Магид Евгений Аркадьевич

Autonomous Perpendicular Parking of a Car-like Robot in an Unknown Environment

2024 · CHAPTER · en

LoRa technology application within the Internet of Flying Things concept for data collection from IoT devices

2024 · CHAPTER · en

The Internet of Flying Things is one of the priority areas for robotics and automation research and applications. A promising case is an application of UAVs in a search and rescue scenario for data collection from pre-positioned IoT devices, which allows assessing and monitoring a current situation in an operation area. This paper presents a concept of using a low-power LoRa protocol to transmit data from smart sensors to UAVs that deliver the data to a dedicated cloud for further processing and analysis. A mock-up of a structure, which is described in the paper, is to be realized using microcontrollers, LoRa modules and smart temperature sensors.

Implementation and Validation of the CautiousBug Algorithm in ROS Noetic

2024 · CHAPTER · en

In this paper, we present an implementation of the CautiousBug algorithm within the Noetic distribution of the Robot Operating System (ROS). Bug algorithms address a challenge of robot navigation in unknown environments without relying on pre-existing maps or constructing new ones. These algorithms utilize odometry data, operate without a map, require minimal computational resources, and can be implemented on relatively simple hardware. A C++ software application was created to simulate a behavior of the CautiousBug algorithm in various environments within the Gazebo simulator. This application allows for an analysis of key metrics, including accumulated yaw, distance traversed and algorithm's runtime. We conducted a set of virtual experiments in the Gazebo to evaluate the CautiousBug performance.

Implementation of Rev1 and Rev2 Bug Family Algorithms in ROS Noetic

2024 · CHAPTER · en

Modern map-dependent algorithms for mobile robot navigation typically overload a CPU and memory with a gradually increasing amount of environmental data. In contrast, Bug family local path planning algorithms operate without mapping and have significantly lower hardware requirements. Bug algorithms use real-time measurements from visual and touch sensors to make immediate decisions on direction of the robot’s motion. This paper presents an implementation of Rev1 and Rev2 Bug algorithms for the Robot Operating System (ROS) Noetic framework with a virtual model of the TurtleBot3 Burger robot. Given algorithms are validated in Gazebo simulation using various convex and maze environments. The results demonstrate superiority of Rev1 method over Rev2 in trajectory length in convex maps. However, Rev2 finds shorter paths in maze type environments.

Modeling Sunlight in Gazebo for Vision-Based Applications Under Varying Light Conditions

2024 · CHAPTER · en

Vision is one of the well-researched sensing abilities of robots. However, applying vision-based algorithms can be challenging when used in different environmental conditions. One such challenge in vision-based localization is dynamic lighting conditions. In this paper, we present a new Gazebo plugin that enables realistic illumination changes depending on a current Sun's position. A plugin's underlying algorithm takes into account various parameters, such as date, time, latitude, longitude, elevation, pressure, temperature, and atmospheric refraction. Virtual experiments demonstrated effectiveness of the proposed plugin, and the source code is available for free academic use.

A Tool for Mass Generation of Random Step Environment Models with User-Defined Landscape Features

2024 · CHAPTER · en

Computer simulations are growing in popularity in robotics research due to their near-zero cost of error and lower labor intensity. One of necessary components of a simulation, in addition to a robot model, is a model of a world in which the robot operates. While it is always possible to construct a world model manually, a demand for automatic tools that generate multiple testing environments with particular user-defined features grows together with integration of data hungry machine learning techniques into robotic algorithms. This article presents a next generation of LIRS-RSEGen tool for constructing virtual random step environments (RSE). The new tool can simultaneously generate multiple RSE models with user-defined specific features that are declared via an intuitive graphical user interface. The resulting models simulate an urban search and rescue environment and can be used with robot models for developing and testing software for localization, mapping, navigation and locomoti on, and are applicable for machine learning due to their relatively low impact on performance and random elements in RSE generation. The constructed worlds’ performance was successfully tested with robot models in the Webots and Gazebo simulators.

