouse Systems Based on an Intelligent Zone-Based Multi-Agent Dispatching Method

О МЕТОДЕ ПОВЫШЕНИЯ ТОЧНОСТИ ПОЗИЦИОНИРОВАНИЯ ГРУППЫ АНПА С ИСПОЛЬЗОВАНИЕМ ГАНС КБ

alex-scherba — 2026-10-05

В статье предложен экономичный способ решения задачи одновременного определения координат группы недорогих автономных необитаемых подводных аппаратов /АНПА/. Полагается, что на борту каждого АНПА отсутствуют дорогостоящие системы точного времени и абсолютные датчики скорости. Представленный способ навигационного обеспечения группы АНПА основан на использовании гидроакустической навигационной системы с короткой базой /ГАНС КБ/. Рассмотрено исполнение ГАНС КБ, которое включает две приемные антенны на борту каждого АНПА и один излучающий маяк с известным местоположением. Для решения навигационной задачи используются измерения разности времени прихода сигнала в каждом такте на приемные антенны ГАНС КБ, разности времени прихода сигналов в двух последовательных тактах на одну приемную антенну и измерения частоты принятого на борту АНПА сигнала. Описан алгоритм оценивания местоположения АНПА на основе фильтра частиц, предназначенный для работы на борту каждого АНПА в режиме реального времени. Представлены результаты работы предложенного алгоритма, полученные в процессе компьютерного моделирования, которые подтвердили его более высокую точность по сравнению с версией алгоритма, в которой измерения разности времени прихода сигналов в двух последовательных тактах на одну приемную антенну не используются.

UAV Stabilization System Based on Machine Vision and Inertial Sensors

Kristina Obukhova — 2026-10-05

Unmanned aerial vehicles (UAVs) are widely used in areas such as remote monitoring in agricultural industry, infrastructure inspection, logistics, and exploration of hard-to-reach areas. However, many existing approaches to stabilization and navigation rely on global navigation satellite systems, whose effectiveness is reduced in conditions with limited or no satellite signal. The objective of this study is to develop a UAV stabilization system based on machine vision and inertial sensors for operation in GNSS-denied environments. The proposed method uses visual-inertial odometry implemented using the OpenVINS framework, fusing camera data and inertial measurements within an extended Kalman filter. The system is implemented as a distributed software architecture with a client-server model, data transfer via TCP, and integration with the robot operating system for real-time data processing and state estimation. The developed system ensures reliable collection, synchronization, and processing of visual and inertial data, and enables UAV position and orientation estimation under correct acquisition and data synchronization conditions. A stabilization mechanism based on maintaining a predetermined hover point and generating corrective control inputs was implemented in a simulation environment. The proposed system lays the foundation for further research to improve estimation accuracy and refine control strategies for autonomous UAV operation

Autonomous Flight System for UAVs to Selected Objects

Mikhail Stepantsov — 2026-10-05

Manual control of an agricultural unmanned aerial vehicle (UAV) with simultaneous video stream analysis imposes excessive cognitive load on the operator, leading to control and observation errors. Additionally, disturbances such as forested areas, urban infrastructure, and complex terrain hinder reliable object surveillance. This paper proposes a system in which the operator provides an initial indication of the object of interest, after which the visual tracking system autonomously maintains target tracking and generates control commands without further human intervention. All computations are performed on an onboard computer additionally installed on the UAV, which does not require communication with a ground station. To ensure compatibility with the output of the PID controller, an algorithm has been developed that allows the use of a standard flight controller without modifying its internal control loops. Experimental validation of the proposed solution was conducted in a custom-developed simulation environment based on Unity. PID controller coefficients were tuned empirically, and the configuration of the CSRT tracker was selected for operation with a low-contrast object with poorly defined boundaries. For the final system evaluation, a satellite antenna mounted on a communication tower was used as the object of interest. The tracking accuracy achieved was 0.05% of the screen diagonal over the entire tracking sequence and 0.007% of the screen diagonal during the steady-state phase, when the center of the bounding box coincides with the screen center, at a display resolution of 720×480 and a processing rate of 45.5 FPS on an Intel Core i7-11700F computing platform

Methods of collecting information using unmanned aerial vehicles for monitoring difficult and remote areas

lebedev-ob — 2026-10-05

This study presents a new approach to modeling the study area, which involves discretizing its space into rectangular segments. A method for
optimizing the coverage of the study area by unmanned aerial vehicle sensors by forming a minimal set of such rectangular segments is described. The geometric parameters (size, orientation, location) of the segments are taken into account to reduce the overall survey time. A typical unmanned aerial vehicle route is considered as a sequence of straight segments defined by control points. An algorithm for constructing an optimal unmanned aerial vehicle route, based on the ant colony principle, has been developed, along with controls for its movement along a given path. An unmanned aerial vehicle route consists of segments that can be classified as «active» (for scanning the area) and «idle» (for moving over areas not subject to survey). In most cases, such a route consists of alternating active and idle segments. A methodology and algorithm for unmanned aerial vehicle movement between control points corresponding to specified route segments are proposed. The process of finding the optimal solution is implemented using two interconnected algorithms: an ant colony behavior model and a control agent behavior algorithm. The applied control method involves continuous replanning of the unmanned aerial vehicle trajectory in real time, taking into account its parallel movement relative to the vector that specifies the direction.

CGAN model for building layout plans generation

Dmitriy Aleksandrovich Kostin — 2026-10-05

The task of layout planning is a fundamental challenge in developing site masterplans, governed by spatial and architectural constraints. Automating plan generation can greatly reduce the effort required to produce design alternatives. While shape grammars and genetic algorithms have been used for this purpose, they require manual adjustment of transformation rules and optimization criteria for each new case. In contrast, deep learning methods can identify patterns and generate designs based on existing developments. Conditional Generative Adversarial Networks (CGANs) allow constraints to be incorporated by using label maps as input. In this work, the Pix2PixHD model – featuring one generator and two discriminators – is applied to the problem. Building layout plans are generated for sites in Vladivostok, Saint Petersburg, Sochi, Seoul, and San Francisco, and the results are compared using numerical image similarity metrics. The primary metrics are Fréchet Inception Distance (FID) and Structural Similarity Index (SSIM), chosen for their ability to compare distributions. The key advantages and limitations of the model are discussed. The paper concludes with an assessment of the applicability of this deep learning approach to solving the layout planning problem in architectural design. Directions for future research have been identified, including the use of hybrid models as well as addressing the need for training on giant datasets.

Hybrid Forest Mapping for Robotic Navigation under GNSS Uncertainty

petukhovigor — 2026-10-05

This paper examines the problem of constructing a navigation-oriented spatial model of a forested area under conditions of limited satellite navigation accuracy. In dense forest stands, GNSS observations are subject to significant errors, which distorts the spatial structure and reduces the distinguishability of objects. A hybrid method is proposed based on the integration of global coordinates obtained using GNSS devices and local geometric measurements of distances and directions. The method incorporates polar measurements relative to control points, as well as inter-object distances, forming a graph model of the spatial structure. Coordinate reconstruction is performed within the framework of a weighted optimization problem, taking into account the heterogeneity of the initial data accuracy. Additionally, the model incorporates contextual environmental factors, including undergrowth, windfall zones, and topography, which affect the observability of objects and the quality of measurements. This improves the applicability of the method in real-world operating conditions of forestry machines and robotic systems. The proposed approach is shown to restore a geometrically consistent site structure, improve object distinguishability, and be robust to GNSS observation errors. The method is implemented using readily available measurement tools and can be applied to both field mapping and forestry machine automation systems.

Testing Methodology for a Computer Vision System for Glass Container Quality Control on a Robotic Conveyor

Unknown — 2026-10-05

This paper presents a comprehensive testing methodology for a prototype cascaded computer vision system deployed on a robotic conveyor mockup for automated quality control of glass containers. The work extends previously proposed approaches to synthetic training data generation by shifting the focus to empirical validation of deep learning algorithms on real product samples using low‑power edge platforms. The physical experimental setup is implemented on a laboratory belt conveyor equipped with a Raspberry Pi 5 single‑board computer, a Hailo‑8L NPU accelerator, a Raspberry Pi AI Camera (IMX500) sensor and a programmable LED ring light. A real COCO‑compatible dataset with key classes of industrial defects is collected and a structured test protocol is developed. The system is evaluated in terms of detection accuracy, inference performance, thermal stability under long‑term load, and robustness to motion blur and illumination changes. Experimental results confirm that the proposed architecture achieves industrial‑level inspection performance on edge devices: mAP reaches 0.95, the false positive rate does not exceed 5%, and the processing speed consistently exceeds 30 FPS. This demonstrates the feasibility of integrating such hardware–software complexes directly into real‑time production conveyor lines.

Effects of Simulator Training Duration on Early-Stage Control Performance of Mobile Vehicle Platforms

petukhovigor — 2026-10-05

This paper examines the impact of simulator training duration on the performance of mobile transport platform (MTP) operators in the early stages of skill development. The primary hypothesis is that the learning effect is uneven, structurally localized, and manifests itself at specific stages of task performance. An experimental study was conducted involving operators with no practical experience who underwent simulator training for 15 and 30 minutes under identical conditions. A remotely controlled mobile platform (RCM) implemented as an unmanned aerial platform (UAP) was used as a representative case of MTP control, ensuring accurate recording of trajectory parameters. The experimental task involved driving a standardized route with checkpoints, divided into time segments. To analyze structural effects, a relative segment effect metric was introduced, allowing for quantitative assessment of differences between training conditions at the level of individual trajectory segments. In addition to the overall task completion time, segmental characteristics, the number of incidents, and missed checkpoints were analyzed using nonparametric statistical methods. The results showed that increasing training duration leads to a statistically significant reduction in overall task completion time. However, the effect is heterogeneous: significant differences are localized in specific trajectory segments characterized by increased control complexity. High variability in performance was also revealed in the early stages of training, along with a lack of a strong relationship between execution speed and control accuracy. These results demonstrate that assessing the effectiveness of operator training should consider not only integral indicators but also the task structure. Although the experiment was conducted on an aerial platform, the proposed approach is applicable to a wide range of mobile transport platforms.

