ICR 2025

Mutual Adaptation Model of Operator and Controlled Object in Ergatiс Robotic System

2025-05-07T07:57:24+00:00

The presented paper deals with the problem of developing a model of mutual adaptation of the operator and the controlled object in an ergative robotic system. The simplest collaborative robotic system is taken as a robotic system, the participants of which are a manipulator type robot and an operator. Mutual adaptation in the control process is considered on the example of a system with service discipline and application of an alternative approach. An example of solving the problem of adapting the service discipline of a mass service system under conditions of unpredictability of the external environment, which inevitably change the optimal setting of the service discipline, is given. The main components of the developed model of the system with adaptive control on the basis of neural network are given. The basic structure of this system is proposed. The paper provides an overview of work in the field of robot control and applications for robotic systems using neural networks.

Evaluation of a weather plugin in Gazebo: a case-study of a wind influence on PX4-based UAV performance

2025-05-07T07:57:25+00:00

Unmanned aerial vehicles (UAVs) are utilized in a wide range of real-world applications including search and rescue missions, 3D mapping, building inspection and reconstruction. In a real-world environment, severe weather conditions significantly impact UAV operations. Wind is one of the most dangerous weather factors that can destabilize and even destroy a UAV. Preparing UAV flight algorithms for real-world conditions in order to minimize risks of damage and malfunctioning is a crucial part of a UAV system development. This paper analyses a wind influence on UAV virtual model performance within the Gazebo simulator. The paper focuses on gazebo_wind_plugin package used for creating a windy virtual environment. PX4-LIRS package is used for simulating a UAV within the Gazebo. Three types of experiments were run to evaluate capabilities of the plugin to simulate typical issues that arise in real-world scenarios: a UAV wind resistance, a wind effect on a UAV tilt angle and on a flight velocity. Virtual experiments demonstrated that gazebo_wind_plugin could serve for PX4 UAV flight scenarios modeling in Gazebo simulation with a satisfactory level of realism.

Proposed calculation of H-bridge switching law in speed con-trol of the single-phase asynchronous motor in PWM mode

2025-05-07T07:57:26+00:00

This article proposes a calculation plan to control H-bridge inverter switching based on changing the law of distribution of non-zero voltage states according to the power law ratio, 0 voltage states according to the division law. uniformity in speed control of a single-phase asynchronous motor in H-bridge inverter switching pulse width modulation. The results are surveyed, analyzed, compared and evaluated using simulation models on software. Matlab-Simulink. In particular, the fundamental harmonic level index and the total amount of higher harmonics (THD) are used to compare, analyze and evaluate the quality of H-bridge inverter output voltages in different switching modes, simultaneously. Meanwhile, evaluate the influence of the zero and non-zero voltage distribution rules of the proposed plan with the SinPWM pulse width modulation technique on the speed oscillation and the speed value set at the voltage frequencies. AC pulse H-bridge inverter output with different average voltage levels and pulse hashing frequencies.

Augmented Reality Interface for UR5e Robot that Transfers Parts to a Human in Collaborative Assembly

2025-05-07T07:57:26+00:00

Collaborative assembly is one of the most common tasks in collaborative robotics, which requires an efficient and user-friendly way of communication with a robot. This paper presents an augmented reality (AR) interface for controlling a UR5e robot in order to get assembly parts from a robot’s desktop in a collaborative assembly process. This interface is based on the concept of a smart mirror that recognizes hand gestures as control signals. A goal of the study is to investigate user perception of a contactless interaction method with the collaborative robot UR5e using this AR interface. 15 participants took part in experiments. During the experiments, the participants demonstrated a quick adaptation and a big interest to working side by side with the UR5e robot using the AR interface. Our results suggest that the developed AR interface is flexible, scalable in terms of development and can be recommended for equipping a workcell for collaborative assembly tasks of different products.

