ICR 2026

Intelligent task allocation between moving objects based on reinforcement learning algorithm

Mikhail Yurievich Medvedev — 2025-10-24

This paper provides a review of task allocation in robot groups using reinforcement learning algorithms. Based on this review, it is concluded that there is a need to develop task allocation methods based on reinforcement learning. The variable number of robots and target points must be taken into account. The paper formulates a task allocation problem for a group of robots functioning in a discrete environment with obstacles. Two variants of Q-algorithm with deep neural network for task allocation are proposed. The first algorithm is a centralized one. It uses a single neural network that approximates the value of the actions of all robots in the group. The second algorithm is a distributed one. This algorithm on board each robot computes the selection values of each target point in the current state. The proposed algorithms are investigated by numerical modeling methods for an environment with a variable number of robots and target points.

Task Allocation in Groups of Mobile Robots under Uncertainty using an Adaptive Ant Colony Algorithm

Unknown — 2025-10-24

This paper addresses the task allocation problem in groups of mobile robots operating under parametric and stochastic uncertainty caused by sensor noise, environmental dynamics, and incomplete information about tasks. A task distribution method is proposed based on a modified ant colony optimization (ACO) algorithm incorporating an adaptive mechanism responsive to environmental changes. A mathematical model of multi-objective optimization is developed, accounting for task completion time, energy consumption, and solution robustness. An enhanced algorithm is presented, capable of adjusting its parameters in response to the emergence of new tasks and robot failures. The objective function integrates total execution time, cumulative energy cost, and robustness to external disturbances. A key feature of the proposed modification is the linear adaptation of the weighting coefficients associated with the pheromone and heuristic components in the probabilistic decision rule, implemented through progressive adjustment of the α and β parameters. Numerical experiments confirm the superiority of the adaptive algorithm over the baseline version. It is shown that dynamic parameter tuning significantly accelerates solution reconfiguration in response to environmental changes. The results demonstrate the viability of the proposed approach and its effectiveness in constructing intelligent control systems for mobile robot groups operating in complex and dynamic environments.

Application of an Aqua-Aero Robotic System for hyperspectral and biochemical monitoring of environmental pollution in water areas

elchin — 2025-10-24

This paper presents the initial testing results of the AARS (Aqua-Aero Ro-botic System), a robotic complex designed for hyperspectral and biochemical mon-itoring of waterbody pollution. The AARS comprises an unmanned boat equipped with a landing platform for UAVs carrying a hyperspectral camera operating in the 400–1000 nm range. Upon detecting pollution zones via hyperspectral imaging, the system transmits the coordinates to the unmanned boat. The boat then navigates autonomously within the contaminated area, performing real-time biochemical ex-press analysis while in motion. For this purpose, the boat is fitted with an autono-mous water sampler capable of collecting eight water samples. Additionally, during operation, the system detects the presence of plankton and cyanobacteria by meas-uring photosynthetic activity, optical density, and chlorophyll concentration. Sim-ultaneously, the following parameters are analyzed: pH, dissolved oxygen concen-tration, temperature, water hardness, transparency, and REDOX potential.

COMPARISON OF THE EFFECTIVENESS OF THE GENETIC ALGORITHM AND THE MODIFIED LOUVAIN ALGORITHM FOR THE UNDERWATER WIRELESS SENSOR NETWORK CLUSTERING MODEL

Unknown — 2025-10-24

This paper presents a simulation model for managing the resources of an underwater wireless sensor network, based on the use of modularity as a metric in the clustering process. The model employs a modified Louvain algorithm, a genetic algorithm, and Dijkstra’s algorithm for constructing optimal message transmission routes. The main focus is on the comparative analysis of the efficiency of the developed algorithms. Simulation results confirm that the proposed model effectively addresses the challenges of energy-efficient clustering and reliable routing. Algorithms that aim to maximize modularity form stable and balanced cluster structures by accounting for the residual energy of sensors and are capable of supporting dynamic reclustering under changing topological or energy conditions. This leads to reduced data loss, balanced load distribution, and extended autonomous operation time of the network. The genetic algorithm, which uses modularity as a target function, demonstrates high adaptability to various network configurations and significantly reduces the number of message retransmissions during data collection. The comparative analysis shows that the genetic approach provides higher transmission reliability and energy efficiency compared to the Louvain algorithm, particularly under dynamically changing underwater conditions. The developed model can be applied to practical marine monitoring tasks where physical constraints, equipment characteristics, and the variability of network topology must be taken into account simultaneously.

Planning the movements of a mobile robot in a three-dimensional environment with constraints on pitch and yaw angles

medvmihal1 — 2025-10-24

In this paper, the problem of motion planning in a three-dimensional environment taking into account constraints on pitch and yaw angles is considered. A hierarchical algorithm for solving the problem for an environment in which obstacles are approximated by spheres is proposed. At the upper level, the feasibility of planar motions is analyzed taking into account the specified constraints and the location of the start and target points. If the analysis confirms that the constraints are met, a decision is made to move to the specified point. Otherwise, a maneuver is performed in order to ensure the required constraints on pitch and yaw angles. Options for performing additional maneuvers are proposed that ensure a significant reduction in the path length during its execution. The motions at the lower level are implemented using an algorithm based on the characteristic visibility graph. This algorithm has advantages over cellular decomposition methods in the case of sufficiently complex scenes. Docking of individual sections of the motion is performed using a smoothing procedure according to the modified Dubins method. The proposed algorithm allows obtaining relatively simple and feasible target trajectories. In addition, the algorithm allows you to build optimal movement trajectories

Set-Input Trees: Discovering New Effective Tree-Based Multiple Instance Learning Algorithms

Lev Utkin — 2025-10-24

We propose gradient-based Set-Input Trees, a novel tree-based architecture for Multiple Instance Learning (MIL) that addresses both classification and regression. Unlike conventional methods relying on fixed aggregation (e.g., min/max pooling), the proposed architecture integrates gradient-boosted trees with an attention mechanism: instances are processed independently, while leaf embeddings are pooled via learned attention weights. This preserves interpretability while capturing bag-level structure. For regression, we introduce a synthetic MIL formulation, feature-to-bag conversion, enabling evaluation on continuous targets. Experiments show outperformance of the proposed algorithms on standard MIL benchmarks comparing to tree-based models. The model’s tree-based design ensures scalability and transparency, bridging instance-level decisions with set-valued predictions. Codes implementing the proposed algorithms are publicly available.

Automated Control System for Magnetically Active Microrobots for Medical Applications

Unknown — 2025-10-24

This paper presents an automatic control system for achieving controlled motion of magnetoactive micro-objects. A mathematical model is developed to describe the dynamics of a spherical body made of magnetically filled polymer under the influence of a moving permanent magnet's magnetic field. The proposed control system architecture enables two distinct motion regimes: rolling with minimal interaction forces against the channel wall, or sliding of the deformed magnetoactive object with significant contact forces. This effect is achieved through optimal positioning and orientation of the magnet, which provides precise control over the normal reaction force between the microrobot and the contact surface. The system has potential applications in various medical procedures, including thrombectomy and vascular biopsy. A key feature of the proposed system is the integration of a computational module that performs real-time calculations of force and kinematic parameters for magnetoactive control. The paper includes numerical simulations demonstrating different motion parameters of the magnetoactive object under identical surface and object conditions, validating the effectiveness of the proposed control architecture and methodology. The results confirm the system's capability to achieve diverse motion patterns through optimized magnetic field control.

Synthesis of nonlinear systems of underwater vehicles based on quasi-linear models

medvmihal1 — 2025-10-24

This study is devoted to the application of one of the modern approaches to the synthesis of nonlinear automatic control systems for autonomous unmanned underwater vehicles. The high requirements for such underwater vehicles necessitate the use of methods that take into account the features of their nonlinear characteristics. At the same time, the task of synthesizing nonlinear control systems is quite complex. This has led to the development of a significant number of different methods for solving this problem. Most of these methods require a preliminary transformation of the equations of a nonlinear object to some special types, but finding the appropriate transformation is very often also a difficult task. A new quasi-linear approach is considered below, in which the desired transformation is reduced to differentiation, and in some cases to integration of partial derivatives of the nonlinearities of the object. This approach leads to simple engineering methods for synthesizing control systems for autonomous underwater vehicles. Based on it, it is possible to create modern control systems for nonlinear objects in the field of shipbuilding, energy, agriculture, instrument engineering and other industries.