Adaptive Nonlinear Hierarchical Control Based on RBF Neural Networks for Line-Following and Obstacle-Avoiding Mobile Robots under Dynamic Model Uncertainties

xuanchiem — 2026-10-05

This paper proposes an adaptive nonlinear neural-network-based control law for a line-following and obstacle-avoidance robot under dynamic model uncertainties. First, the mathematical model of the line-following and obstacle-avoidance robot (LF&OAR) is unified with the conventional line-following model by employing LiDAR and camera sensors to construct a virtual line along the obstacle boundary and reconnect it with the main line, while the robot dynamics are represented in a general form. The proposed controller is developed based on a hierarchical control structure. At the kinematic level, the desired linear and angular velocities are generated using the Lyapunov approach to ensure smooth line-tracking performance, even when the reference path has large curvature. At the dynamic level, a controller is designed based on the general dynamic model of a differential-drive mobile robot using the super-twisting sliding mode control (STSMC) technique. The unknown nonlinear functions in the general dynamic model are approximated by radial basis function neural networks (RBFNNs), and an adaptive law is derived to update the network weights online. Simulation results and comparisons with a conventional super-twisting sliding mode controller demonstrate the effectiveness of the proposed method in terms of fast convergence and small tracking errors. Furthermore, experimental results confirm the feasibility of implementing the proposed control law on a real robotic platform.

AI-Driven Trichome Quantification for Robotic Phenotyping Systems

Unknown — 2026-10-05

This paper addresses the challenge of accurately quantifying basil (Ocimum basilicum) leaf glandular trichomes in the context of rapidly expanding urban vertical farming, where optimized monitoring of plant traits is essential for improving essential-oil yield and quality. To overcome the limitations of manual counting, we developed an automated approach using deep-learning object detection. Basil plants of the “Queen of Sheba” cultivar were grown under controlled LED lighting with varying spectral compositions, and microscope leaf images were acquired, annotated, and used to train YOLO models. Automated trichome detection revealed significant differences between their quantities on adaxial and abaxial leaf surfaces and demonstrated that different lighting treatments resulted in various trichome quantities. Moreover, correlations between essential oil yield and trichome quantity can be observed. Deep learning enabled fast, objective, and reproducible quantification of basil microstructures. The developed method represents a promising tool for integrating AI-based phenotyping into city farm monitoring and robotized automatization systems, supporting the optimization of cultivation parameters and receiving of high-yielding plants.

A Cloud-Edge Collaborative Framework for Personnel Identification via Wearable Smart Glasses

Hao Chen — 2026-10-05

This paper presents a cloud-edge collaborative framework for intelligent personnel identification and crowd sensing based on wearable AI smart glasses in complex environments. To overcome the computational and power constraints of edge devices, the system adopts a cloud-edge architecture where the smart glasses perform lightweight face detection, adaptive ROI cropping, and quality-aware compression, while the cloud server executes high-dimensional feature extraction and identity matching. For identity verification, a temporal multi-frame weighted smoothing strategy integrates cosine similarity scores across consecutive frames, achieving an overall accuracy of 96.1% in experiments. For database-free scenarios, an identity-agnostic crowd sensing pipeline combines human keypoint-based frontal orientation filtering, appearance-based semantic role discrimination for staff exclusion, and a cross-frame tracking mechanism fusing person re-identification features with spatial kinematic association, attaining a pedestrian recognition accuracy of 97.8% and achieving complete staff filtering in the tested controlled scenarios. Experimental results on real-world hardware demonstrate that the proposed framework effectively balances real-time responsiveness, analytical accuracy, and privacy preservation, providing a scalable solution for intelligent personnel management in complex dynamic environments.

Adaptive Energy-Shaping Error-Dynamics Control for Omni-Mecanum Mobile Robots

hoangduclonglqdtu — 2026-10-05

This paper presents a novel adaptive energy-shaping control framework for trajectory tracking of omni-mecanum-wheeled mobile robots operating under significant parametric uncertainties and external disturbances. By exploiting the intrinsic mechanical structure of the system, the tracking error dynamics are reformulated into a strictly dissipative energy-based representation, enabling systematic controller synthesis with clear physical interpretation. The proposed approach constructs the control law directly in Cartesian space and integrates adaptive mechanisms to compensate for unknown inertial, frictional, and load-related parameters. Rigorous Lyapunov-based analysis guarantees global asymptotic tracking in the nominal case and uniform ultimate boundedness of all closed-loop signals in the presence of uncertainties. Comparative simulations on the full nonlinear robot model demonstrate that the proposed controller achieves faster convergence, reduced overshoot, and smoother control effort compared with conventional adaptive backstepping and sliding-mode methods. These results highlight the effectiveness, robustness, and practical applicability of the proposed strategy for high-precision motion control of omnidirectional mobile robots in uncertain environments.

Interface for controlling a group of unmanned aerial vehicles, taking into account the operator's characteristics

boraplyton — 2026-10-05

This article examines the problem of controlling a swarm of unmanned aerial vehicles (UAVs) operating in a changing environment, with degraded communication quality and incomplete a priori information about the current situation, potentially leading to communication loss. It is shown that requirements must be imposed on the swarm of UAVs themselves, on the operator who controls the swarm and monitors its movements, as well as on the types of interfaces and their qualitative characteristics. A method for studying operator capabilities is proposed, based on determining the time characteristics of switching operating modes depending on observed phenomena and control capabilities. As part of the study, various methods of information processing are identified and their quality assessed. Based on these characteristics, the feasibility of controlling a swarm of UAVs and preventing loss of control over the swarm is assessed. To prevent false alarms, a justification for the most important characteristics is proposed. Modeling results show that the proposed approach reduces the impact of possible distractions to maintain control over the swarm of UAVs.

A Fixed-Time Adaptive Barrier-Based Integral Super-Twisting Sliding Mode Control for Uncertain Wheeled Mobile Robots

hoangduclonglqdtu — 2026-10-05

This paper proposes a novel fixed-time adaptive barrier-based integral super-twisting sliding mode control strategy for uncertain wheeled mobile robots subject to wheel slip, skidding, and external disturbances. Unlike existing prescribed-performance sliding mode approaches, the proposed controller integrates a barrier Lyapunov function with an adaptive integral super-twisting mechanism to ensure strict state constraint satisfaction while eliminating the need for predefined performance bounds. The controller guarantees that tracking errors evolve within dynamically shrinking safe sets and converge to zero within a fixed time independent of initial conditions. A new nonlinear sliding manifold is developed by combining barrier transformation and fractional integral dynamics, which improves transient performance and robustness. An adaptive disturbance compensation law is incorporated to estimate unknown uncertainties without requiring prior knowledge of their bounds. Rigorous stability analysis is conducted using Lyapunov theory, and new sufficient conditions are derived to guarantee fixed-time convergence and boundedness of all signals. Theoretical results demonstrate that the proposed control scheme achieves faster convergence, reduced chattering, and improved robustness compared to conventional sliding mode, non-singular fast terminal sliding mode, and fixed-time prescribed performance controllers. Simulation studies involving straight-line and U-shaped trajectory tracking validate the effectiveness and superiority of the proposed method under severe disturbances and wheel slip conditions.

A Robust RTK Navigation and Positioning Method for a Quadruped Robot in Complex Dynamic Environments

Wei Zhang — 2026-10-05

A quadruped robot requires continuous, stable, and high-precision positioning for autonomous navigation in complex dynamic environments. However, its onboard GNSS observations are easily affected by tree canopy blockage, near-ground multipath, and endogenous electromagnetic interference from the platform. As a result, traditional dual-frequency RTK methods often suffer from long float periods, unstable ambiguity fixing, and poor trajectory continuity in complex scenes. To address this problem, this paper proposes a robust RTK positioning method that integrates wide-lane guidance constraints with a progressive Gaussian filtering update mechanism. The proposed approach follows a procedure of wide-lane guidance, progressive float estimation, and narrow-lane stepwise fixing. Based on dynamic experimental data collected from a quadruped robot in a campus environment, this paper compares the proposed AGF-WLL1 method with the conventional EKF-L1L2 method. The results show that the proposed method reduces the horizontal and vertical errors from 0.221 m and 0.319 m to 0.128 m and 0.178 m, respectively. The narrow-lane fixed-solution ratio increases from 67.054% to 77.286%, and the float-solution ratio decreases to 16.836%. The experimental results demonstrate that the proposed method enhances fixed-solution stability, trajectory continuity, and positioning robustness for quadruped robot navigation in environments with blockage and interference.