User Perception and Evaluation of a Portrait Drawn by KUKA KR3 AGILUS Manipulator

2025-05-07T07:57:28+00:00

Creating an artistic portrait is considered one of the most complex areas of a generative art. Currently, generative artificial intelligence (AI) has made significant advancements in generating portraits of people based on photographic images. This has opened up new opportunities for utilizing AI achievements in robotic artistry. However, questions related to evaluation and human perception of such works, including their artistic value, remain open. In our work, we present experimental results of integrating the processing of human photographs using the generative model InstantID with a drawing robot and studying users' reactions to the obtained images. In our experiments, we generated graphic portraits of people and constructed their images using KUKA KR3 AGILUS robot. We then asked participants to rate the resulting portraits based on three parameters: similarity/recognition, beauty/aesthetics, and evoked emotions. The results of the experiments can be used in social robotics, including art therapy that involves social robots and human-robot interaction.

Tracking of Mobile Objects with an UAV and a DNN Controller

2025-05-07T07:57:29+00:00

The success of the Ingenuity mission on Mars underscores the growing need for autonomous navigation technologies in UAVs. However, detecting and tracking moving objects with unknown dynamics remains a challenge in planetary exploration. Current optimal control algorithms outperform classical controllers but struggle to generate control signals within the required operating time, leading to high computational costs. We propose a Deep Neural Network (DNN) architecture pre-trained with a Model Predictive Control (MPC) for horizontal motion control, coupled with a Proportional-Integral-Derivative (PID) controller for altitude and orientation of a UAV for mobile target tracking. This approach reduces computational costs, significantly improves the speed of control signal generation, and maintains performance similar to MPC. Control commands are computed from the estimated trajectory using visual information and UAV states. Testing is conducted in Gazebo with the Parrot Bebop 2.0 and Husky Robot.

NavTopo: Leveraging Topological Maps For Autonomous Navigation Of a Mobile Robot

2025-05-07T07:57:30+00:00

Autonomous navigation of a mobile robot is a challenging task which requires ability of mapping, localization, path planning and path following. Conventional mapping methods build a dense metric map like an occupancy grid, which is affected by odometry error accumulation and consumes a lot of memory and computations in large environments. Another approach to mapping is the usage of topological properties, e.g. adjacency of locations in the environment. Topological maps are less prone to odometry error accumulation and high resources consumption, and also enable fast path planning because of the graph sparsity. Based on this idea, we proposed NavTopo – a full navigation pipeline based on topological map and two-level path planning. The pipeline localizes in the graph by matching neural network descriptors and 2D projections of the input point clouds, which significantly reduces memory consumption compared to metric and topological point cloud-based approaches. We test our approach in a large indoor photo-relaistic simulated environment and compare it to a metric map-based approach based on popular metric mapping method RTAB-MAP. The experimental results show that our topological approach significantly outperforms the metric one in terms of performance, keeping proper navigational efficiency.

Stabilizing manipulator trajectory via collision-aware optimization

2025-05-07T07:57:31+00:00

In this work, we develop an optimization-based solution for a manipulation planning among obstacles. This task is particularly challenging for collaborative manipulators when the operations include movements through the singular configurations. Trajectory optimization for this case requires a collision model, which allows converging to a valid solution from an invalid initial guess. By improving the initial trajectory approximation and optimizing collisions with advanced obstacle representation, it significantly enhances trajectory planning accuracy and efficiency. In our work, we propose an approach based on the combining of obstacles. We use Octomap as the baseline (obstacles are represented as a set of cubes). This creates the need to calculate each cube when calculating the trajectory. Enlarging cubes to parallelepipeds allows one to obtain a collision-free trajectory faster. We have conducted a set of experiments with Octomap and with the representation of obstacles in the form of enlarged cubes. Experiments have shown that enlarging cubes to parallelepipeds reduces planning time and increases the success rate.

Vectorized visibility graph planning with neural polygon extraction

2025-05-07T07:57:32+00:00

This article proposes an approach for path planning in environments with obstacles. The methodology integrates recent neural obstacle polygon extraction with visibility graph path planning, complemented by the integration of vectorization techniques. That significantly enhances path planning efficiency. The modular design of the neural network method, encompassing contour detection, vertex identification, and polygon approximation modules, facilitates improved performance compared to traditional methods. Furthermore, the investigation into vectorization's impact on the intersection operation accelerates algorithm speed, contributing to faster path planning processes. Experimental results validate the efficacy of the integrated approach, showcasing notable improvements in path planning efficiency, especially with the utilization of vectorization techniques. The study systematically addresses challenges such as slow graph construction and inaccurate obstacle detection, providing a robust solution for optimizing path planning processes. Moreover, the implementation of multiple modules in the methodology enables its versatility for testing with various environments. This versatility allows researchers to assess the method's performance across diverse scenarios and visualize the results effectively. Overall, the integrated approach offers a comprehensive solution for optimizing path planning in complex environments, demonstrating its potential to streamline path planning processes and improve mobile robot navigation.