Foot Trajectory Generation for a Quadruped Home Robot

dobrynin_da — 2025-10-24

This article discusses the construction of a foot trajectory for a quadruped walking robot with 12 degrees of freedom. A mathematical model of the robot leg for calculating inverse kinematics is given. The trajectory of the ro-bot’s foot motion is synthesized for uniform rectilinear motion of the robot. The foot trajectory is a basic element for constructing a smooth and stable gait of a walking robot. The trajectory of the robot’s foot consists of a support phase and a transfer phase. In the transfer phase, elliptical and rectilinear trajectories of the foot are used. Dependences of the robot leg drive parameters on time are constructed. The article analyzes these dependencies. It is shown that the maximum speeds and accelerations of the robot leg drives are achieved in the transfer phase. The possibility of optimizing the trajectory parameters for the specified drive parameters to obtain the maximum speed of the walking robot is shown. The trajectory of the robot leg prototype is tested on an experimental stand. It is shown that the executed trajectory has parameters that are in good agreement with the calculated values. In conclusion, practical conclusions are given on optimizing trajectory parameters in order to obtain maximum speed.

The unmanned complex of assess marksmanship performance

- azadbayramov — 2025-10-24

Recently, in the training of military soldiers and officers, much attention has been paid to improving the efficiency and accuracy of shooting from small arms. Considering that hundreds and thousands of soldiers and officers participate in military exercises, there is a need to create robotic systems that would quickly assess the accuracy of shooting without the participation of a human operator. This paper presents the developed unmanned robotic complex of assess marksmanship performance for small arms fire. The complex is designed for effective training of soldiers and officers in shooting at close and long distances (20 ¸ 1500 m.). The operating principle of the system is briefly described and technical characteristics are given. It is shown that the proposed unmanned robotic system has significant advantages over foreign analogues. The shooting evaluation system uses automated LOMAH software, hardware, and a PTZ camera to conduct and automatically evaluate shots at various types of targets according to the shooting task. An electric Power Supply Unit, a Motor Control Unit and a software algorithm for controlling the complex have been developed. It has been shown that the developed robotic complex has a number of advantages over existing analogues.

Determination of Forces Acting in the Joints of the Lower Extremities During Rehabilitation

Oksana Viktorovna Emelyanova — 2025-10-24

Currently, robotic stands are becoming widespread, and are used for the rehabilitation of the lower extremities based on the continuous passive development of one or more joints. In this case, additional excessive forces arise, which have a significant impact on the joints and the rehabilitation process. Direct measurement of these forces is not possible, as they are internal. The article considers the joint movement of the patient's leg and the rehabilitation exoskeleton as a human-machine system (HMS), with the help of which movement is provided in the joints of the lower extremities at a given law of change in the angle of rotation of the thigh. The method of estimating internal forces is based on the Dalembert principle, which allows us to consider and describe HMS under dynamic equilibrium conditions. The system of equations recorded for individual parts of the system allows real-time calculation of the forces acting in the joints based on data, including that obtained using a system of sensors installed on the exoskeleton. Thus, by calculation, it is possible to obtain the values of the forces acting in the thigh, knee and ankle joints. The results of numerical modeling, taking into account experimental studies, showed a significant effect of the coefficient of friction on horizontal forces arising in the femoral, knee and ankle joints, and a change in the angle of rotation of the thigh leads to an increase in the moments of resistance on the femoral and knee joints. This will help the rehabilitation doctor control the movement of the exoskeleton links and generate loads in passive and active modes of operation.

EffiAtt-Emotion: A Lightweight Emotion Recognition Model for Adaptive Human-Robot Interaction in Collaborative Robotics

Yakushev Dmitrievich Alexey — 2025-10-24

Effective and safe interaction in collaborative robotics hinges on the robot's ability to adapt its behavior to the human's cognitive and emotional state. This paper introduces EffiAtt-Emotion, a lightweight deep learning model for facial emotion recognition designed specifically for resource-constrained onboard robotic systems. With only 13.3 million parameters, the model demonstrates a high and balanced performance, achieving an accuracy of 80.6% and a macro F1-score of 0.806 for eight distinct emotion classes. We present a framework for integrating EffiAtt-Emotion into a collaborative robot's (cobot's) control loop, enabling real-time adaptive behavior. A use case on a collaborative assembly line is detailed, where the cobot adjusts its speed, provides assistance, or enters a safe stop mode in response to the human operator's recognized emotions, such as confusion, frustration, or fear. This approach directly addresses documented challenges in HRI, where human factors like stress and fatigue contribute to decreased performance and safety risks. By making the robot aware of the operator's state, our system aims to enhance task efficiency, ergonomics, and critically, to improve safety in line with the human-centric principles of Industry 5.0.

Algorithms for Calculation of a Modular Wheeled Mobile Robot’s Configuration

Ildar Shamilevich Nasibullayev — 2025-10-24

The paper presents algorithms for calculating the configuration of a modular wheeled robot when moving along an arbitrary plane trajectory. The robot is a system of two-wheeled modules, sequentially connected using a ball joint. Two variants of the module wheelset are considered: fixed and with controlled rotation relative to the vertical axis. Mathematical models of the coupling equation are constructed on the basis of continuity in a ball joint. For the first option, an iterative algorithm of the second order of accuracy relative to the time step of the leading robot is proposed. The solution of the coupling equation is carried out analytically. For modules with a controlled wheelset, the configuration is determined using an algorithm that uses a second-order numerical scheme to solve a nonlinear parametric coupling equation. It was found that when moving a robot with a fixed wheelset, the trajectories of the driven models deviate from the trajectory of the leading one. The magnitude of the deviation increases with the growth of the trajectory curvature and accumulates from the first module to the last. The proposed algorithms allow for kinematic studies of a modular wheeled robot. To perform dynamic studies, a computer model of the robot in the ROS environment and using the Gazebo simulator is proposed. The results of computer experiments are in good agreement with the results of mathematical modeling.

Near Field Scanning System of Home Walking Robot

dobrynin_da — 2025-10-24

The paper considers the principles of constructing a machine vision system for a domestic four-legged robot. The machine vision system constructs a grid of distances to the support surface in places where the robot’s legs can be placed. Scanning along the robot’s line of motion is performed using a laser lidar. The proposed system is based on the use of an inexpensive 2D lidar. A model of surface scanning for determining support areas is presented. The influence of scanning parameters and distance measurement accuracy on the accuracy of constructing a height map is analyzed. It is shown that the accuracy of measuring the parameters of horizontal and vertical surfaces depends on their location relative to the robot’s body. Conclusions are made and recommendations are given for improving the accuracy of measurements. The conclusions made allow choosing a laser lidar that is most suitable for the scanning system. Testing on a simulator confirmed the high-quality scanning of the robot’s near zone. The proposed scanning method allows reducing the cost of the control system, reducing the requirements for computing power and increasing the battery life of the domestic robot.

Lamarckian Evolution Based Algorithm for Multi-robot Path Planning Problem

Anna Borisovna Klimenko — 2025-10-24

This study addresses the issue of decreasing computational resources spent on solving the multi-robot path planning (MRPP) problem. We propose a novel algorithm inspired by Lamarckian evolution principles integrated into the traditional Darwin evolutionary algorithm. The proposed Lamarckian Evolution-Based (LEB) algorithm aims to enhance both the convergence rate and the accuracy of the MRPP problem solution, thereby reducing overall computational effort. Through extensive simulations, the LEB algorithm demonstrated superior performance compared to standard approaches such as Genetic Algorithm (GA), Particle Swarm Optimization (PSO) and Simulated Annealing (SA). Despite requiring additional computational resources per iteration due to added local optimization stages, the LEB algorithm achieves significant reductions in total resource expenditure and improves solution quality within a specified computational budget. Our findings highlight the potential of Lamarckian-inspired strategies for multi-robot navigation tasks efficiency.