Robust Geometric Filtering via Linear Constraint

yuqiangjin — 2026-10-05

Geometric filters are known to be highly sensitive to model mismatch — a critical limitation in real-world applications. To address this issue, this paper introduces a novel filtering framework grounded in linear constraint enforcement. We begin by analyzing how model mismatch propagates through the filtering process in group-affine systems, revealing the fundamental mechanism behind error accumulation in conventional geometric filters such as the invariant extended Kalman filter (InEKF). Building on this insight, we design a constrained filtering strategy that actively regulates the gain matrix along the direction of model deviation, thereby suppressing error growth while preserving the intrinsic geometric structure of the underlying Lie group manifold. Notably, the proposed method achieves this enhanced robustness without sacrificing computational efficiency, making it suitable for real-time deployment. Finally, simulations and real-world experiments on INS/GNSS integrated navigation demonstrate that the proposed method consistently achieves substantially improved estimation accuracy compared to the standard InEKF, offering an effective solution for robust filtering in complex environments.

Stabilization System Quality in Synthesized Optimal Control

Elizaveta Shmalko — 2026-10-05

Stability is crucial for practical implementation of computed optimal controls onboard, as it compensates for discrepancies that arises, as a result, of
objectively existing differences in the models used during control design and the real objects. The synthesized control approach solves optimal control problems by strategically placing stable equilibrium points in state space, activated at equal time intervals. However, stability alone proves insufficient for maintaining optimal performance; equilibrium points must also exhibit specified quality characteristics to minimize sensitivity to external disturbances and model uncertainties.
Stable equilibria can be classified as aperiodic or oscillatory. Near oscillatory points, the system's differential equations fail to ensure strict contractivity in state space mappings, potentially causing unpredictable error growth between the actual object state and its model-based estimate. Thus, oscillatory equilibria are undesirable for control synthesis.
This paper imposes additional quality requirements on the stabilization dynamics approaching equilibrium points namely to the object’s movement to the stability point. A numerical approximation method is proposed for synthesizing stabilization systems meeting these criteria. Optimal trajectories are first generated around the target equilibrium for various initial conditions via optimal control solutions. Symbolic machine learning then serves as a universal function approximator, trained on these supervised trajectories to minimize approximation error across the trajectory set.
A computational experiment demonstrates stabilization synthesis for a dif-ferential-drive mobile robot using symbolic regression, a form of evolution-ary machine learning. Results confirm that high quality stabilization system is little sensitive to disturbances.

Monocular Vision-Based Positioning System for Object Localization in Collaborative Conveyor Applications

Elvira Chebotareva — 2026-10-05

This paper presents a monocular vision-based relative localization system for a collaborative robotic cell comprising a mobile conveyor and an industrial robot. The main challenge is that movement of the conveyor and/or the robot may disrupt traditional camera-robot calibration, which requires time-consuming manual readjustment. The proposed solution enables a camera, statically mounted on the conveyor frame, to detect a calibration ArUco marker attached to the robot base with known coordinates, compute the camera's position relative to the robot, and then determine the coordinates of a target object on the conveyor within the robot base coordinate system. The key advantage of this approach is that calibration is performed online in each frame, allowing the conveyor and the robot to move independently without losing positioning accuracy. A mathematical model of coordinate transformation, including rotation matrices and homogeneous transformations, is developed. The system is integrated with a real KUKA KR3 robot. The experiments demonstrated a mean absolute error in the XY plane of 2.4 mm for the X-axis and 3.1 mm for the Y-axis, which is acceptable for most object grasping tasks. These findings confirm the system's functionality and suitability for flexible manufacturing cells with frequently changing layouts.

Analysis of Deep Learning Methods in Video Analytics, Robotic Monitoring, and Aquaculture Management with an Empirical Comparison of Lightweight YOLO Models for Fish Detection

Nghia Le Van — 2026-10-05

This paper presents a review of state-of-the-art deep learning methods for video analytics in aquaculture over the period 2025–2026, complemented by an empirical evaluation of two lightweight YOLO-family neural network models for the fish detection task. Four key research directions are analyzed: fish disease detection, segmentation and phenotyping, tracking and biomass estimation, and the development of lightweight models. Based on contradictions identified between existing studies, an experimental comparison of two architectural configurations - the original YOLOv26 and YOLOv11 integrated with the Reparameterized Multi-Head attention module (RepMHD) - was conducted on the public DeepFish dataset, which contains images of fish in natural tropical marine habitats. The results show that the YOLOv11 + RepMHD configuration achieves higher Precision (0.978 vs. 0.960) and a comparable mAP@0.5 value (0.991 vs. 0.990) with 2.28 million parameters and 5.6 GFLOPs, whereas the original YOLOv26 demonstrates a higher mAP@0.5:0.95 value (0.816 vs. 0.812) with 2.38 million parameters and 5.2 GFLOPs. These findings indicate that the choice between these two strategies should be guided by the specifics of the application - particularly the priority assigned to minimizing false positive versus false negative detections.

Monocular Depth-Based Object Localization for Collaborative Robotic Grasping

Elvira Chebotareva — 2026-10-05

This paper investigates the application of monocular depth estimation for 3D object localization in collaborative robotic manipulation tasks. We propose a method for converting relative depth maps obtained using the Depth Anything V2 Small model into metric depth maps by employing small 5×5 cells ArUco markers with a physical size of 3×3 cm, which do not obstruct the operator or reduce the available workspace. Pre-calibration on 10–30 images achieves a diameter measurement accuracy of mean absolute error (MAE) of 0.307-0.671 cm (relative error 3.6-7.9%) and height measurement accuracy of a 1.73–2.13 cm MAE (relative error 14.6-16.7%). The full processing cycle for a single frame takes 1.19 seconds on a mid‑range laptop without a dedicated GPU, which confirms the method’s suitability for operation under modest frame rate requirements. The proposed approach is applicable in collaborative scenarios with low‑speed object motion and can be recommended for robots equipped only with an RGB camera, provided that pre‑calibration is performed. The method is characterized by its ease of implementation, low cost, and the absence of a need for specialized hardware, making it attractive for a wide range of human-robot interaction tasks in constrained spaces.

Robust and Effective Semantic-aided Point Cloud Matching for Mobile Robot Navigation

kirill_mouraviev — 2026-10-05

Navigation of a mobile robot is a challenging task which requires precise localization of the robot to plan routes and follow them. Localization refinement is commonly done by matching data from the robot's LiDAR sensors and corresponding area of the map. Conventional geometry-based methods of point cloud matching may fail with shapeless objects (e.g. trees) or with weakly overlapped scans. Addition of semantic features to geometric ones may help to address this issue. However, semantic segmentation models may produce segmentation errors, so additional techniques are required to handle these errors while matching. In this paper, we propose a robust and computationally efficient point cloud matching method based on 2D projection of semantic data and iterative filtering of extracted features. Usage of 2D projections makes our method memory- and computationally-effective, whereas clusterization and integration of RANSAC technique keeps robustness to segmentation errors. We test our approach in a simulated indoor environment and on real outdoor data, comparing it with state-of-the-art semantic-free point cloud matching methods. The experimental results show that our approach outperforms competitors on both memory consumption and matching precision, and works with proper speed for real-time operation.

Autonomous Service Platform for Spraying UAVs: System for Mixing Solutions

Andrey Leonidovich Ronzhin — 2026-10-05

Maintenance of agricultural spraying drones (UAVs) currently requires manual dosing of highly concentrated and toxic products (with dilution ratios of 1:100 and higher), which is not always performed with sufficient accuracy. The developed multi-channel modular system for mixing working solutions is designed for the automated preparation of liquids and refueling of UAVs. The developed system ensures versatility due to the ability to prepare solutions in a wide range of required dilution ratios from 1:10 to 1:1000 and higher while maintaining high dosing accuracy. A fundamental distinction of the developed system is the inclusion of dosing modules of two different designs, each optimized for a specific range of dosing volumes and concentrate dilution ratios. The system fully automates the preparation and refueling process, eliminating the need for the operator to be constantly present in the hazardous area near operating UAVs and minimizing their exposure to highly concentrated toxic substances. The system is designed for use as part of mobile or stationary autonomous service platforms for servicing UAV fleets.

Lightweight Multi-Task Emotion Recognition for Human-Robot Interaction

Alexey Dmitrievich Yakushev — 2026-10-05

Abstract. We address the problem of real-time emotion perception on resource-

limited robotic platforms. Instead of pursuing marginal accuracy gains with

computationally heavy architectures, we investigate the performance limits of a

streamlined multi-task setup. The proposed method performs 26-way emotion

classification alongside continuous Valence, Arousal, and Dominance (VAD)

regression using a shared MobileNetV3-Large backbone. To prevent the

classification head from being biased by the heavily skewed label distribution of

the EMOTIC dataset, we apply Asymmetric Loss (ASL). Furthermore,

decoupling the feature projections for discrete and continuous outputs mitigates

gradient interference between the two tasks. On the EMOTIC test split, the model

achieved 25.33% mean average precision for classification and a mean error rate

of 0.9146 for VAD prediction. Given the fine-grained, context-dependent nature

of the labels, this provides a robust baseline for edge-oriented models. On the

deployment side, after ONNX conversion and OpenVINO optimization, CPU

inference requires only 3.29 ms per frame (over 300 FPS). This leaves substantial

computational headroom for concurrent robot control loops. The results

demonstrate that massive models are not strictly necessary to extract reliable

affective signals on edge devices.