Optimal Control Problems in Collaborative Multi-Agent Robotic Systems

2025-05-07T07:57:33+00:00

The paper considers human-robot collaboration from the control point of view as the joint activity of robots and people in a single multi-agent system to achieve the control goal with the optimal value of a given quality criterion. The paper proposes a mathematical statement of the control problems for such complex plant. In the proposed formulation, the role of a person is described in the form of an additional control vector, with the help of which the influence of phase constraints is reduced in terms of functional minimization. So, the optimal control and the control general synthesis problems for collaborative systems include additional control vector in phase constraints. The problems are formulated for a single collaborative cell and for multi-agent collaborative system. The problem statements for multi-agent systems include also dynamic phase constraints to avoid collisions between robots. The example of optimal control problem for two mobile robots with two controlled phase constraints is presented. For solution of the problem the method of the synthesized optimal control was used.

Development of a dynamic model of an underwater manipulator in identification form

2025-05-07T07:57:34+00:00

The paper proposes a new method for constructing dynamic models of underwater manipulators installed on board autonomous underwater vehicles. These models are designed to identify unknown parameters of the interaction of an underwater manipulator with a viscous medium during its movement. A feature of these models is that unknown parameters, such as coefficients of added masses and damping coefficients, are included in the equations of dynamics linearly with respect to known kinematic coefficients. The correctness of the proposed models and the possibility of identifying these coefficients were verified during numerical simulation of the movement of an underwater vehicle with a manipulator mounted on it at various initial parameters of a viscous medium.

Decentralized Unlabeled Multi-Agent Navigation in Continuous Space

2025-05-07T07:57:36+00:00

In this work, we study the problem where a group of mobile agents needs to reach a set of goal locations, but it does not matter which agent reaches a specific goal. Contrary, in this work we focus on the decentralized scenarios, when each agent acts individually relying only on local observations/communications and is free to move in arbitrary direction at any time. Our iterative approach involves agents individually selecting goals, exchanging them, planning paths, and at each time step choose actions that balance between progressing along the paths and avoiding collisions. The proposed method is shown to be complete under specific assumptions on how agents progress towards their current goals, and our empirical evaluation demonstrates its superiority over a baseline decentralized navigation approach and a comparison with the centralized TSWAP algorithm reveals its efficiency in minimizing trajectory lengths for mission accomplishment.

Method for Maximizing the Number of Detected Keypoints on Homogeneous Underlying Surfaces

2025-05-07T07:57:36+00:00

RGB images are commonly used as input data for image processing systems, but, the informative value of RGB visualization can be insufficient, especially in situations where the area to be processed is homogenous and a mixture of terrain objects fall on a single pixel of the image. Spectral data is essential for effective processing homogenous areas of the underlying surface. In this paper, the authors demonstrate the advantages of using the spectral indices visualization as an input for keypoint detection and matching tasks. Position shift detection problems require process of finding key points in the image. However, in such tasks, it is not always possible to find a sufficient number of keypoints using only RGB data. This problem appears when you trying to process homogeneous types of surfaces such as forests, deserts, and water spaces. The Main task of this paper was to develop a method based on a combination of spectral channels to achieve an increase in the number of keypoints detected on a pre-defined homogeneous underlying surface. The results of the conducted study showed that the number of detected keypoints on average increased 1.44 times in the calculated index images compared to RGB images and, in turn, resulted in an average 1.592-fold increase in the number of key point matches in the current and previous images.