Fish Disease Detection and Counting Using Enhanced YOLOv11 for Application in Aquatic Robots

Nghia Van Le — 2025-10-24

The rapidly growing global population creates urgent demand for protein and fish-based food sources. Consequently, the fisheries sector faces the dual challenge of meeting seafood supply needs while protecting marine ecosystems. AI and robotics-enabled smart aquaculture emerges as a critical solution to this dilemma. In high-density aquaculture systems, where disease transmission can escalate uncontrollably and cause catastrophic losses, computer vision integration has demonstrated exceptional efficacy through timely detection and enumeration of diseased fish. This paper analyzes a disease detection and counting system based on an enhanced YOLOv11 architecture. The model exhibits remarkable capability in identifying and localizing diseased specimens within complex underwater environments. On our manually annotated two-class dataset (diseased/healthy fish), it achieves 98.2% mAP@0.5, 76.9% mAP@0.5:0.95, 96.4% precision, and 95.2% recall. Critically, the system delivers not only real-time processing but also high accuracy, enabling prompt identification of piscine dermal pathologies for immediate intervention. This prevents widespread outbreaks and substantial economic damage. These findings pave the way for novel research trajectories and advanced intelligent solutions in future aquaculture development.

Trajectory Tracking Control of Surface Mobile Buoys Based on A-Star - NMPC Algorithm

YunLi Nie — 2025-10-24

针对传统浮标机动性不足、轨迹跟踪精度低的问题，设计了一种新型水面移动浮标，并设计了一种融合A*算法和非线性模型预测控制（NMPC）的轨迹跟踪算法。首先，建立三自由度水面移动浮标的数学模型。其次，设计A*-NMPC算法，将A*算法的全局路径规划能力与NMPC的动态优化特性相结合，形成协同控制框架。最后，通过仿真实验，对比分析了与线性模型预测控制（MPC）的轨迹跟踪效果。结果表明，A*-NMPC在航向突变场景下表现出显著优势，最大纵向误差较MPC降低50%，横向误差降低80%，且具备动态避障和快速响应能力。

Investigation into the Impact of Propellers on the Rudder Efficiency of Ships during Steering Operations

LiBin Du — 2025-10-24

In order to observe the influence of the wake behind the propeller on the rudder, this paper conducts an interference experiment on the propeller-rudder system using the B4-55 propeller and the NACA0020 marine rudder, and analyzes the hydrodynamic characteristics of the rudder behind the propeller wake. The computational domain is divided into a rotational domain and a fluid domain, and the domain is discretized using cut-cell meshes. The discrete equations are solved using the model. The rotation motion of the propeller is simulated by using the slip grid. Data information is transferred through the interface to complete the hydrodynamic performance simulation of the propeller system. This paper analyzes the hydrodynamic performance of marine rudders under different rudder angles, and explores the influence of the propeller on the resistance, lift and limit rudder angle of the rudder. The distribution characteristics of the velocity field and pressure field in the computational domain were analyzed.

Development of UAV control system architecture with integrated visual navigation for agricultural monitoring

Unknown — 2025-10-24

The development of unmanned aerial vehicles (UAVs) equipped with first-person view (FPV) cameras is advancing rapidly, posing novel engineering and scientific challenges. Research areas related to the implementation of operator assistance systems and partial flight automation are becoming increasingly relevant. This work is devoted to the development of the hardware component of an FPV UAV control system for partial flight automation based on visual navigation principles. The developed FPV UAV control system architecture features an original integration of a flight mode controller into the existing conventional flight controller-based control system. The paper presents an algorithm and describes the operational principles of the flight mode controller, which serves as the core element integrating the visual navigation system into the UAV’s control system. Testing of the proposed control system architecture was conducted using a custom-built UAV designed for monitoring tasks. The UAV’s maximum takeoff weight is 5.5 kg, and its propeller diameter is 22 inches. The average command latency of the flight mode controller is 0.42 ms, which increases the total control delay by 5.3% at a refresh rate of 250 Hz. The proposed solutions for visual navigation system integration allow the operator to control the UAV and switch flight modes without significant delays.

Robust UAV Control Using Confidence Weighted Multi Sensor Integration

petertrfi — 2025-10-24

This paper presents an adaptive control framework that enables a unmanned robot to follow its planned trajectory including cases when the reliability of its sensor suite changes unpredictably. The main contribution is the adaptive mode of operation: the system dynamically adjusts the threshold value of each sensor stream, recalculates their weights in real time, and recombines control system. While high precision measurements are available, the controller runs in a nominal linear quadratic mode. When observability drops—because a sensor drifts, saturates, or fails—the algorithm transitions smoothly to a prediction only mode that relies on the process model, and, if necessary, to a robust H infinity mode that preserves stability under worst case disturbances. The underlying estimation layer is based on an innovation driven weighting rule embedded in an Extended Kalman Filter, making the approach independent of any particular sensor type. Hardware simulation and outdoor testing confirm that adaptive mode significantly improves tracking accuracy, shortens recovery time after faults, and eliminates the need for manual retuning when the sensor configuration changesю

Closed-Loop Chain Linkage-Based Hand Exoskeleton: A Lightweight and Modular Solution for Hand Rehabilitation

Shifa Sulaiman — 2025-10-24

Prosthetic hands are vital assistive devices that significantly enhance quality of life for individuals with upper limb amputations, enabling them to regain autonomy and perform essential daily activities. However, many existing prosthetic solutions are hindered by high costs, excessive weight, and complex actuation systems that limit accessibility and usability. This paper introduces a novel prosthetic hand design that addresses these limitations through a mechanically efficient and cost-effective approach. The proposed system employs five motors, one per finger, combined with a closed-loop chain linkage mechanism that transmits motion patterns across joints of each finger. This eliminates a need for multiple actuators per finger, thereby reducing an overall weight, power consumption, and cost of the device. We present a complete design methodology, a mechanical architecture, and functional analysis of the prototype. The results demonstrate feasibility and advantages of the approach, offering a promising direction for developing affordable, efficient, and user-friendly prosthetic hands.

Synthesis of process dynamics control using current and previously conducted measurements of its state

Vugar Adam Hashimov — 2025-10-24

The article investigates a method for synthesizing optimal control of processes (objects, systems) with lumped parameters described by systems of nonlinear autonomous differential equations with ordinary derivatives. To synthesize current process control values, it is proposed to use information about its state both at the current and past moments of time. For the values of control actions, a linear dependence on the measured process states is proposed and used. To determine the optimal values of the feedback parameters involved in these dependencies, the corresponding formulas for the gradient of the objective functional were obtained. Computer experiments were conducted using gradient-type optimization methods. An analysis of the influence of the number of measurements used for feedback on the quality of control is made.