Path Planning with Motion Primitives in Dynamic Environments: SIPP on Lattices

Marat Antonovich Agranovskiy — 2026-10-05

Autonomous navigation in dynamic environments is a critical challenge, particularly when spaces are shared with other mobile agents whose future trajectories are known. While traditional grid-based planners efficiently find collision-free paths, their reliance on stop-and-turn mechanics over $2^k$-connected grids produces piecewise-linear trajectories that are kinodynamically highly sub-optimal for differentially constrained robots. In this paper, we present an adaptation of Safe Interval Path Planning (SIPP) that operates on state lattices, utilizing precomputed, kinodynamically smooth motion primitives. To efficiently handle dynamic environments, we rasterize the spatiotemporal swept volumes of moving obstacles directly onto the grid, treating grid cells as atomic units of space, whose resolution is typically dictated by inherent localization noise. We perform a comprehensive comparative analysis between our lattice-based approach and $2^k$-connected grid planners across diverse topological environments. Our evaluation considers a broad spectrum of performance metrics, including planning time, path angularity, cumulative heading change (angle-over-length), and bending energy. The results demonstrate that while the expanded state space of lattice-based search increases computational overhead, it yields trajectories with significantly superior kinodynamic properties. Specifically, our method achieves a reachability comparable to highly connected grids while ensuring smooth, continuous, and physically executable paths ready for real-world deployment.

Modeling the Dynamics of a Collaborative Parallel Kinematic Robot with Elastic Actuators for Automated Periorbital Massage

Unknown — 2026-10-05

This paper presents the development of a dynamic model of a collaborative robotic system based on parallel kinematics with series elastic actuators (SEA) for automated periorbital massage. The proposed system is intended for rehabilitation and therapeutic applications, including meibomian gland dysfunction treatment and postoperative recovery. A mathematical model of the delta manipulator dynamics is developed, incorporating actuator dynamics, SEA elasticity, and contact interaction with soft tissues. Three control strategies are compared through computational simulation: rigid positioning (high-stiffness SEA), passive compliance (soft SEA without force feedback), and active compliance (hybrid position–force control with trajectory correction based on contact force measurement). The simulation results demonstrate that the hybrid controller significantly reduces peak contact force and force tracking error compared to both rigid and passive compliance modes, while maintaining acceptable position accuracy for periorbital massage. The proposed active compliance approach effectively limits interaction force to a specified safe level, ensuring patient safety even under large-amplitude motions. The developed model provides a foundation for designing SEA-based collaborative robots for delicate facial massage procedures.

System Analysis of Video Stream Coherence in the Design of Hybrid Behavior Recognition Models

Ilya Ivanovich Kovalenko — 2026-10-05

Designing video analytics systems under small-data conditions requires accounting not only for the target recognition task but also for the acquisition regime of the sensor. In this study, a two-stream CNN-ViT model with time-color coding is used as an experimental instrument to analyze two contrasting observation conditions: a static camera, where spatial and temporal representations remain mutually consistent, and active tracking, where camera stabilization introduces decoherence between these streams. The experiments show an inversion of generalization behavior across these regimes. In the coherent setting, the higher-capacity configuration provides the best recognition performance. In the decoherent setting, reducing model capacity by 62.5% prevents collapse of generalization and improves accuracy by 61.2 percentage points. In addition, removing a low-consistency class from the tracking domain, which reduces the dataset by 15%, further improves system performance by 6.9 percentage points. These results indicate that, under small-data conditions, the admissible capacity of the computational core should be selected with respect to sensor behavior and the consistency of the observed motion patterns rather than from task complexity alone.

Towards Latency-Stable Robotic Perception Pipelines in ROS 2: Eliminating Non-Maximum Suppression for Real-Time Target Tracking and Distance Estimation

newyork00 — 2026-10-05

Visual perception is a fundamental component of autonomous
navigation in mobile robotics. However, for real-time control loops such
as Proportional-Integral-Derivative (PID) or Model Predictive Control
(MPC) systems, temporal determinism is as critical as spatial accuracy.
Traditional object detectors rely on Non-Maximum Suppression (NMS),
a post-processing stage whose execution time depends on scene complex-
ity. This dependency introduces stochastic inference latency, commonly
referred to as latency jitter. Such temporal instability propagates through
the perception pipeline, degrading multi-object tracking consistency and
significantly impairing monocular distance estimation, ultimately leading
to unstable robotic control behavior.
This paper proposes a latency-stable perception pipeline implemented in
Robot Operating System 2 (ROS 2) that fundamentally eliminates the
NMS bottleneck. By integrating an NMS-free detector (YOLOv10)
with the ByteTrack tracking algorithm inside an asynchronous multi-
threaded ROS 2 node, computational workloads are isolated from com-
munication callbacks, preventing head-of-line blocking and artificial la-
tency variance.
Experimental validation conducted in a dynamic Gazebo simulation envi-
ronment demonstrates a 78.2% reduction in latency standard deviation
(from 18.22 ms to 3.97 ms) compared to a classical baseline pipeline
(YOLOv8 + SORT). The improved temporal stability directly
enhances geometric consistency: the root mean square error (RMSE) of
monocular distance estimation decreases from 1.710 m to 1.449 m, while
the proposed system maintains errors below 1.5 m in 61.3% of frames
versus 26.1% for the baseline. The results confirm that eliminating NMS
is a key requirement for reliable deployment of vision-based perception
systems in real-time robotic platforms.

HELP: Hierarchical Embodied Language Planner for Household Tasks

ALEXANDR V. KORCHEMNYI — 2026-10-05

Embodied agents tasked with complex scenarios, whether in real or simulated environments, rely heavily on robust planning capabilities. When instructions are formulated in natural language, large language models (LLMs) equipped with extensive linguistic knowledge can play this role. However, to effectively exploit the ability of such models to handle linguistic ambiguity, to retrieve information from the environment, and to be based on the available skills of an agent, an appropriate architecture must be designed. We propose a Hierarchical Embodied Language Planner, called HELP, consisting of a set of LLM-based agents, each dedicated to solving a different subtask. We evaluate HELP on a household task and perform real-world experiments with an embodied agent. We also focus on the use of open source LLMs with a relatively small number of parameters, to enable autonomous deployment.

Multi-Agent Path Finding in Multi-Floor Warehouses with Elevators

mishgun — 2026-10-05

This work studies multi-robot navigation in multi-floor facilities equipped with elevators. The baseline mathematical model is Multi-Agent Path Finding (MAPF) in a discrete space-time environment. Unlike the classical single-floor MAPF setting, multi-floor warehouses include vertical transport re- sources (elevators) with limited throughput and nonzero travel time between floors. In the existing literature, elevator-control tasks and MAPF tasks are usu- ally considered separately: elevator papers do not account for detailed coordina- tion of many mobile agents, while MAPF papers assume single-level geometry or simplified vertical transitions. The goal of this study is to formalize MAPF for multi-floor environments with elevators, and to adapt and evaluate existing co- operative planning algorithms under elevator-specific constraints.

An Efficient Navigation System for Autonomous Mobile Robots on Rough Terrain Environments

Xuan-Tung Truong — 2026-10-05

This paper presents an efficient navigation framework for autonomous mobile robots operating in rough terrain environments. The proposed approach integrates terrain perception, traversability analysis, and motion planning to enable safe and reliable navigation in complex off-road conditions. An Extended Kalman Filter (EKF) is employed to enhance robot localization, providing robust pose estimation in noisy and unstructured environments. A 2.5D elevation grid map is constructed to represent the local terrain geometry, from which a traversability cost function is formulated to evaluate terrain feasibility. To address global path planning, a Traversability-A* algorithm is proposed by extending the classical A* method with terrain-aware cost integration, allowing the planner to avoid hazardous regions. For local trajectory optimization, a Timed Elastic Band (TEB) planner is adopted to ensure kinematic feasibility and dynamic consistency. The proposed framework is implemented on a skid-steering mobile robot and validated in Gazebo-based simulation environments with uneven terrain. The results demonstrate improved navigation safety and path quality compared to conventional methods.

An Efficient Camera-LiDAR-Based 3D Human Detection and Tracking Algorithm for Socially Aware Mobile Robot Navigation Systems

Xuan-Tung Truong — 2026-10-05

Reliable real-time perception of surrounding humans is essential for mobile robots operating in dynamic human-populated environments. However, maintaining stable human identities and accurate 3D positions in real time remains challenging, especially under sensor noise, partial occlusion, and onboard computational constraints. This paper presents a camera–LiDAR-based framework for 3D human detection and tracking in socially aware mobile robot navigation. The method combines RGB images from an Azure Kinect camera with 3D point clouds from a Velodyne LiDAR. Humans are first detected and tracked in the image domain using YOLOv8 and BoT-SORT, which provide 2D bounding boxes and persistent identity labels. LiDAR points are then transformed into the camera frame, projected onto the image plane, and associated with the tracked detections. To reduce the influence of background points, Euclidean clustering is applied to the associated point sets, and the selected clusters are used to estimate human positions in 3D space. A Kalman filter is further maintained for each tracked individual to smooth the estimated positions and improve temporal consistency. The resulting 3D human states are integrated into a ROS-based mobile robot navigation framework, allowing the robot to perceive nearby humans and adjust its motion accordingly. Experiments conducted on a real mobile robot platform in several indoor scenarios show that the system provides stable 3D human tracking, supporting its practical use in socially aware navigation tasks.