Overcoming obstacles with a home walking robot

2025-05-07T07:57:37+00:00

The paper considers the principles of building a home walking robot capable of walking on complex surfaces – stairs and ladders. A home walking robot must walk confidently on various flat surfaces and be able to step over small obstacles. To detect obstacles and scan the surface, the robot has a simplified scanning system based on inexpensive TOF sensors for measuring distances. A mathematical model of the robot’s legs, which is used to control the drives, is presented. A model of the robot’s sensor system is presented, which is used to detect obstacle parameters and build a height map. Two ways of overcoming obstacles are considered – stepping over small obstacles and walking on wide stairs. An estimate of the maximum size of obstacles is given, which are determined by the geometric dimensions of the robot’s legs and the parameters of the sensor system. The principles of walking on narrow stairs are formulated. Modeling of the proposed ways to overcome obstacles is performed. The simulation showed acceptable stability of the robot when walking. It is shown that the proposed methods of overcoming obstacles have low computational complexity. This reduces the computing power requirements of the home walking robot control system and increases battery life.

Adaptive Fast Terminal Sliding Mode (FTSM) Control and High Gain Observer (HGO) for Multi-Motor Web Transport Systems

2025-05-07T07:57:38+00:00

The research object of the article is the web winding system, which has a complex MIMO structure, has variable parameters and is affected by noise. In reality today, the controllers of this system mainly use linear control methods. However, linear control methods do not always provide the desired quality. Therefore, research, development and application of nonlinear algorithms to improve system quality is essential. A substantial amount of developments in the field of regulators for complex systems with variable parameters has been realized in the field of robotics and manipulation systems. This paper presents the results of synthesizing the adaptive fast terminal sliding controller based on the neural network for the web winding system for elastic material. The survey and evaluation results by simulation on Matlab-Simulink software show that this controller ensures the quality requirements even when the system is affected by random input noise and model parameters change. The application of the developed solutions in robotics will improve the quality of control and increase the efficiency of mechanical systems.

Motion Control of The Mobile Robot with Propellers which Working on the Periodic Jamming Effect

2025-05-07T07:57:39+00:00

We consider a robot moving along a horizontally located rod. The de-sign includes two bushings, one of which, due to frictional forces as a result of jamming, wraps around the rod and is held on it, while the other one moves freely. The principle of operation of the propeller is based on the effect of periodic jamming. The movement of the mobile robot can be seen in the video on the link: https://drive.google.com/file/d/1QD5V7YwKOKPHlcZa2MQDy_Q4YiMs6KKv/view?usp=drive_link.

Aerial manipulation system for automated installation seismic activity sensors

2025-05-07T07:57:40+00:00

The development of unmanned aerial vehicles (UAVs), the improvement of their production technologies and control methods are constantly expanding the range of tasks in which the use of UAVs is relevant and economically feasible. One area in which the use of UAVs can provide significant benefits is the transportation of seismic activity sensors in hard-to-reach or dangerous areas. The work proposes an approach to the automated transportation, installation and collection of seismic activity sensors, which requires precise landing of the UAV. Positioning of the UAV during landing is carried out using technical vision and a fractal ArUco marker. The work describes the development of an original design for a gripper device and a transport fastening for the sensor. The principles of constructing the hardware of the gripper control system and its integration with the design and equipment of the UAV are considered, and the algorithms for the functioning of the gripper control module are described. Experiments were carried out that confirmed the effectiveness and performance of the proposed solutions. The collection of seismic activity sensors from the surface was successful in 80% of cases with deviations from the target landing point of up to 115 mm in the horizontal plane. The advantage of the developed gripper device is its relatively low weight - 1020 g, with the mass of the transported cargo being equal to 2760 g. The proposed kinematic scheme and its implementation using a screw-nut transmission eliminates the energy costs for fixing the transported cargo in flight. The low mass of the gripper and low requirements for the accuracy of positioning of the UAV at the landing point distinguishes the developed solutions from other existing ones and proves their promise in performing the tasks of manipulating and transporting ground objects using unmanned aerial systems.

Development of a computer vision system for recognizing grain seed samples

2025-05-07T07:57:42+00:00

This article presents the development of a computer vision system for recognizing suitable samples of grain crops for further formation of the seed fund. The goal of the work is to create a reliable and efficient algorithm that can accurately determine the presence and characteristics of damage in images of wheat, oats, and peas. Modern machine learning methods, such as convolutional neural networks, are introduced to train the recognition model. The development process included the collection and preparation of training data, selection and configuration of the neural network architecture, as well as testing and optimization of the algorithm. A comparison was made of the computer vision libraries YOLO, FASTER R-CNN, VISSL, OpenCV. The resulting system demonstrated high accuracy in recognizing defects and morphological characteristics of seeds in test images, with an accuracy of up to 87%. The developed system can be used in various applications related to the automation of agricultural processes, product quality analysis, photo sorters, phenotyping devices, as well as in seed quality control systems and in intelligent control systems for agricultural production processes in crop production.