Improving of brain-machine interaction performance by sensory and transcranial magnetic stimulation

Susanna Gordleeva — 2025-10-24

This study investigates the potential of transcranial magnetic stimulation (TMS) and vibrotactile stimulation to enhance the performance of motor imagery-based brain-machine interfaces (MI-BMIs). Our findings demonstrate that TMS applied to the dorsolateral prefrontal cortex (DLPFC) with specific parameters (5 Hz frequency, 6-minute duration, 90% of the resting motor threshold) induces preactivation of the occipital region, facilitating sensorimotor integration and reducing MI-BMI latency. Additionally, integrating vibrotactile feedback during MI-BMI training enhances event-related desynchronization (ERD) over the contralateral motor cortex for non-dominant hand imagery and increases motor cortical excitability, as evidenced by elevated motor-evoked potentials (MEPs). These results highlight the complementary roles of central (TMS) and peripheral (vibrotactile) stimulation in optimizing MI-BMI systems, offering promising avenues for neurorehabilitation and improving user proficiency in BMI control

Development of a synthesis method for self-adjusting correction devices for DC brushless motors

proa2931 — 2025-10-24

The article is devoted to the solution of the problem of synthesizing self-adjusting correction devices (SCD) for brushless DC motors (BLDC) used as actuators of various mechatronic systems, including multi-stage manipulators and mobile robots. The specified correction devices allow to significantly increase the dynamic control accuracy of these loaded motors by generating an additional control signal fed to the input of the electronic speed controller of the motors, which ensures stabilization during operation at a given nominal level of their dynamic parameters taking into account the current values of external load moments reduced to their output shafts. To solve this problem, an equivalent model of a BLDC described by a system of three first-order differential equations and a method for synthesizing typical SCDs used for collector motors are used. To confirm the efficiency of the synthesized SCDs, modeling and a full-scale experiment using an electromechanical stand were carried out.

Feedback Control Approach Based on Classes of Zonal Control Actions

Samir Zakir Quliyev — 2025-10-24

This paper presents a feedback control synthesis approach for nonlinear dynamic systems with lumped parameters under partial observability and time delays. We propose a zonal feedback control framework, where the state space of observable components is partitioned into finitely many zones, each associated with constant or linear feedback gains. Unlike traditional feedback laws that continuously depend on the instantaneous state, the control input here changes only when the system transitions between predefined zones. This reduces the infinite-dimensional feedback synthesis problem to a finite-dimensional parametric optimization problem, enabling efficient computation of control parameters. The control design problem is formulated as the minimization of an objective functional that accounts for both the system’s dynamic response and control energy, averaged over uncertain initial conditions and parameter variations. To solve this problem, we derive analytical formulas for the gradient of the objective functional with respect to the zonal feedback parameters by employing perturbation techniques and adjoint systems. These gradient formulas allow the implementation of first-order optimization methods, such as the projected gradient algorithm, to determine optimal zonal gains while satisfying control constraints. The presented approach has potential applications in robotics, autonomous systems, and other domains where real-time computation and robustness to uncertainties are critical.

Multi-Criteria Approach to Path Planning for Unmanned Tractors Considering Energy Constraints and Soil Compaction

Ramil Faizullin — 2025-10-24

Autonomous agricultural vehicles operating under the Controlled Traffic Farming (CTF) paradigm face complex routing challenges when minimizing soil compaction, total mission time, and station placement under battery constraints. This paper introduces Multi-Objective Coordinated Autonomous Routing and Placement with Fixed Lanes (MO-CARP-FL), a novel multi-objective evolutionary algorithm designed to optimize the coordinated routing of homogeneous autonomous tractors over a predefined field traffic lane. The algorithm simultaneously addresses five conflicting objectives: minimizing soil compaction using a logarithmic saturation model, minimizing total route time, reducing the number of charging stations, preserving spatial coherence in assigned routes, and balancing workload among tractors. Chromosomes encode both routing and station placement decisions, and custom crossover and mutation operators preserve structural feasibility. A soil compaction model and energy-aware constraints are integrated into the evaluation function. Experimental simulations demonstrate that MO-CARP-FL produces environmentally sensitive routing plans while reducing field degradation. The proposed method is validated through CTF field scenarios, and its results are visualized to provide interpretable insights into route distribution, station usage, and soil impact. This work contributes to multi-objective optimization in agricultural logistics by addressing both environmental impact and operational efficiency in autonomous field operations.

SI-RRT and ST-RRT* For Prioritized Multi-Manipulator Planning: Empirical Evaluation

Nuraddin Masud ogli Kerimov — 2025-10-24

Multi-manipulator systems are essential in warehouse automation, assembly, manufacturing, and logistics, requiring robots to coordinate within shared workspaces while avoiding collisions. This coordination is often formulated as multi-manipulator motion planning, which is notoriously hard to solve due to the high dimensionality of the configuration space. To this end, decoupled approaches are often used, such as prioritized planning. In the latter approach, robots are assigned priorities and plan trajectories sequentially, treating higher-priority agents as dynamic obstacles. Thus, the performance of a prioritized planner critically depends on the underlying single-agent planner. Recent advanced algorithms in this area, tailored specifically for manipulation planning in the presence of moving obstacles (high-priority robots), include Space-Time RRT* (ST-RRT*), which enhances RRTConnect for the time dimension, and Safe-Interval RRT (SI-RRT), which enriches RRTConnect with safe interval planning. While prior research demonstrated that SI-RRT outperforms ST-RRT* in the single-agent scenarios, their effectiveness in the multi-robot settings remains unexplored. This paper presents the first comprehensive evaluation of ST-RRT* and SI-RRT within prioritized planning for multi-manipulator systems. We compare both planners using the TAPAS dataset (containing diverse pick-and-place tasks) and synthetic random tests, scaling from 2 to 8 robots. On synthetic tests, SI-RRT achieves lower path lengths and makespan despite increased planning time for 6–8 robots due to constrained time corridors, caused by raised scene complexity resulting from reduced makespan. On the TAPAS benchmark, SI-RRT consistently demonstrates the same faster computation, shorter trajectories, better success rate and reduced makespan across all tested scenarios.

A method of positional-force control of an underwater manipulator

Vladimir Fedorovich Filaretov — 2025-10-24

The article is devoted to solving the problem of developing a method for constructing positional-force control systems for electric drives (ED) of multi-link underwater manipulators (UM) mounted on autonomous underwater vehicles operating in the mode of landing on the ground or on work sites. To solve this problem, a comprehensive method is proposed. First, for an ED of each degree of freedom of the UM, a self-adjusting correction device is constructed, which makes it possible to stabilize the variable dynamic parameters of this ED at a given nominal level. This stabilization compensates for the influence of Coulomb and viscous frictions, the interactions between the UM links and influence of a viscous medium, on the quality of control. Then, for each ED sliding observers are constructed, which make it possible to obtain information about the current values of the external torques, speeds and accelerations of rotation of the output shafts of all ED. Using this information and measurements of only position sensors of the specified ED, it is possible to construct position-force regulators for them that minimize the selected quadratic cost function. This minimization makes it possible to ensure not only the precise movement of the UM tool along specified spatial trajectories in the presence of significant impacts from the viscous medium, but also to create the required force effects on underwater work objects when performing various technological operations. The operability and effectiveness of the constructed positional-force control systems is confirmed by the results of computer modeling.

Methodology for personalized human–robot collaboration that accounts for individual characteristics of human biological signals

Rinat Romanovich Galin — 2025-10-24

This paper explores a methodology for integrating deep learning techniques into personalized human-robot collaboration (HRC) systems by decoding individual EEG-based neurophysiological patterns. Rather than focusing on biometric identity recognition, the proposed approach centers on detecting individual typological features of EEG dynamics – unique cognitive-motor response patterns that emerge during interaction with robotic systems. EEG data were collected from 30 participants under standardized conditions, with frontal channels (F3, F4, F7, F8) used for signal acquisition. The study introduces a targeted adaptation of classical CNN architectures – AlexNet and MobileNetV2 – to the structural and dynamic characteristics of EEG spectrograms, including low spatial resolution, single-channel input, and domain-specific noise. Two adapted models – SimpleAlexNet and LiteMobileNet2D — were evaluated for their ability to classify EEG-based pattern types under constrained computational conditions. Experimental results show that LiteMobileNet2D achieves a superior balance between accuracy and generalization, maintaining low overfitting despite aggressive model simplification. SimpleAlexNet, while prone to overfitting, demonstrated acceptable performance in medium-complexity tasks. These findings confirm that individualized EEG pattern recognition is feasible with lightweight neural models, enabling real-time adaptation of HRC behavior to the operator’s current physiological state and laying a foundation for scalable, context-aware HRC systems.