Control of Mobile Robots with Load-Bearing Supports Discretely Interacting with the Supporting Surface

bordyugov_denis — 2026-10-05

A schematic of a two-section mobile robot of the «robo-train» type is presented. The robot is equipped with braking supports that interact discretely with a flat surface and with a linear actuator. Its motion is achieved by extending or retracting the rod of the linear actuator: one section remains stationary due to the interaction of all its deployed supports with the supporting surface, while the other section is in a free state and performs rotational motion about a vertical axis passing through a single fixed support. The motion principle of the two-section robot is based on the alternation of fixation and displacement phases of the sections. This enables the «robo-train» to follow a prescribed trajectory with minor deviations and to achieve high maneuverability. It is established that the motion pattern is determined by the geometric parameters of the system, the operating modes of the linear actuator, and the control algorithm governing the interaction of the deployable braking supports with the supporting surface. To evaluate the motion performance of the «robo-train», the concept of «dimensional passability» is introduced, which refers to the robot’s ability to move within narrow curvilinear regions bounded by rigid walls. The design and operating principle of a mobile robot with deployable supports and a rotating internal body are also considered; such a robot can be incorporated into the «robo-train» as a leading section. Basic motion trajectories of this robot are presented, the combination of which enables motion in various directions and allows different operating modes of the two-section robot to be realized.

An Adaptive State Observer for a Class of Time-variant Bilinear Systems Under Condition of Partial Parametric Uncertainty

Olga Zarina — 2026-10-05

An adaptive observer for a bilinear time-varying dynamic system under condition of partial parametric uncertainty is proposed. The problem is solved under the assumption that unknown parameters are contained in the control matrix/vector. The state matrix of the system is assumed to be of a certain form. The relevance of such assumption is demonstrated via example of the real technical system (robotic leg) described by the Euler-Lagrange equation. The key idea of the proposed algorithm is a new parameterization of the object. Two functions are defined based on the state vector of the matrix. The crucial point of this algorithm is that one of these functions can be found analytically using known and measured system signals. The proper solution for this function allows to eliminate the nonlinear part of the equation. After that application of linear filters transforms the initial system to a linear static regression model containing unknown constant parameters. The unknown parameters are estimated using a gradient algorithm, and estimates of unknown state vector components are calculated based on the obtained estimates. The proposed method is applicable to a wide class of bilinear systems, including Euler-Lagrange systems describing many real technical objects and robotic systems.

ROBOTICS IN ENVIRONMENTAL MONITORING OF SURFACE WATER BODIES

vladmal — 2026-10-05

Current real-world events are exposing a critical weakness in our standard operational frameworks. Against the backdrop of escalating hybrid conflicts, the biological protection of water resources demands a radical overhaul of our fundamental containment strategies. The traditional, stationary model of sanitary surveillance has proven to be systemically rigid and immobile. The destruction of the Kakhovka Hydroelectric Dam and the large-scale flooding in the Orenburg region provided harsh empirical evidence of the total paralysis of fixed laboratory networks in disaster zones. In such scenarios, the time factor becomes fatal. The resulting diagnostic vacuum - where no testing can occur- transforms water into an ideal vector for the mass transmission of pathogens, thereby necessitating a decisive paradigm shift from the traditional method of bringing the sample to the instrument to the urgent delivery of the instrument directly to the contamination site. Targets and methods. Development, validation, and field testing of the "VirBac-1" robotic universal mobile complex was the main goal of research group. A primary technical hurdle was overcoming the physical barrier known as the "size paradox," which refers to the difficulty of detecting extremely small viral particles in large water volumes using traditional filters that either clog instantly or fail to trap the target. To solve this, the authors deliberately abandoned the classic sieve effect in favor of electrostatic adsorption on modified nitrocellulose membranes; in simpler terms, the system uses a specialized filter with an electric charge that acts like a magnet for viral particles, trapping them regardless of their physical size. The architecture of the complex integrates this colloid-chemical principle into an automated membrane concentration cycle operating under strictly controlled pH levels and ionic strength of the solution to ensure maximum capture efficiency. Results. The empirical data exceeded initial projections. The resulting cyber-physical prototype completely eliminates human error during the pre-analytical stage while ensuring full operational autonomy powered by lithium-ion batteries. Crucially, the total analytical cycle time has been radically compressed: whereas standard regulatory frameworks require forty-eight to seventy-two hours. Conclusion. The "VirBac-1" complex reduces this time to an operationally actionable window of one point two to one point five hours. In the context of a flood or a bio-terror event, a delay of two or three days renders laboratory data useless for emergency response, whereas a result delivered in ninety minutes allows for immediate public health intervention. Consequently, the "VirBac-1" robotic complex establishes a fundamentally new technological architecture for continuous environmental monitoring of water bodies. Modeling of hydrological disaster scenarios has convincingly demonstrated that the complex is ready for large-scale, in situ screening of surface waters, effectively bypassing the logistical constraints that cripple traditional laboratory systems.

Safe Use of Artificial Intelligence and Robotics: Issues of Legal Terminology

Matvei Sergeevich Chernyshov — 2026-10-05

This article addresses the issues of developing legal terminology in the field of regulatory governance of safe artificial intelligence and robotics use. The authors argue that the absence of an adequate conceptual framework constitutes a fundamental obstacle to the effective regulation of relations concerned with the safe development and deployment of AI systems. Particular attention is given to the situation in the Russian Federation, where dedicated legislation remains at the drafting stage. The article conducts a comparative analysis of legislative approaches to the definition of artificial intelligence in the different jurisdictions and classifies the identified approaches into two broad categories: the “anthropomorphic” approach and the “technical” approach. Drawing on this analysis, the authors propose an original conceptual framework of legal definitions for Russian AI legislation. A novel legal definition of “artificial intelligence” is formulated, reflecting its dual nature as both a technology and a scientific discipline. The article substantiates the need for statutory recognition of the categories of “trusted” and “prohibited” AI systems. A three-tier model of legal subjects is proposed, comprising the AI system developer, operator, and user. The authors conclude that the establishment of a sound and coherent terminological foundation is a necessary precondition for safeguarding the rights and safety of individuals in the context of advancing artificial intelligence and autonomous robotics technologies.

Differentiable LQR as a Structured Policy Output Layer for Mapless Goal-Reaching Navigation of a Mobile Robot

glebdudyrev — 2026-10-05

This paper addresses the problem of moving a mobile robot to a given goal in an unknown environment using local sensory data. A standard MLP policy trained with the PPO algorithm and directly generating linear and angular velocity commands is considered as the baseline approach. The proposed method is a policy architecture whose actor output is augmented with a differentiable linear–quadratic regulator layer. This architecture implements a two-level control generation scheme: the neural component of the policy produces a local control intention, while the embedded LQR layer transforms it into the final control action. The paper describes how this layer can be integrated into a continuous actor–critic policy without changing the environment action space and without moving the regulator outside the policy itself. Experimental evaluation is performed in ROS2/Gazebo using a TurtleBot3 robot in goal-reaching tasks in a free environment and in an environment with static obstacles. The comparison is conducted under a unified setup with identical observations, reward function, training and evaluation protocol, and metrics. The obtained results show that the proposed architecture does not degrade the robot’s basic ability to reach the goal and, in a more complex environment, produces a meaningful difference in behavior and trajectory quality compared with a standard MLP policy, which makes the embedded differentiable LQR layer a promising direction for further research.

Real-Time Stereo-Vision Semantic Navigation for Apple Orchards Using YOLOv5

Bushra Ali — 2026-10-05

Autonomous navigation in orchards is highly challenging due to dense tree canopies, variable lighting, grass, uneven terrain and frequent GNSS signal blockage. This paper presents a complete, purely stereo-vision navigation system that relies exclusively on data from a single forward-facing Intel RealSense RGB-D camera. The proposed learning-based semantic algorithm uses a YOLOv5 segmentation model to perform real-time pixel-level segmentation of the ground (road) between rows. The segmentation output is processed through a distance transform and triangle fitting to extract both heading angle and lateral offset. A hybrid steering controller with mode switching generates smooth and robust control commands while end-of-row detection is performed directly from the segmentation mask which enables fully closed-loop operation without any external positioning sensors. The system was evaluated in an apple orchard under realistic natural lighting conditions. Experimental results show a mean absolute lateral error of 0.1651 m and a root mean square error of 0.2056 m relative to the row centerline. The results indicate that accurate and stable row following can be achieved using only low-cost RGB-D data, making the proposed system a feasible navigation solution for orchard robotics.

Development of an Algorithm for Solving the Inverse Dynamics Problem for Underwater Manipulators

Alexander Protsenko — 2026-10-05

The paper discusses the development of a new method for calculating load moments for all joints of underwater multi-link manipulators installed on unmanned underwater vehicles. By introducing additional terms into the equations of motion, it takes into account all the key features of the interaction of the manipulator links with a viscous medium, including the effects of currents. To achieve this, the attached masses and moments of inertia are represented as symmetric positive-definite matrices with constant coefficients. This distinguishes the proposed approach from known methods that use variable coefficients dependent on the instantaneous velocity and direction of motion. This representation ensures the linearity of the dynamic model with respect to the identifiable hydrodynamic parameters, creating the basis for their subsequent precise determination. To account for constant underwater currents in the proposed modified algorithm, the linear velocity of the manipulator base is assumed to be equal to the difference between the vehicle speed and the current velocity, allowing for the determination of the longitudinal and transverse components of the viscous friction forces and moments. Furthermore, a correction is introduced into the expressions for the acceleration of the link's center, accounting for the constant current velocity. This corrects the calculation of the forces and moments due to the attached masses and moments of inertia. Thanks to these modifications, the proposed algorithm enables the correct modeling of the manipulator's dynamics in a real underwater environment, where the presence of currents is a significant factor influencing the formation of moments in the joints.