Integration of Ant Colony Algorithm and Neural Networks for Task Management and Allocation in Groups of Mobile Robots

2025-05-07T07:57:43+00:00

This paper addresses the use of intelligent algorithms for task allocation and trajectory planning in groups of mobile robots. Structural schemes and methodologies for management and planning have been developed, where the ant colony algorithm is used for generating initial solutions for task distribution and route searching, while a neural network is applied for subsequent optimization and adjustment of robot trajectories in real-time. The overall structural scheme of the management and planning system includes elements such as a task distribution unit among robots, an initial solution generator, a trajectory optimizer, a robot control subsystem, and a monitoring and adjustment unit. Special attention is given to the implementation of the task distribution subsystem (initial solution generator) based on the ant colony algorithm and the movement trajectory optimization block based on a neural network. The initial solution generator is used to find the initial set of routes and task distribution strategies among a group of mobile robots using the ant colony algorithm, which simulates the foraging behavior of ants. The neural network analyzes the formed set of routes for efficiency and safety, adjusting them according to current conditions and new information about obstacles or changes in tasks. Simulation results are presented using a group of mobile robots working in a field with various obstacle configurations. The simulation results conclude the feasibility of using the developed approach in real-world task allocation scenarios in a group of robots.

About one way to scan a surface for a home walking robot

2025-05-07T07:57:45+00:00

The paper discusses the principles of building a home quadruped ro-bot’s vision system for stepping over obstacles and walking up stairs. The vi-sion system measures the distance to the support surface at the places where the robot’s legs may be positioned. The proposed system is based on the use of sev-eral TOF sensors to measure the distance to possible reference zones. A surface scanning model is presented to determine the reference areas. The issues of the influence of surface properties on the measurement results are considered. The influence of the instability of the sensor tilt angles and its width of the radiation pattern on the measurement accuracy is analyzed. It is shown that the use of sensors with a small angle of inclination makes it possible to measure the pa-rameters of horizontal surfaces with sufficient accuracy. Conclusions are drawn and recommendations are given to improve the accuracy of measurements. The test on the simulator showed a good match of the measured profile of the stage with the theoretical results. The proposed scanning method allows to reduce the cost of the control system, reduce the requirements for computing power and increase the battery life of the home robot.

Recursive Estimation Algorithms for AUV Collaborative Navigation in Case of Abnormal Outliers in Measurements

2025-05-07T07:57:46+00:00

To solve the problem of navigation for multiple autonomous unmanned underwater vehicles, an extended Kalman filter with an additional option of rejection of abnormal outliers in the measurement errors, and a correntropy extended Kalman filter have been developed. They have been compared to each other in terms of accuracy, consistency and computational complexity in the scenarios with Gaussian and non-Gaussian noise in the measurements, containing pulse-type interference.

Algorithms for Planning Trajectory of a Modular Wheeled In-pipe Robot

2025-05-07T07:57:46+00:00

The 2D and 3D mathematical and computer models of the kinematics of a modular wheeled robot when it moves inside the pipe for inspection of the inner surface are presented in this paper. In the 2D model the transition of the leading module from a linear trajectory to an arc with a constant radius and subsequent exit to a straight trajectory is considered. A numerical algorithm for the driven modules motion based on the position of the leading module and the constraint equation is constructed. As the analysis results of the simulation data, the dependence of the deviation between leading and driven modules trajectories versus the radius of the curved section is revealed. The optimal trajectory of the robot movement inside the curved pipe is synthesized in which all modules
remain within the safety zone, considering the deviations of the modules from the trajectory of the first one. An analytical 3D model is obtained which allows calculating the following movements: along the helical line trajectory (working mode of pipe inspection); dynamic change in the pitch of the helix; transition to movement along the circumference or along the pipe. The deviations of the tra-
jectory of the driven modules from the trajectory of the leading module are analyzed and structural methods of reducing the magnitude of these deviations are proposed. Computer models were coded with the C++ and Python programming languages. The visualization of the modular robot movement was carried out by the 3D modeling software Blender.