Efficient Human-Robot Interaction via Deep Perception and flexible Motion Planning

Thanh Nguyen Canh — 2025-10-24

Human-Robot Interaction (HRI) is an emergent field propelled by advancements in artificial intelligence, yet achieving seamless human understanding and responsive robot control remains a significant challenge. This paper introduces an efficient, integrated HRI system for a custom-built dual-arm robot, presenting two primary contributions. First, we propose a sophisticated perception system that leverages deep learning to interpret human states. This system employs a Multi-task Cascaded Convolutional Neural Network (MTCNN) for robust face detection, a Deep Convolutional Neural Network (DCNN) for recognizing emotions, and a Long Short-Term Memory (LSTM) network to identify dynamic gestures from image sequences. Second, we detail a flexible dual-arm robot control system built on the Robot Operating System (ROS). This control system utilizes the Rapidly-exploring Random Tree (RRT) algorithm for efficient path planning, enabling the robot to translate recognized human cues into corresponding actions. Comprehensive evaluations on benchmarks, including the WIDER FACE and FER2013 datasets, validate the perception models. The proposed system was validated through both simulation and physical experiments, demonstrating high accuracy in perception and control. The results highlight the framework’s effectiveness in creating fluid and responsive interactions for complex HRI scenarios.

Refined 3D Object Localization with Monocular Camera using Depth Estimation and Geometric Refinement

Thanh Nguyen Canh — 2025-10-24

Accurate 3D object localization is a fundamental requirement for applications in industrial robotics, augmented reality, and autonomous navigation. While traditional multi-view systems are precise, their hardware complexity and cost limit widespread adoption. Monocular vision offers a cost-effective alternative but struggles with the inherent challenge of inferring depth from a 2D image, often leading to significant localization errors. This paper presents a novel methodology that overcomes these limitations, achieving high-precision 3D localization using only a single camera. Our proposed framework integrates three synergistic stages. First, a camera calibration process using a checkerboard pattern corrects lens distortion and establishes a real-world metric coordinate system. Second, we employ the YOLOv8 model for real-time 2D object detection and the ZoeDepth network to generate a dense depth map from the monocular input. To mitigate spatial inaccuracies that arise when objects are positioned off-center, we introduce a geometric object position refinement technique. This method adjusts the object’s center based on its projected image coordinates and depth information. Experimental results demonstrate the superiority of our approach, which achieves an average position error of just 4.6mm. This represents a significant 59.29% improvement in localization accuracy compared to using the standalone YOLOv8 detector, showcasing our method's effectiveness for robust 3D localization.

EKF-FOGS Enhanced Observer For Autonomous Tracked Vehicle Control In Slippery Terrain

Hai An Nguyen — 2025-10-24

Autonomous tracked vehicles play a crucial role in precision agriculture, automated construction, and emergency rescue operations. However, maintaining positioning accuracy under slippery terrain conditions remains a significant challenge. This paper proposes the EKF-FOGS (Extended Kalman Filter with Focused Observation for Guided Steering) observer—a state estimation solution capable of maintaining high accuracy when vehicles operate in environments with high slip ratios. The system integrates data from encoders, IMU, and depth cameras while applying an adaptive mechanism that automatically adjusts to changing terrain conditions. The Bekker-Wong soil mechanics model is employed to simulate track-terrain interaction, providing a realistic testing environment. This method addresses the error accumulation problem of odometry by incorporating absolute position references independent of inertial and distance measurements. MATLAB simulations demonstrate that EKF-FOGS significantly improves position estimation accuracy compared to traditional EKF, achieving average errors below one meter in various high-slip scenarios. These results confirm the method's potential for widespread application in GPS-denied positioning systems, particularly for high-precision outdoor tasks.

Trajectory Tracking and Orientation Stability of Planar Cable Robots Using PSO-Tuned Sliding Mode Control

Duc-Chi Nguyen — 2025-10-24

While cable-driven parallel robots (CDPRs) feature wide workspaces, strong payload
capacity, and good scalability, they are hindered by nonlinear dynamics, redundant actuation,
and non-negative tension constraints. Traditional control strategies, including PID, adaptive
methods, and robust nonlinear approaches, often suffer from chattering, parameter sensitivity,
and difficulties in ensuring feasible cable tensions. This paper proposes a hybrid control framework
that integrates Sliding Mode Control (SMC) with Particle Swarm Optimization (PSO) to
achieve precise trajectory tracking and orientation stability in a four-cable planar CDPR. The
SMC ensures robustness against model uncertainties and external disturbances, while PSO optimally
tunes the controller gains within bounded ranges to mitigate chattering and enhance
stability. A null-space–based positive tension converter is incorporated to enforce non-negative
tensions, minimizing deviation from the nominal wrench. Simulation results on circular trajectories
demonstrate that the proposed approach maintains tracking accuracy with position
errors below 3%, suppresses angular deviations within 0.34°, and ensures feasible tension profiles
within 0–60 N. The framework highlights the effectiveness of PSO-optimized SMC in addressing
robustness, stability, and physical constraints, offering practical insights for deploying CDPRs in
automated and precision tasks.

Synthesis method for UAV automatic landing system on a marine moving platform using UWB technology

Anton Gubankov — 2025-10-24

The paper deals with an approach to synthesis a local navigation system for unmanned aerial vehicles (UAV) using ultra bandwidth (UWB) technology. Proposed navigation can be implemented to an automatic landing system on the deck of a marine moving platform, which use autonomous on-board UAV in difficult weather conditions (fog, drizzle). This system ensures sufficiently large range, accuracy, accessibility and low price for local navigation. A Kalman filter is used to estimate the position and speed of the UAV relative to the moving platform, and a nonlinear controller has been synthesized to control the movement of the UAV, which allows for a smooth and accurate approach of the UAV to the moving platform. At the same time, the specified controller generates the desired orientation angles of the UAV as control actions, which allows it to be used as an add-on above typical flight controllers. The results of simulation have confirmed the feasibility and effectiveness of the proposed local navigation system.

Hybrid Extended State Observer with Enhanced Sliding Mode Control for Tracking and Disturbance Rejection for Mecanum Wheeled Mobile Robots Applied in Automotive Manufacturing

huytran25 — 2025-10-24

This paper introduces an Adaptive Sliding Mode Control (ADSMC) framework integrated with an Extended State Observer (ESO) to enhance the robustness of a Mecanum Four Wheel Autonomous Vehicle (MFWAV) for automotive intralogistics. where mobile robots are subjected to external disturbances and model uncertainties that degrade tracking accuracy. The proposed control law is systematically derived from the robot’s nonlinear dynamic model through the construction of an appropriate Lyapunov candidate function, ensuring closed-loop stability under varying operating conditions. The ADSMC–ESO scheme is designed to provide robust disturbance rejection and adaptive compensation for parametric variations while maintaining high tracking precision, offering improved adaptability and robustness for modern automotive manufacturing.

Path planning for smooth operations of nonholonomic mobile robots using piecewise-uniform turning maneuvers

Minh-Tuan Nguyen-Thai — 2025-10-24

Path planning for smooth operations is important for nonholonomic mobile robots to operate safely and efficiently. This study proposes an approach to design a piecewise-uniform path planning. The whole plan of the path is divided into smaller segments. Either the linear velocity or the linear acceleration of each segment is constant. Similarly, either the angular velocity or the angular acceleration of each segment is constant. Because of the uniformness, the limitations on the velocities and accelerations can be easily satisfied. This approach can make use of any path planning algorithm that outputs a path consisting of straight lines, for example, rapidly-exploring random tree star algorithm and A star algorithm. Piecewise-uniform turning maneuvers are then added to the raw path. They include not only the turns from one straight segment to another but also a smooth starting curve and a smooth stopping curve to satisfy the robot’s pose at the beginning and the required pose at the end of the path. Two examples are presented, showing that the proposed method successfully generates appropriate paths and reduces the operation time of the mobile robot.