Method of Tracking Underwater Moving Objects by means of Unmanned Surface Vessels in Marine Environments

Irina Evgenevna Naidanova — 2026-10-05

The paper considers a method of forming and tracking trajectories of underwater moving objects (trajectory processing) by means of a small unmanned surface vessel, taking into account external disturbing effects from the marine environment. The proposed method is based on the processing of data received from a hydroacoustic station mounted on an unmanned surface vessel, and can be applied when it is not possible or impractical to deploy stationary hydroacoustic stations or when it is necessary to promptly complete a mission to track underwater objects in the water area where there is no appropriate infrastructure. Fuzzy logic methods are used to track moving objects, in particular to associate their measurements, an extended Kalman filter is used to estimate the position and velocity of underwater objects, and elementary rotation matrices are used to account for the varying orientation of a small unmanned surface vessel. In addition, this method takes into account the time delay in the propagation of the hydroacoustic signal. The simulation results confirmed the operability and effectiveness of the proposed method of tracking underwater objects.

A Communication-Efficient Modified Consensus-Based Auction Algorithm for Distributed AUV Task Allocation

Firas Houssein — 2026-10-05

The problem of target positions allocation in a formation of autonomous unmanned underwater vehicles (AUVs) under conditions of limited underwater communication is considered. A consensus-based auction algorithm (CBAA) is used to solve this problem, combining decentralized task selection and an assignment approval process between agents. A drawback of classical CBAA is the significant time required to execute multiple consensus phases, especially in an underwater environment where acoustic communication channels are characterized by low bandwidth and significant data transmission delays. A modification of the CBAA is proposed, making it need only a single consensus phase followed by local calculation of the final allocation by each agent based on the full set of bids from all participants. This approach eliminates the need for repeated consensus phases and shifts some of the workload from the communication system to the AUV's onboard computing system. A comparative analysis of the communication complexity of the original and modified algorithms was conducted. The results confirm the potential of the proposed approach for rapid target positions allocation in large groups of AUVs.

Positioning system for autonomous underwater vehicles equipped with multilink manipulators near work objects

Roman Pavlovich Vasilenko — 2026-10-05

Autonomous underwater vehicles (AUVs) are designed for efficient long-distance travel. When equipping such vehicles with multilink manipulators (MMs), it is necessary to select a vehicle position and orientation near the work object (WO) that will allow it to be stabilized and ensure the successful and safe mission execution. This paper proposes an approach to the synthesis of positioning systems for AUVs equipped with MMs near WOs during manipulation operations in the presence of underwater currents. This approach consists of the following main components: estimation of the underwater current velocity vector based on the analysis of visual navigation system data in the area near the WO; using flow simulation, the heading angles are determined at which propulsion and steering complex (PSC) capabilities are sufficient to stabilize the AUV under the effects of currents; the AUV with MM position relative to the target WO is determined, taking into account the onboard technical vision system (TVS) field of view and the boundaries of the MM workspace; simplified operation execution modeling is performed to assess the MM’s effects on the AUV; the target position and orientation of the AUV near the WO are selected based on a quadratic quality criterion.

Graphs of Convex Sets for Lifelong Manipulation Motion Planning

Milana Dmitrievna Krivova — 2026-10-05

Manipulation has great potential not only for industrial settings with predefined trajectories, but also for tasks such as human-robot interaction, where more sophisticated planning methods are required. One promising direction is planning with Graphs of Convex Sets (GCS) in the configuration space. This approach has been shown to be effective on many robotic manipulation tasks, but it still has a major bottleneck: constructing convex sets in high-dimensional setups is computationally expensive. Moreover, obtaining a sufficiently good decomposition of the whole scene in advance can be difficult and time-consuming. To this end, in this paper we focus on the online planning setup in which goal states are sent to the system iteratively. Instead of requiring a good decomposition of the entire scene beforehand, we refine and expand the graph of convex regions only when needed, as new goal states arrive. In this way, the decomposition is built incrementally rather than fully precomputed. We evaluate our approach on different planning scenes and compare it with sampling-based and optimization-based baselines in terms of planning time and trajectory quality. The results show that our method is faster while remaining competitive in trajectory quality.

Feedback Control of a Discrete System with Memory within the Class of Zonal Control Actions

Samir Zakir Guliyev — 2026-10-05

The problem of optimal control for a discrete-time dynamical system
over a finite horizon is considered under uncertainty in the initial state values and
in the external parameters acting on the system. The control is generated according
to a feedback law that depends only on the observable part of the state vector
and its history. A functional form of the feedback law is introduced on the basis
of a family of zonal controllers: the space of the observable part of the state is
partitioned into a finite number of zones, and within each zone the control is defined
by fixed affine coefficients, including a memory term, that is, a dependence
on the previous measurement. Such a parametrization reduces the original optimal
control synthesis problem to a finite-dimensional optimization problem with
respect to the coefficients of the zonal law. For the cost functional averaged over
the set of admissible initial states and external parameter values, analytical formulas
for the gradient with respect to the feedback parameters have been derived.
First-order optimality conditions are formulated, and a projected gradient algorithm
is proposed for computing the optimal zonal coefficients in the presence of
constraints. Results of numerical experiments on test problems are presented;
they illustrate the effectiveness of the proposed approach and make it possible to
compare different variants of zonal controllers.

Informed Reinforcement Learning for mobile robot planning and control tasks based on the D* algorithm

smakopyan — 2026-10-05

Autonomous navigation of mobile robots in partially unknown and dynamic environments requires control strategies that are both optimal and robust to disturbances.This paper proposes a hybrid navigation framework that integrates global path planning, Pontryagin-based optimal control, and informed reinforcement learning. The system is implemented as a hierarchical ROS2-based architecture, where D*-Lite generates a global path, an optimal control module computes nominal control actions, and a TD3-based RL agent learns corrective policies to compensate for disturbances.

Unlike conventional RL approaches, the proposed method leverages optimal control solutions as informative priors, allowing the agent to learn residual control policies instead of learning from scratch. This significantly reduces the exploration space and improves training efficiency.

Experimental results in a Gazebo simulation with dynamic obstacles and physical perturbations demonstrate improved stability, faster convergence, and better trajectory tracking compared to standalone methods.

The proposed approach provides a scalable and robust framework for integrating model-based control and data-driven learning in mobile robot navigation.

Application of Three-Dimensional Graph Modeling for the Investigation of Multipath Data Routing Algorithms in Underwater Acoustic Networks of AUVs

viktor_tiulnikov — 2026-10-05

This paper examines a class of telecommunication networks with mobile nodes, an important subclass within which is the so-called geography-aware networks. Their distinguishing feature is the availability of information about the geographic coordinates of all nodes in the network to each individual node, whereby each node is aware of the complete topology of the network graph, which enables rapid identification of the required number of information transmission routes between any two nodes. The purpose of this article is to investigate multipath routing algorithms. To this end, we propose a method for the automatic synthesis of adequate test models capable of representing networks of virtually unlimited complexity. The proposed solution is based on a compositional approach, in which a relatively simple network fragment that satisfies given constraints is first formed, and then the resulting model is constructed as a composition of copies of this fragment. To investigate the efficiency of multipath routing algorithms, we propose a compositional method for the random synthesis of test models of complex networks that satisfy constraints on distances between the vertices. This method was applied to the investigation of two routing algorithms, resulting in a large body of illustrative model data. The obtained results, presented in the form of graphs, demonstrate an increase in the gain in message transmission time as the queue length in the target flow grows. While the efficiency of multipath routing is insignificant under low network load conditions, its usefulness increases with growing network load. A similar increase in efficiency is demonstrated when transitioning from an algorithm that does not allow intersecting paths to an algorithm that allows such intersections.

Onboard Vehicle Detection for Autonomous UAV Reconnaissance under Degraded Visual Conditions

pdthong — 2026-10-05

Autonomous unmanned aerial vehicle (UAV) reconnaissance imposes three simultaneous requirements on onboard target detectors: high accuracy for small targets that frequently appear in medium-altitude imagery, robustness under realistic image degradation, and sufficient compactness for deployment on embedded hardware with strict power and latency constraints. This paper presents UAVR-YOLOv8, an enhanced YOLOv8-based detector designed to address these three requirements within a unified framework. The proposed model incorporates three main components: a P2 detection branch to better exploit shallow features for small-target detection, a Coordinate Attention module to enhance spatial feature representation, and a curriculum augmentation strategy covering motion blur, low-light, fog, sensor noise, and atmospheric haze. The model was trained and evaluated on a selected aerial image dataset consisting of UAV images containing three major target groups: main battle tanks, armored personnel carriers, and military trucks, captured under various terrain and daylight conditions. In addition to conventional evaluation metrics, this study introduces a robustness index to quantify model reliability under degraded imaging conditions, thereby extending the assessment beyond average accuracy on clean images alone. Experimental results on the NVIDIA Jetson Orin Nano show that the proposed model maintains strong detection accuracy on clean images, significantly improves performance retention under the five considered degradation types, and achieves real-time inference within the allowed power budget. The full ablation study further confirms that all three proposed modifications contribute positively, with small but consistent synergistic gains when integrated into a single design. These results indicate that an appropriate combination of proven techniques, together with evaluation under realistic operating conditions, can provide a practical detector for autonomous UAV reconnaissance deployment.