Parameter identification algorithm for a LTV system with partially unknown state matrix

2025-05-07T07:57:48+00:00

In this paper an adaptive state observer and parameter identification algorithm for a linear time-varying system is developed under condition that the state matrix of the system contains unknown time-varying parameters of a known form. The state vector is observed using only output and input measurements without identification of the unknown parameters. When the state vector estimate is obtained, the identification algorithm is applied to find unknown parameters of the system. Since the general mathematical model is considered, the proposed algorithm can be applied for different technical objects, such as mobile robots, manipulators, electro-mechanical systems, etc.

System for Placing Seismic Sensors Based on Actions of UAVs Group with Optimized Flight Plan

2025-05-07T07:57:49+00:00

A trajectory planning system for a group of unmanned aerial vehicles (UAVs) designed for seismic exploration tasks is proposed. The developed system is capable of efficiently allocating tasks and planning trajectories even considering large state spaces. The system has low computation cost, considers possible failures of UAVs in the group and provides online trajectory replanning in case of possible collisions. Series of experiments were conducted on real UAVs and in the Gazebo simulation environment, which demonstrated the computational efficiency of the system: the time for distributing tasks among UAVs was 0.007 s, and the maximum time for calculating one trajectory was 0.103 s. A comparison of the proposed solution with other methods for 7 UAVs showed that in terms of the total task completion time, the system outperformed the ECBS-TA method by 1% and the PC-TAFF method by 0.05%. In terms of individual mission execution time, the system surpassed the ECBS-TA method in 6 out of 7 UAVs and PC-TAFF in 2 out of 7 UAVs. The conducted tests demonstrated the effectiveness of the developed solution and the feasibility of its use.

Fish Image Classification Based on MobileNetV2 with Transfer Learning Technique for Robotic Application in Aquaculture

2025-05-07T07:57:51+00:00

Intelligent aquaculture helps solve the problems of traditional aquaculture by applying smart production methods. The combination of technology and the art of farming has opened the door to the smart seafood industry, where artificial intelligence, IoT and automation systems are effectively applied and are completely changing management. From water quality management to monitoring the health of farm animals, harvesting, and feeding are all done automatically and quickly with high efficiency. Among them, fish classification is an essential task in smart aquaculture. In the field of image classification DCNN has achieved remarkable successes. However, applying DCNN models also requires a large amount of data to train the model and a long training time. Therefore, applying a model based on MobileNetv2 has solved these problems. In this article, the fish image classification method based on MobileNetv2 with Transfer learning is applied. We used the MobileNetV2 network pre-trained by the ImageNet dataset as the base network and added the following layers to the base model and Softmax classifier, this new model is Adam optimized. This method finally achieved a classification accuracy of 98.18% in the test dataset of 330 images of 5 fish types.

A Technique of Distributed Missions Assignment Problem Solving in Heterogeneous Groups of Aerial Rescue Robots

2025-05-07T07:57:51+00:00

This research considers a problem of missions assignment in the heterogeneous group of aerial rescue robots, which function without obstacles on plain landscape. A new technique of distributed missions assignment problem solving is proposed, based on the metaheuristic optimization algorithms independent runs, which are launched on ARRs in a distributed manner. The novelty of the technique is that blocks of metaheuristics iterations with varying sizes are formed and distributed through the robotic group with the usage of efficient algorithms for the block size forming and assignment. This improves the makespan of distributed missions assignment problem solving significantly without considerable degrading of the overall missions assignment result.

Prediction of Deformations on Elastic Objects Using an LSTM Model

2025-05-07T07:57:52+00:00

A volumetric deformable object does not have a fixed representation. Its contour changes as forces are applied on it.

The number of distinctive traits like corners or points of high curvature are modified and the level of detail required for its description varies as its shape becomes simple or more intricate.

Furthermore, the type of deformation it will undergo varies drastically depending on its material and original shape. For these reasons mobile robots face a challenging problem when on the need of manipulating everyday objects like toys, cushions or dishwashing sponges.

In this paper we propose an LSTM neural network that learns to predict the deformation of the contour of an elastic object, taking as input the original contour and the sequence of position and force measurements of the robotic finger that interacts with it.