Numerical Solution to Estimate the Transfer Function Describing the Towed Underwater Vehicle

Nguyen Trong Tuyen — 2025-10-24

Synthesizing the control system for objects with distributed parameters is carried out in two steps: first, a model with centralized parameters is established instead of the original object, then the synthesis of the controler is performed using well-known methods for linear systems. Therefore, this article will propose and investigate a solution estimating the transfer function (TF) describing the towed underwater vehicle (TUV) by an approximate model with the required error based on the real interpolation method (RIM). This solution allows manipulation of real numbers, which overcomes the limitations of previous estimation methods. It reduces the amount of calculation and allows to obtain an approximate model of arbitrary order without any difficulty. In addition, the proposed algorithm can be easily realized on computers.

This study includes main contents: establishing the model of the Towed cable (TC)- Underwater vehicle (UV) system, proposing an estimation algorithm and building a program to automatically estimate the TF, which associates between displacement at the end of cable attached to the UV with displacement at the point attached to the winches, evaluating the impact of sea waves on the towed underwater vehicle.

Design of Synergetic Controller based on Nonlinear State Observer for Twin Rotor MIMO System

hoangduclonglqdtu — 2025-10-24

This paper presents the design of a robust controller for the Twin Rotor MIMO System (TRMS) by integrating Synergetic Control Theory (SCT) with a Nonlinear State Observer (NSO). TRMS is a benchmark system widely used to evaluate advanced control algorithms due to its highly nonlinear dynamics, strong coupling, parameter uncertainties, and susceptibility to external disturbances. Conventional controllers often struggle to ensure both stability and precise trajectory tracking under such conditions. To overcome these challenges, the proposed method leverages SCT to guarantee convergence of pitch and yaw angles to desired trajectories while preserving system stability in the presence of unknown disturbances. Since synergetic control laws require full state information, a nonlinear state observer is developed to estimate unmeasured states with high accuracy. The stability of both the controller and observer is rigorously proven using Lyapunov functions. Simulation results demonstrate that the combined SCT–NSO approach effectively eliminates disturbances, achieves fast convergence, and ensures accurate trajectory tracking, outperforming traditional methods. The proposed framework provides a promising solution for robust control of complex nonlinear systems such as TRMS and paves the way for further extensions to terminal synergetic control designs.

Research of Nonrecursive Federated Filtering Algorithms Under Non-White Noise Measurement Errors

Yulia Litvinenko — 2025-10-24

The paper considers the design of the algorithms estimating the dynamic system state under the presence of non-white noise measurement errors. The estimators can be based on centralized processing, when all available measurements are fed to one centralized computer, or on decentralized schemes. The main advantage of centralized architecture is its ability to provide mean-square optimal estimate of the linear system state using the linear measurements with Gaussian system noise and measurement errors, however, this architecture features low reliability and high computational load.

The described decentralized processing methods are based on federated filtering algorithms (FFA), where the state vector of dynamic system is estimated by weighting the estimates of the local filters (LF), processing the local measurements, in the master filter (MF). Decentralized methods based on FFA have become widespread and also often applied to navigation of different robot-aided systems. The federated filtering algorithms are computationally simpler and immune to false measurements, however, they generally fail to provide optimal estimates, and the calculated accuracy characteristic is not a real error covariance matrix of the generated estimate.

The proposed FFA is based on nonrecursive processing of measurements in LF. It has been shown that when LF tuning conditions are met, the MF estimates and covariance matrices coincide with the estimates and covariance matrices for the optimal centralized Kalman filter. It has been noted that the use of a nonrecursive measurement processing scheme creates a good background for applying factor graph optimization (FGO) methods in the problem of nonlinear measurements processing using FFA.

The obtained results are illustrated by the example of navigation system correction.

Robust control for line-following robots with model uncertainties and dynamic disturbances

xuanchiem — 2025-10-24

This paper proposes a robust nonlinear control law for line-following robots under dynamic disturbances caused by model uncertainties. The control objective is to ensure that the robot accurately tracks the desired trajectory, even in the presence of line paths with varying curvature. The control strategy is designed based on a cascade structure, consisting of two loops: the kinematic loop and the dynamic loop. The outer loop, corresponding to the kinematic controller, is developed using a Lyapunov function. The dynamic control law in the inner loop is derived from differential geometric methods and guarantees finite-time stability. System uncertainties and disturbances are compensated by introducing a predefined term in the control law. The closed-loop stability is rigorously proven using Lyapunov theory. Simulation results, along with comparisons to the LQR controller, demonstrate the effectiveness and robustness of the proposed approach, highlighting its fast convergence, small tracking error, and strong disturbance rejection capability.

Development of a Method for Synthesizing Disturbance Identification Systems for Unmanned Aerial Vehicles

proa2931 — 2025-10-24

This paper proposes a method for synthesizing a disturbance identification system for quadcopter-type unmanned aerial vehicles (UAVs) that allows determining the magnitudes of force and moment disturbances acting on all degrees of freedom of the UAV and caused by the uncertainty of the parameters and the presence of unknown external influences. We present a complete mathematical description of the UAV dynamics and kinematics taking into account the aerodynamic drag, as well as the uncertainties of the parameters and the influence of the environment, which are represented in the dynamic equations by additive unknown functions which are subject to estimation. To determine the values of these unknown functions in real time, sliding mode observers are proposed. The disturbance data can then be used by adaptive control systems to compensate for their negative effects and improve the accuracy of motion. Numerical simulation is performed, the results of which confirm the operability and effectiveness of the proposed solutions. The efficiency and accuracy of observer estimates are illustrated for several scenarios, including the movement of the UAV to a given point and along a predetermined trajectory.

Estimation of Coordinates and Motion Parameters of an Accompanied Vessel Based on Factor-Graph Optimization

Vladislav Karaulov — 2025-10-24

The signal from a noise-emitting source propagates due to the phenomenon of sound refraction in the marine environment along various trajectories (rays) and is received by the antenna array of an autonomous underwater vehicle from a certain direction (bearing) at different elevation angles. To reduce the uncertainty region of the target coordinates, the paper uses the values of signal correlation across elevation angles, which are compared with reference data computed on a grid of distances and formed based on a ray representation model of the marine environment field. Information on hydroacoustic conditions in the observation area, including measurements of the vertical sound speed distribution, is used as input data for the field model. The proposed solution is based on a combined approach, the essence of which is that the problem of estimating the coordinates and motion parameters of an accompanied vessel from bearing measurements and correlation values of received signals is formulated within a Bayesian framework, and factor-graph optimization methods are used to construct suboptimal algorithms. To evaluate the effectiveness of the proposed algorithm, a simulation program was developed and used to conduct predictive simulation modeling of the considered problem on various episodes and under different hydroacoustic conditions. The results obtained were compared with a non-recursive algorithm constructed within the Bayesian approach. An analysis of the algorithms' consistency was conducted. It is shown that the algorithm based on factor-graph optimization has a significant advantage in computational cost. The results obtained were compared with a non-recursive algorithm. An analysis of the algorithms' consistency was performed. It is shown that the algorithm based on factor-graph optimization has a significant advantage in computational costs.

Recursive Batch Smoother with Multiple Linearization for Single-Beacon Navigation Problem

alexey_isaev — 2025-10-24

A problem of single-beacon navigation of an autonomous underwater vehicle (AUV) is considered. It is solved in terms of the Bayesian approach and belongs to the class of problems with essential nonlinearity, the posterior probability density function (PDF) of which changes from multi-extremal to single-extremal during its evolution. As a rule, traditional recursive Kalman-type algorithms do not work in this case. To solve the single-beacon navigation problem, an algorithm based on the combined use of recursive and nonrecursive data processing schemes, called a recursive batch smoother with multiple linearization, has been designed. Its main essence lies in the simultaneous use of a bank of batch algorithms – recursive iterative smoothing linearized filters, which allows identifying the point of time at which the PDF becomes single-extremal. After this moment is identified, it becomes possible to switch from the problem solution using a bank of batch algorithms to the solution with a single recursive iterative Kalman filter, thus decreasing the computational complexity. The possibilities of reducing computational complexity by applying methods used in factor graph optimization algorithms are discussed. The advantages of the proposed algorithm are shown by the simulation.