Velocity observer design via reducing to linear regression for the Euler-Lagrange systems class

Sofya Devyataya — 2026-10-05

In this paper, we consider the problem of velocity estimation for Euler-Lagrange systems. For these systems, classical observers are quite limited because, in the general case, the equations have quadratic (bilinear) dependence on generalized velocities. The proposed approach is based on the existing state observer designed for bilinear systems. This observer requires that the unknown multiplier of the state is integrable. Of course, not every arbitrary system satisfies this condition. However, Euler-Lagrange equations naturally contain partial derivatives and, what is more important, full derivatives of the mass matrix elements, as well as more complicated expressions that come from them. An advantage of this bilinear systems observer is that it reduces the velocity estimation problem to a simple linear regression form. Thanks to this, we can use different adaptive, robust, and even finite-time estimation algorithms. In the paper, we formulate some requirements that guarantee the presence of the full derivative of the natural logarithm of the mass matrix determinant in the equations of motion. This makes explicit integration possible. For the sake of completeness, we partly reproduce the existing procedure for observer synthesis with minor modifications. Also, simulation results for the cart-pendulum system implemented in MATLAB/Simulink are shown. These results demonstrate that the proposed observer converges well and has better robustness to measurement noise compared to the dirty derivative approach.

Mitigating Tracking Failures in Sparse-View Gaussian SLAM via Feature-Guided Initialization

phuc_nx — 2026-10-05

Gaussian Splatting has recently emerged as a powerful paradigm for real-time SLAM with high-fidelity scene representation. However, accurate camera tracking remains highly dependent on reliable pose initialization, as photometric optimization requires a good initial estimate to converge. Iterative Closest Point (ICP)-based initialization often fails under challenging conditions such as large viewpoint changes or limited inter-frame overlap, leading to drift and degraded tracking performance. In this work, we address this limitation by introducing a feature-based initialization strategy using SuperPoint and LightGlue, enabling robust correspondence estimation beyond geometric overlap. The estimated pose is obtained via 3D-2D geometric optimization and further refined through photometric alignment, forming a hybrid two-stage tracking pipeline. We evaluate the proposed method on both synthetic and real-world datasets, namely the AriaMultiagent and ReplicaMultiagent datasets. Experimental results demonstrate improved tracking accuracy over the baseline, particularly under sparse-view and large-motion conditions, while maintaining comparable performance in well-conditioned scenarios.

Target Designation Method Using Underwater Video Camera Calibrated with Multi-link Manipulator

Michail Dmitrievich Belousov — 2026-10-05

The paper focuses on developing a method for creating accurate target designations for a multi-link manipulator installed on AUV for various technological operations. These target designations consist of spatial straight lines passing through observed points on work object surfaces, created with the help of underwater monocular video cameras mounted on the AUV. The protective housings of these video cameras consist of thick glass that separate different media due to environment high external pressure, and usually introduce significant optical distortions in the captured images. To correct these distortions, a new calibration method for underwater video cameras is proposed. This method uses a known ray-tracing approach to account for light refraction at media boundaries (with unknown pre-determined refractive indices: air-glass-water) and eliminate lens distortions. This method is applicable for calibrating video cameras with glass of various shapes (flat, cylindrical, etc.). It not only enables accurate determination of the spatial line parameters on which a visible point of an observed object lies but also – after camera calibration – ensures accurate designation targeting of working tools of these manipulators to target work points. The effectiveness of this method has been confirmed by experimental results.

Development of Control System Synthesis Method for Unmanned Aerial Vehicles Moving in a Group

Anton Gubankov — 2026-10-05

The paper deals with a control system synthesis method for a group of unmanned aerial vehicles (UAV) which can implement search or monitoring applications (missing people, fish exploration, aerial photography and so on). Very often this requires efficiency and occurs in adverse weather conditions of the sea, highlands, deep forest, high latitudes, etc. In such conditions, navigation and communication with the ground control station may be difficult. Group control strategy "leader-followers" reduces the risks to the equipment and simplifies the creation of the flight task. To do this, a UWB beacon is placed on the leader, which allows the followers to determine the distance to him and receive data on his movement parameters. As a result, all the followers can provide navigation and maintain a set formation relative to the leader, without communication with the ground control station. The simulation results confirmed the operability and effectiveness of the proposed method UAV group control method.

Shot-range 3D Scanning System for Home Walking Robot

Dmitry Anatolievich Dobrynin — 2026-10-05

This paper presents a prototype of a shot-range vision system for a domestic quadruped robot. The vision system generates a grid of distances to the supporting surface and the nearest obstacles in locations where the robot’s legs might be positioned. Scanning along the robot’s path is performed using a laser lidar. The lidar is rotated by an additional servo drive. The proposed prototype system utilizes an inexpensive 2D lidar. The design of the scanning system is presented, and options for using laser scanners are proposed. A mathematical model of the scanning system for converting lidar data into the walking robot’s coordinate system is presented. The mathematical model is used to construct an obstacle height map in the robot’s near field. The height map can be used to detect obstacles in front of the robot and plan movements on complex surfaces. This paper discusses lidar data processing, which enables the required scanning accuracy to be achieved. Testing on an experimental setup confirmed the key parameters for scanning accuracy. The proposed scanning system reduces the cost of the control system, lowers computing power requirements, and increases the battery life of a domestic walking robot.

Simulation of Static Interactions of a Leg with a Support for a Walking Robot

Dmitry Anatolievich Dobrynin — 2026-10-05

One of the main limitations when using walking robots is the limited capacity of the power supply. At the design stage of a robot, a detailed analysis of how to reduce energy consumption is required. This requires an analysis of the robot’s energy consumption in static and dynamic operating modes. This article discusses the interaction of the robot’s foot with the support surface in a static mode when the robot is stationary is investigated. The design of the leg of a four-legged walking robot is described. A mathematical model of the robot’s leg is given, which is used to calculate the moments acting on the robot’s leg drives. The data for calculating the static moments of the robot’s drives when the robot is not moving relative to the support surface are presented. Criteria for minimizing the energy consumed by a walking robot in this mode have been obtained. The experiments carried out on the stand with a prototype of a walking robot leg showed a good qualitative agreement between the experimental results and the simulation results. This allows you to significantly reduce the energy consumption of a walking robot in static mode.

Hydrodynamic Performance Analysis of Rim-Driven Propeller Based on CFD

LiBin Du — 2026-10-05

The Shaftless Rim-Driven Thruster (RDT) is an integrated propulsion system that highly combines the duct, motor, and propeller. As the most critical component in RDT design, the rim-driven propeller addresses the shortcomings of traditional shaft-driven propellers, such as large spatial requirements, complex structures, and low efficiency. In this paper, a hydrodynamic performance simulation of a self-developed rim-driven propeller is conducted under open-water conditions to analyze the flow field characteristics, and the open-water hydrodynamic performance curves are successfully plotted to evaluate the thrust and torque coefficients. Furthermore, the propeller is investigated under oblique flow conditions to evaluate its hydrodynamic performance variations in complex underwater environments. The simulation results indicate that the oblique flow causes the propeller to bear significant alternating loads, which leads to a highly disordered flow field on the suction side of the blades. This phenomenon ultimately results in a noticeable decline in the overall hydrodynamic performance of the thruster. These findings provide important theoretical guidance for the further structural optimization and practical application of rim-driven propulsion technology.

Method for identifying elastostatic parameters of a multi-link manipulator using a three-axis accelerometer

Ivan Grigorev — 2026-10-05

The elastostatic (compliance) parameters of an industrial manipulator (IM) are used to calculate its deformations under its own weight or external forces. The calculated deformations are compensated by the manipulator control system, which allows to increase positioning accuracy in absolute coordinate system. However, the compliance parameters are usually unknown. This article proposes a method for identifying these parameters based on a consumer-grade three-axis accelerometer mounted on the IM flange and operating in inclinometer mode. During data collection for identification, the IM is moved to several positions and loaded with a known mass in each position. In each position, the accelerometer measures the flange tilt angle before and after loading. Numerical optimization methods are then used to find elastostatic parameters that accurately predict the IM deformations when an external force is applied to its flange. The proposed method, unlike existing methods, does not require expensive or bulky high-precision measuring equipment and lends itself well to automation. Furthermore, the method allows the use of uncalibrated consumer-grade microelectromechanical accelerometers, which are very inexpensive and widely available. The results of mathematical modeling confirm the method's feasibility.

Study of Energy Efficiency of Thunniform Locomotion Using a Robotic Fish Example

illya_mitin — 2026-10-05

Abstract. The development of biomorphic underwater robots that utilize locomotion principles of aquatic organisms is one of the actively advancing areas of modern robotics. Of particular interest is the design of robots implementing thunniform locomotion, characteristic of highly efficient swimmers.

This paper presents a fish-like robot whose design is based on the morphology of the yellowfin tuna and optimized for experimental studies of swimming kinematics under laboratory conditions. The robot is equipped with a caudal propulsion system that generates forward motion through periodic oscillations of the tail fin, mimicking the natural swimming kinematics of fish.

The article describes the structural features of the robot, its internal layout, sensors, and principles of maneuvering in three-dimensional space. Experimental studies of swimming kinematics were conducted in a pool using video recording followed by computer vision data processing. Based on marker coordinates on the robot’s body, swimming speed was determined under different operating modes of the tail actuator.

The dependence of swimming speed and energy efficiency on tail beat frequency at constant amplitude was investigated. It was shown that swimming speed increases approximately linearly with increasing frequency. The minimum Cost of Transport (COT) value was 6.22 J·kg⁻¹·m⁻¹, which is comparable to biological prototypes and the most efficient biomorphic robots. The obtained results can be used in the further design of energy-efficient fish-like underwater vehicles.