Even though the representation of the contour we use is still of fixed length, our model makes reasonable predictions of deformation for interactions it has not seen before for the entire interaction and not only for the next frame as other models in the literature do.

Also, since it uses only local information of pieces of the object, we envision that this work can be extended to cover the case of representations of varying length.

Robo-surveillance of a Solar Park for Interdisciplinary Education

2025-05-07T07:57:53+00:00

The field of intelligent mobile robotics is highly interdisciplinary by nature, spanning interest from mechanical, electric, and electronic engineering to computer science, physics, cognition and neuroscience. Yet, there are few opportunities during undergraduate education to experience work and communication between students from these disciplines. For this reason, we devised a scenario where knowledge from these diverse disciplines could be demonstrated and used for experiments by students during a one-week workshop. This scenario consists of trials with two different robots. The first phase of the trials makes use of a simple differential robot where the basics of electronics, communications over WiFi and Bluetooth networks are taught. The second phase makes use of an omnidirectional robot whose task is to survey a solar park to detect dusty solar panels. Its job is only to navigate through the park detecting the panels and sending one photo of the panel to a nearby workstation. This workstation processes the images captured by the small robots to detect whether the panels in them are clean or dusty, using a U-net neural network. Here we report our experience on the first two editions of this interdisciplinary workshop.

Camera based Navigation: AI vs Image Processing

2025-05-07T07:57:56+00:00

We compare the performance of rule-based image processing algorithms for autonomous navigation with that of a neural network approach based on an Nvidia CNN architecture. We detail our implementation of Hough-Transform-based and pixel-based navigation algorithms and discuss their advantages and disadvantages, including a quantitative comparison of their navigation performance in terms of speed, stability and reliability. We describe how rule-based algorithms can be used to train neural networks that result in enhanced navigation with high stability and reliability. We discuss the benefits of training neural networks with binary images rather than color images. We present our experiments on autonomous navigation with model cars racing on unknown tracks using the different algorithms and conclude with some recommendations for the use of each algorithm.

The robotic control system of UAV based on artificial intelligence technologies

2025-05-07T07:57:57+00:00

The paper presents the results of developing an algorithm for controlling and navigating an unmanned vehicle using the artificial intelligence method. The remote control in difficult conditions limits the capabilities of the unmanned aerial vehicle and reduces the efficiency of the entire system. In this case, the aircraft can perform tasks depending on the variability of external factors. An algorithm has been developed for using a visual or inertial navigation system to determine the coordinate and time characteristics of an unmanned aerial vehicle in the event of non-functioning of the global satellite navigation system. In this case, the developed algorithm for the functioning of the artificial intelligence of the on-board computer itself determines the coordinate-time characteristics of the unmanned aerial vehicle in autonomous flight mode. The proposed algorithm and control architecture are applied to an unmanned aerial vehicle intended to perform a reconnaissance flight. Then, in the paper there have been presented the architecture of control and navigation system based on artificial intelligence for winged UAV, the electrical diagram of the UAV control system, the electrical diagram of the magnetometer, the electrical diagram of the barometer and electrical diagram the driver in UAV.

Algorithms and Software for Simulation of Intelligent Systems Based on Multi-Agent Neurocognitive Architectures

2025-05-07T07:57:58+00:00

The work presents an approach to the design of intellectual decision-making systems, which consists in the synthesis of cognitive architectures and multi-agent systems. As a cognitive architecture, architecture is used, the nodes of which are multi-agent systems. This approach is based on the computing abstraction of the processes of multi -agent exchange of information between the neurons of the brain, in which individual neurons of the brain are considered as rational software agents that perform cooperative interaction with each other in order to maximize their local target functions. A simulation model of a decision-making system has been developed. The software implementation of the process of interaction between agents is presented within the framework of a multi-agent neurocognitive decision-making model. The algorithm and a program for modelling multi-agent neurocognitive architectures have been developed. Such architectures, through the collaboration of agents, can provide the search for suboptimal solutions under conditions of partial uncertainty. The developed software can be used to simulate the decision-making process for performing tasks in a real environment. In particular, the presented formalism of multi-agent neurocognitive artificial intelligence systems is experimentally used to ensure goal-directed behavior of autonomous mobile robots.