CautiousBug path planning algorithm package for ROS Noetic

2024 · CHAPTER · en

The problem of robot navigation in unknown environments can be addressed without using a ready map or constructing a new one. The Bug family algorithms focus on sensor odometry data. The CautiousBug algorithm stands out from the rest of the algorithms by a considerable property: as its name implies, it does not make risky decisions that can lead to serious deterioration in performance indicators. This difference allows it to compete with other algorithms, whose errors in some cases lead to a significant increase in a trajectory length. This paper presents a practical implementation of the highly theoretical CautiousBug algorithm within the Noetic distribution of the Robotic Operating System (ROS).

Torque control of a wheeled humanoid robot with dual redundant arms

2024 · ARTICLE · en

Most of conventional controllers are prone to consume more computational time for controlling a wheeled humanoid robot. A nonlinear trajectory control of a wheeled humanoid robot using a model-based controller with less computational load and energy consumption is presented in this article. The upper body of the humanoid robot consists of two 6 degrees of freedom redundant arms, a three degrees of freedom torso and a two degrees of freedom neck. The nonholonomic mobile platform consists of two actuated wheels and two caster wheels. Nonlinearity of the robot dynamic model and coupling between various branches of the upper body are taken into account. The dynamic model derived using Newton–Euler approach along with decoupled Natural Orthogonal Complement matrix approach is used to derive dynamic equations of the upper body and the wheeled platform. Zero moment point based stability approach is used for verifying stable motion of the wheeled humanoid robot with minimum energy and time consumption to complete a task. The proposed computed torque controller is compared with other similar controllers developed for the same robot model to prove advantages of the proposed torque controller. Simulation results are experimentally validated with the real robot.

Electromyography-Based Biomechanical Cybernetic Control of a Robotic Fish Avatar

2024 · ARTICLE · en

This study introduces a cybernetic control and architectural framework for a robotic fish avatar operated by a human. The behavior of the robot fish is influenced by the electromyographic (EMG) signals of the human operator, triggered by stimuli from the surrounding objects and scenery. A deep artificial neural network (ANN) with perceptrons classifies the EMG signals, discerning the type of muscular stimuli generated. The research unveils a fuzzy-based oscillation pattern generator (OPG) designed to emulate functions akin to a neural central pattern generator, producing coordinated fish undulations. The OPG generates swimming behavior as an oscillation function, decoupled into coordinated step signals, right and left, for a dual electromagnetic oscillator in the fish propulsion system. Furthermore, the research presents an underactuated biorobotic mechanism of the subcarangiform type comprising a two-solenoid electromagnetic oscillator, an antagonistic musculoskeletal elastic system of tendons, and a multi-link caudal spine composed of helical springs. The biomechanics dynamic model and control for swimming, as well as the ballasting system for submersion and buoyancy, are deduced. This study highlights the utilization of EMG measurements encompassing sampling time and 𝜇-volt signals for both hands and all fingers. The subsequent feature extraction resulted in three types of statistical patterns, namely, Ω,𝛾,𝜆 serving as inputs for a multilayer feedforward neural network of perceptrons. The experimental findings quantified controlled movements, specifically caudal fin undulations during forward, right, and left turns, with a particular emphasis on the dynamics of caudal fin undulations of a robot prototype.

Comparative analysis of neural network models performance on low-power devices for a real-time object detection task

2024 · ARTICLE · en

A computer vision based real-time object detection on low-power devices is economically attractive, yet a technically challenging task. The paper presents results of benchmarks on popular deep neural network models, which are often used for this task. The results of experiments provide insights into trade-offs between accuracy, speed, and computational efficiency of MobileNetV2 SSD, CenterNet MobileNetV2 FPN, EfficientDet, YoloV5, YoloV7, YoloV7 Tiny and YoloV8 neural network models on Raspberry Pi 4B, Raspberry Pi 3B and NVIDIA Jetson Nano with TensorFlow Lite. We fine-tuned the models on our custom dataset prior to benchmarking and used post-training quantization (PTQ) and quantization-aware training (QAT) to optimize the models’ size and speed. The experiments demonstrated that an appropriate algorithm selection depends on task requirements. We recommend EfficientDet Lite 512×512 quantized or YoloV7 Tiny for tasks that require around 2 FPS, EfficientDet Lite 320×320 quantized or SSD Mobilenet V2 320×320 for tasks with over 10 FPS, and EfficientDet Lite 320×320 or YoloV5 320×320 with QAT for tasks with intermediate FPS requirements.