Synthesis of Robust Controller based on Stability Criterion of Forced Oscillations

Iuliia Сергеевна Zaitceva — 2026-10-05

The article is devoted to the development of a classical -synthesis method for a class of nonlinear dynamic systems where nonlinearity is described by harmonic linearization, and the parameters of the harmonic input signal – amplitude and frequency – are specified within certain ranges. The key distinction of the proposed approach from the classical one lies in explicitly accounting for the range of amplitudes and frequencies of the input signal as a synthesis parameter set by an engineer based on operational requirements, rather than implicitly estimating it according to the worst-case scenario. The method relies on the stability criterion of forced oscillations in the first harmonic approximation and transforms the synthesis problem into an amplitude-frequency-explicit design scheme with guaranteed stability bounds. The results are illustrated using an example of servo drive control with saturation nonlinearity and combined dynamic nonlinearity. It is shown that for static nonlinearities, the method provides explicit stability boundaries. For nonlinearities with growing uncertainty, the proposed approach reduces conservatism in estimation: the synthesized controller ensures formal stability guarantees within the operating range of amplitudes, whereas the classical H∞ synthesis with conservative weighting produces values exceeding the stability threshold, indicating inability to certify stability under the same uncertainty description. Thus, the method enables formal stability guarantees within engineer-specified operational ranges by explicitly incorporating amplitude and frequency constraints, thereby reducing conservatism compared to classical worst-case approaches while maintaining compatibility with standard robust control frameworks.

Numerical Simulation of Dynamics Deformation of Bio-Inspired Robotic Wing

waleed786 — 2026-10-05

Interest in bio-inspired robotics has grown considerably in recent years, driven largely by the potential of these systems to achieve lightweight construction and high-energy efficiency. In line with this growing field, the present study employs numerical simulations to investigate dynamic deformation in a bio-inspired robotic structure. While the process of dynamic deformation can be described, using a range of constitutive mathematical models such as elastic, elastoplastic, and elastoviscoplastic frameworks this work specifically applies a linear elastic model to examine the deformation behavior. The computational results obtained from this simplified approach demonstrate a strong correlation with experimental data reported in the literature. This high degree of agreement not only validates the numerical model employed but also suggests that, under specific conditions, a linear elastic approximation may be sufficient to capture the essential dynamic characteristics of certain bio-inspired systems. The findings provide a foundational step toward understanding more complex deformation mechanisms.

Fuzzy Logic Application for Underwater Robot with Undulating Fin Control

Aws Ahmad — 2026-10-05

This paper presents the development and analysis of control algorithms for an autonomous underwater vehicle (AUV) equipped with an undulating fin propulsor. A brief review of existing studies on dynamic and kinematic models of underwater robots with flexible fins is provided. A mathematical model of the vehicle interacting with the environment is implemented in the Simscape-MATLAB framework. To solve the problem of depth and course control, the Active Disturbance Rejection Control (ADRC) method is applied. This approach extends the system model by introducing an additional state variable that represents factors not included in the mathematical description. The virtual state is estimated online by an observer and is used to generate the control signal, which makes it possible to compensate for external disturbances. In addition, fuzzy logic methods based on physical and hydrodynamic principles as well as expert knowledge are used to transform generalized control actions into meaningful fin-motion parameters. The advantage of the proposed approach is that it does not require an exact analytical model of AUV dynamics, especially the relationship between fin kinematics and vehicle motion. Simulation results under rapidly varying external disturbances confirm the effectiveness of the proposed mathematical model and control strategy.

Autonomous control of a medical stand for urological surgeries

efremov — 2026-10-05

This article discusses the development of a medical stand for automated retrograde intrarenal surgery. An analysis of the possibilities for automated ureteroscope control using the medical stand is conducted. Alternative developments in this area are considered. The kinematics of a ureteroscope in the surgeon's hands are examined, and its kinematic diagram is provided. The process of tracking the trajectories of the ureteroscope and the surgeon's hand during surgery is conducted. The relationship between the movement of the surgeon's hand and the rotation of the ureteroscope is determined. The range of motion sufficient for performing surgery using the medical stand is determined. Based on the obtained data, the parameters of the medical stand are determined. An algorithm for operating the stand during retrograde intrarenal surgery is proposed. The algorithm includes: defining the workspace by processing an X-ray image, constructing a graph in this space using Delaunay triangulation, and planning the shortest trajectory of movement of the distal end of the ureteroscope towards the stone using Dijkstra's algorithm. The advantages of using the selected methods to solve this problem are revealed. A system for controlling the stand in remote control mode using console handles, depending on the handle rotation speed, is proposed. Graphs of the voltage and rotation speed of the ureteroscope handle versus time during console handle rotation are presented, obtained experimentally using a stand prototype. The data obtained may further assist in the development of a medical stand for retrograde intrarenal surgery.

Synthesis of Control of the Movement of Measurement Points of the State and Power of Heat Sources at the Ends of the Rod

Vugar Adam Hashimov — 2026-10-05

Using the example of controlling the dynamics of the movement of sensors for measuring the temperature state on a rod, the problem of synthesizing control actions in systems with distributed parameters is investigated. Feedback is achieved by measuring the state of the process using moving remote measuring devices (sensors). Linear dependencies of the current values of control actions on the boundary conditions and on the movement of measuring devices from the measured temperature values both at the current and at past moments of the process state are proposed. In the problems, the constant feedback parameters, that is, the coefficients involved in these dependencies, are synthesized. As a result, the feedback control problem is formulated as a problem of parametric optimal control of an object with distributed parameters. To determine the optimal values of the feedback parameters, formulas for the gradient of the objective functional were obtained, which make it possible to use effective numerical methods of first-order optimization. The results of numerical experiments are presented.

Development of a method and algorithm for determining the reaction forces of the joints of the patient's lower extremities

Oksana Viktorovna Emelyanova — 2026-10-05

This paper proposes a method and algorithm for determining the reaction forces occurring in the hip and knee joints, as well as the interaction forces between the musculoskeletal system and the external environment, using mathematical modeling. This allows for the assessment of movement dynamics and load distribution across the joints, muscles, and bone structures of the musculoskeletal system to analyze the patient's functional state and subsequently restore lower extremity motor function.

The diagnostic and rehabilitation device is presented in this work as a crank-slider mechanism equipped with a linear drive, ensuring linear motion of the patient's heel. The mathematical model of the patient's lower limb motion in this device is based on the solution of a direct kinematics and dynamics problem and is represented as Newtonian and Lagrange differential equations of the second kind for a constrained mechanical system. These equations allow for the active and reactive forces acting on the thigh, lower leg, and heel, depending on the driving force generated by the electric drive.

The aim of the study is to develop a method and algorithm for determining reaction forces in the hip and knee joints to assess the dynamics of movements and diagnose the condition of the patient's joints, as well as to analyze the distribution of loads on the elements of the musculoskeletal system when it interacts with the external environment.

Graphs were constructed showing the dependence of forces occurring in the hip and knee joints, heel pressure on the support surface, and frictional forces (over time) on the hip rotation angle. It was established that rheological parameters μ_iof the lower limb muscular system significantly influence the force responses in the hip and knee joints: their increase leads to increased joint loads.

Monitoring the duration of the muscle contraction phase and the nature of limb control provides an objective assessment of the effectiveness of rehabilitation measures, which allows for the optimization of recovery programs for patients with affected joints.

Method of Adaptive Control of Pure Feedback Plants Based on the Actor-Critic Framework

Mikhail Yurievich Medvedev — 2026-10-05

This article proposes a backstepping-based method for design of adaptive control for uncertain pure feedback nonlinear systems using reinforcement learning. Pure feedback nonlinear systems describe classes of electromechanical systems, mobile robots, and moving objects. When setting the problem, the constraints on control and state variables of the plant were taken into account. The adaptive control problem is solved using the one-step Actor-Critic method. Radial basis functions are used to approximate the control policy. To estimate the value of a state, an analytical expression is used that depends on the control error at two time steps. A new algorithm for updating weight coefficients and parameters of radial basis functions is proposed. The article demonstrates the effect of increasing the accuracy of a closed-loop system by dynamically adjusting the parameters of radial basis functions. The proposed algorithm does not contain uncertainty in calculating the increments of weight coefficients. The closed-loop Lyapunov stability property of the system is demonstrated. The theoretical analysis is supported by numerical simulation examples.

Adaptive Navigation Method for a Robotic System with Discrete Mode Switching Based on Environmental Un-certainty Assessment

Petr Trefilov — 2026-10-05

The paper considers the problem of generating a navigation solution for a robotic system operating under changing environmental conditions, degradation of measurement channels, and incomplete a priori information about the current situation. It is shown that single-mode navigation algorithms with a fixed measurement processing structure do not always provide the required state estimation quality when the availability and reliability of navigation sources change. An adaptive navigation method based on discrete switching of operating modes depending on an integral uncertainty estimate is proposed. Within the method, diagnostic features are formed to characterize the consistency of measurements with the predicted state, the level of estimation uncertainty, and the structural availability of navigation channels. Based on these features, an integral uncertainty index is calculated and used to select one of the predefined processing modes: nominal data fusion, partial degradation of measurement information, or limited correction. To prevent false transitions, the selected mode is confirmed over several consecutive steps. Numerical simulation results show that the proposed approach reduces the influence of degraded measurements on state estimation and decreases the position error compared with a baseline single-mode algorithm.