A method for industrial manipulators for approximating paths with an excessive number of fly-by points

Ivan Maksimovich Grigorev — 2025-10-24

Computer vision systems are often used to automatically generate paths for the working tool of an industrial manipulator. However, such paths often contain an excessive number of points and small random noises and cannot be directly used to set the movements of the manipulator. The article proposes a method for approximating such paths by a sequence of straight lines and circles to reduce the number of commands to the manipulator and to mitigate noise. The method is based on search through possible variants for dividing the path into segments and comparing the total errors when fitting a straight line or circle into each of the segments. The article provides three ways to search through combinations: a complete search, the fastest search and an intermediate variant that provides the user with a choice between the speed of the search and the probability of finding the optimal combination. In addition, equations are given for calculating the parameters of straight lines and circles, as well as for estimating the approximation error; a method for calculating the endpoints of straight line segments and circular arcs is shown. The efficiency of the method was tested on the path obtained from a profile laser scanner.

Aerodynamic Characteristics and Dynamic Modeling of Long Endurance Unmanned Aerial Vehicle

thehungmfti — 2025-10-24

This paper investigates the aerodynamic characteristics and dynamic behavior of a long endurance unmanned aerial vehicle (UAV). The aerodynamic parameters were analyzed using simulation tools including Ansys Fluent, Missile Datcom, and OpenVSP. The results indicate that the UAV achieves optimal flight efficiency within the angle of attack range of 4°– 6°, with stall occurring at 12°. Based on these findings, a longitudinal dynamic model was developed by the classical method with motion equations to evaluate the influence of aerodynamic and dynamic parameters on flight stability. The study demonstrates that the UAV maintains controllability and reliability under various flight conditions. These findings provide practical implications for UAV design, manufacturing, and improvement, thereby contributing to the foundation for mastering and advancing modern aviation technology in the future.

Comparison of methods for covariance estimation for factor graph optimization of the Bayesian estimation problem

Dmitrii Alexandrovich Cherginets — 2025-10-24

Factor graphs optimization is becoming a common approach to states and parameters estimation in navigation tasks due to the simplicity of the system description and combining various measurement sources, as well as the high accuracy of estimates. To obtain the best estimate, it is necessary that the noise covariance matrices of the corresponding measurements are known a priori. However, in practice, this is very difficult to achieve due to the complexity of methods for estimating noise parameters and the non-stationarity of these parameters, which may require the evaluation or correction of the covariance matrices used. Recently, several approaches have been presented to obtain an estimate of the covariance matrix in relation to factor graph optimization, but no comparison has been made with other approaches. This paper is devoted to comparing two methods for estimating measurement and system noise covariance matrices. First is a factor graph method where both covariances and state vector are estimated using algorithms based on factor graph, second is a combined Bayesian – factor graph method where a sequential nonlinear Bayesian algorithm is used to estimate covariances and factor graph is used to estimate the state vector. The following metrics were used for comparison: the mathematical expectation and the standard deviation of the estimates of covariance, and a coefficient of accuracy of the system state estimates with the obtained estimates of covariance.

Generation of synthetic images to expand datasets for computer vision systems used on robotic conveyors

Vitaly Ivanovich Bondarenko — 2025-10-24

The paper presents a comprehensive approach to expanding datasets using synthetic image generation for advanced computer vision systems applied to robotic sorting in automated glass product quality control systems. The proposed method integrates synthetic data generation with deep learning models for defect detection in glass container production. The system combines composite rendering using the Pillow library and automatic annotation generation in COCO format to create photorealistic variations of original objects with controlled variability in lighting and background conditions. The defect generation module integrated into the computer vision system realistically simulates production defects in glass containers, including chips, cracks, microbubbles, scratches, label wear, and shape deformations, significantly enhancing data variability and realism. Experimental validation demonstrated the method’s effectiveness for generating balanced datasets for training robust computer vision models. A cascaded approach using YOLO for rapid region of interest extraction, EfficientDet and RetinaNet for precise defect identification, and Vision Transformer for verification achieves high accuracy in real-time industrial conditions. The system successfully generated 1000 synthetic images in approximately 15 seconds, maintaining 100% annotation accuracy and area deviation of less than 1%. This approach demonstrates high performance and is designed for integration into a comprehensive computer vision system architecture using cascaded models (YOLO, EfficientDet, Vision Transformer) for real-time quality control on production lines, effectively addressing class imbalance and data scarcity problems characteristic of industrial applications.

Distributed Swarm Robot Control Using Fuzzy Logic, Chaos Theory, and the Dragonfly Algorithm

hongnt — 2025-10-24

When swarm robots move in an environment with obstacles to search for an unknown target, the challenge is that each individual robot must be able to maintain swarm cohesion, avoid obstacles, and achieve the widest possible coverage of the search space. Therefore, this paper proposes a solution that applies fuzzy logic to determine the steering angle of robots in the swarm, which helps adjust collision avoidance behaviors and regulate robot distribution within the environment. At the same time, the paper integrates chaos theory and the Dragonfly Algorithm (DA) to improve the quality of swarm control, enabling the robots to move more flexibly, avoid obstacles smoothly, and achieve the best possible coverage of the target search space during the entire process.

A Fuzzy logic control for tactile antenna-based wall-following in autonomous robots

lacduong — 2025-10-24

Currently, autonomous robotic systems primarily rely on vision for environmental perception, while their tactile sensing and interaction capabilities remain limited. This study introduces a bio-inspired tactile antenna designed to enhance physical interaction and environmental awareness in autonomous robots. A fuzzy logic control strategy is developed to enable robust wall-following behavior along both straight and curved surfaces. The proposed approach is validated through simulations and physical experiments, which demonstrate that the tactile antenna significantly improves navigation performance and adaptability in complex environments.

Optimizing the Trajectory of Robotic Manipulator: Reinforcement Learning for the Generation of Initial Guess

Elena Rubleva — 2025-10-24

This paper describes the challenge of trajectory planning for robotic manipulators operating in obstacle-rich environments. The goal is to propose a novel approach that integrates the TrajOpt algorithm with reinforcement learning (RL) to improve the efficiency and accuracy of motion planning. TrajOpt, a numerical optimization-based method, generates successful trajectories, while RL produces high-quality initial trajectory estimates, serving as strong starting points for optimization. This integration enables TrajOpt to find feasible paths more efficiently and reduces collision risks. RL further refines initial trajectories, enhancing navigation in complex environments. Experimental results highlight the benefits of merging machine learning with traditional optimization methods. The RL-based approach achieved an 82% success rate in navigating obstacle-dense environments, outperforming the classical RRTConnect method, which achieved 68%. The proposed system improves planning efficiency and reliability in complex scenarios with obstacles by using reinforcement learning (RL) alongside TrajOpt. This research offers valuable insights into improving robotic manipulator performance and lays a foundation for future advancements in motion planning. It demonstrates that integrating reinforcement learning with existing algorithms can drive significant progress in addressing the complexities of navigating obstacle-laden environments, opening new directions for research and development.

Robust Sliding Mode Control for the Electro-Optical Observatory Drive System of a Patrol Ship

pageplan33286 — 2025-10-24

. Electro-optical tracking devices play a crucial role in modern civil monitoring and observation systems, such as coastal surveillance, maritime navigation assistance, and environmental monitoring. However, the tracking performance of their drive systems can be significantly degraded by external disturbances such as ocean waves, platform vibrations, and environmental uncertainties. To address these challenges, this paper proposes a robust sliding mode control (SMC) strategy integrated with a disturbance observer for the electro-optical observation drive system of a maritime monitoring platform. The proposed controller is designed to enhance robustness against nonlinearities and external disturbances while ensuring high tracking accuracy. A stability analysis is provided to guarantee convergence of the tracking error under bounded disturbances. The effectiveness of the approach is validated through numerical simulations in MATLAB/Simulink, where the proposed SMC demonstrates superior performance compared to conventional control schemes. Results confirm that the controller ensures precise tracking and robust disturbance rejection in realistic maritime conditions, making it a promising solution for improving the operational reliability of shipborne electro-optical observation systems.

Fast Path Planning with Hierarchical Approach Based on 3D Scene Graphs

Xuan-Tung Truong — 2025-10-24

In the literature on autonomous systems in general, and mobile robots in particular, the primary goal of path planning methods is to meet both com pleteness and real-time criteria, ensuring they can be effectively deployed in real-world systems. For Unmanned Aerial Vehicles (UAVs), the requirement for real-time performance becomes even more critical due to constraints of resources.
The most of existed methods use the volumetric map combined with a single-level planning approach. This is computationally expensive even in small scenes. There were many efforts in reducing this burden of computing. The ESDF map encodes the shortest distance from every point in the environment to the nearest obstacle enabling collision-checking faster. Oleynikova et al. has tried to skeletonizing the environment and then creating sparse graph. This work suggests the use of hierarchical planning based on the novel presentation of environment, namely 3D Scene Graph (3DSG), to save computational resources more. Hierarchical path planning is an approach that divides the path finding process into different levels of abstraction, allowing for more efficient and scalable solutions. The 3DSG is inherently suitable to hierarchical planning because it is organized as multi-layers at different levels of abstraction and includes the relationship between nodes at the same layer as well as across different layers.
Moreover, we also did the experiments for illustrating the performance of planning on 3DSG versus sparse graph. We do A-star and D* Lite with the same configuration on 3DSG and sparse graph on some kind of environment. From the result, we figured out that the efficiency of hierarchical planning on 3DSG is superior compared to sparse graph. A summary video of the proposed system is available at https://youtu.be/6leSRoO16KQ.

Toward Collaborative Robots: A Novel Hybrid Slider–Crank Mechanism with a Soft Variable-Stiffness Connecting Rod

hiep_trinh — 2025-10-24

Robots designed for human interactions require not only precision
but also adaptability and safety during physical contact. Traditional
rigid robots perform well in structured industrial environments
but are limited in collaborative applications because of their inherent
lack of compliance. Soft robots, by contrast, offer intrinsic safety and
adaptability but struggle with accurate modeling and controllability. To
address these limitations, this paper proposes a novel design strategy
for robotic mechanisms that integrates rigid and soft components to exploit
the advantages of both principles. This strategy is demonstrated
through a hybrid slider–crank mechanism incorporating a soft, variablestiffness
connecting rod into the conventional rigid framework. The soft
link, constructed as a pneumatic pressure–responsive elastomer composite,
allows active stiffness modulation: elevated pressure increases rigidity
for accurate force transmission, whereas reduced pressure enhances
compliance for safer interaction. A theoretical model of the mechanism
is developed to describe stiffness variation, dynamic force transmission
under different pressurization levels. The design and fabrication process
of the soft link is presented, followed by indentation tests to characterize
pressure-dependent stiffness and dynamic experiments to evaluate
force transmission in the hybrid mechanism. Experimental results confirm
that pneumatic pressurization effectively tunes the stiffness of the
connecting rod, leading to controllable changes in dynamic response and
output force. This work demonstrates the feasibility of embedding soft
variable-stiffness elements into core structural components of classical
mechanisms, paving the way for collaborative robots that combine precision,
adaptability, and safety.

Actuator Faults Detection and Isolation Based on Nonlinear Observer for Quadcopter

thiempham — 2025-10-24

This work presents a simplified FDI strategy tailored to the quadcopter’s nonlinear dynamics. The approach employs a nonlinear observer to reconstruct system states and generate residuals by comparing observer-based estimates with measured angular rates. Fault-sensitive decision statistics are computed over finite sample windows and evaluated against adaptive thresholds to detect and isolate loss-of-effectiveness (LOE) actuator faults. Simulation studies demonstrate that the proposed method achieves accurate and timely fault detection and correct isolation over a range of fault severities, while maintaining low computational overhead. These results indicate that the observer-based FDI scheme offers a practical solution for onboard implementation in resource-constrained quadcopter platforms.

Cooperative Optimal Multi-Robot Formation Arrangement for Object Pushing

23025027 — 2025-10-24

This paper presents a method for coordinating and optimizing multi-robot formations to push objects along predefined trajectories under frictional constraints. A Particle Swarm Optimization (PSO) framework is proposed to determine the optimal contact point positions on the object boundary. The optimization combines criteria derived from grasp and Gram matrix analysis to generate a wide range of wrench to handle diverse manipulation tasks and uniform distribution of pushing directions. These are integrated into a into a comprehensive cost function that guides the swarm toward stable and controllable configurations. Based on the selected positions, a simple PID-based controller is employed to generate the desired wrench for path tracking. Simulation results demonstrate that the strategy has produced suitable contact point locations and provided efficiency and stability in the process of pushing the object. The combined approach offers a practical and computationally efficient framework for pushing-based manipulation, with potential applications in robotic handling of objects that cannot be grasped.

An Efficient Navigation Algorithm for Unmanned Surface Vehicles in Dynamic Environments

Xuan-Tung Truong — 2025-10-24

Unmanned surface vehicles (USVs) are increasingly being deployed for tasks in complex maritime environments such as harbors, coastal monitoring, and port patrols. Safe navigation in congested, dynamic waters requires an integrated approach to perception, mapping, path planning, and motion control. In this paper, we present a fully integrated navigation system for a catamaran-style USV operating in a busy harbor scenario. The proposed system closes the loop from environmental sensing to control actuation. It fuses data from GPS, IMU, 3D LiDAR, and camera sensors to perform simultaneous localization and mapping (SLAM) and obstacle detection in real time. A tightlycoupled LiDAR-inertial odometry algorithm (LIO-SAM) provides robust localization within a feature-rich environment, while an Extended Kalman Filter (EKF) combines GPS/IMU data for improved odometry. For path planning, a global planner based on the A* algorithm computes efficient routes on an occupancy map, and a local planner using the Dynamic Window Approach (DWA) refines the trajectory to avoid dynamic obstacles. A twoloop PID controller executes smooth motion control of the USV’s linear and angular velocities. We validate the system in a high fidelity Gazebo simulation of a crowded harbor, demonstrating that the USV can safely navigate to target waypoints while avoiding static and moving obstacles. The results highlight that the integration of SLAM with global A* planning and local DWA obstacle avoidance significantly enhances autonomous navigation capabilities. The proposed framework emphasizes system-level efficiency and is
extensible to incorporate advanced vision-based perception (e.g., water segmentation and object detection) to further improve safety in dynamic environments. This work contributes an efficient navigation architecture for USVs and provides insights for future field deployment in real maritime environments.

Variational Autoencoder for Efficient Image Representation in Deep Reinforcement Learning for Mobile Robot Navigation

Long Phi Nguyen — 2025-10-24

In this paper, we present a deep reinforcement learning (DRL) policy for autonomous mobile robot navigation in cluttered environments. A key challenge in applying DRL to vision-based navigation is the high dimensionality of raw image observations, which often leads to poor sample efficiency and unstable training. To address this issue, we propose a representation learning module based on a variational autoencoder (VAE), trained on simulated depth images to compress visual input into a compact latent representation. This encoding preserves essential spatial information such as obstacle boundaries and free space, while filtering out redundant details, resulting in a more structured and informative observation space. The latent features, combined with odometry data, are then used as input to the DRL policy, enabling the robot to reliably reach specified targets while avoiding collisions. Extensive simulation experiments in diverse and cluttered environments demonstrate that our method achieves a success rate of approximately 66.7%, which is significantly higher than baselines trained directly on raw depth images or with FFT-based encoders under the same training budget. These results highlight the effectiveness of combining VAE-based representation learning with DRL for robust and efficient autonomous navigation.