Магид Евгений Аркадьевич

Visual data processing framework for a skin-based human detection

2021 · CHAPTER · en

For a large variety of tasks autonomous robots require a robust visual data processing system. This paper presents a new human detection framework that combines rotation-invariant histogram of oriented gradients (RIHOG) features and binarized normed gradients (BING) pre-processing and skin segmentation. For experimental evaluation a new Human body dataset of over 60000 images was constructed using the Human-Parts dataset, the Simulated disaster victim dataset, and the Servosila Engineer robot dataset. Random, Liner SVM, Quadratic SVM, AdaBoost, and Random Forest approaches were compared using the Human body dataset. Experimental evaluation demonstrated an average precision of 90.4% for the Quadratic SVM model and showed the efficiency of RIHOG features as a descriptor for human detection tasks.

Structural Model of a Training Computer Program for Improving Professional Skills of a Student in a Role of a District Polyclinic Physician

2021 · ARTICLE · en

An important feature of a medical education in the context of the COVID-19 pandemic is a migration of classes into an online or mixed format, which requires simulated-based teaching methods with elements of robotics and artificial intelligence. We analyzed computer programs for maintaining medical records of patients that are employed by various polyclinics of Kazan city. Based on the analysis, we summarized requirements for a training computer program that could develop medical records maintaining skills of university students. The proposed program focuses on modeling situations associated with pre-hospital stage processing of medical records for patients. The paper presents an improved architecture of the program and an example of its use.

Function Approximation Technique Based Immersion and Invariance Control for Unknown Nonlinear Systems

2020 · ARTICLE · en

A function approximation technique based immersion and invariance (FATII) control method is proposed in this letter. Firstly, an unknown control system is restructured as the combination of an auxiliary system and a variation term from the original system. The variation term is treated as a time-varying uncertainty and parameterized by a group of weighted chosen basis functions. These weights are estimated at every time instant and the change of the estimates is governed by an update law. The update law is defined based on an immersion and invariance approach such that both the system state and the estimation error converge to zero. The FATII method is model-free and thus applicable to a wide range of systems. The asymptotic stability of the proposed method is established and its feasibility is verified under simulations.

Automatic Tool for Gazebo World Construction: From a Grayscale Image to a 3D Solid Model

2020 · CHAPTER · en

Artificial intelligence based framework for robotic search and rescue operations conducted jointly by international teams

2020 · CHAPTER · en

Many countries suffer from various natural disasters, including heavy rains, that are associated with further flood and landslide disasters. Based on our experiences of different disasters response, we develop a joint international operation framework for a disaster site management with distributed heterogeneous robotic teams that consist of unmanned aerial, ground, surface, and underwater vehicles. The artificial intelligence-based information collection system, which is targeting to become a worldwide standard, contains interaction protocols, thematic mapping approaches, and map fusion processes. The project provides a new working framework and control strategies for heterogeneous robotic teams’ cooperative behavior in sensing, monitoring, and mapping of flood and landslide disaster areas. In this paper, we present an overview of the system and a first stage toward robot interaction protocols development and the system modeling within robot operating system’s Gazebo environment.

Laser Rangefinder and Monocular Camera Data Fusion for Human-Following Algorithm by PMB-2 Mobile Robot in Simulated Gazebo Environment

2020 · CHAPTER · en

Pilot Virtual Experiments on ArUco and ArTag Systems Comparison for Fiducial Marker Rotation Resistance

2020 · CHAPTER · en

Comparative Analysis of Monocular SLAM Algorithms Using TUM & EuRoC Benchmarks

2020 · CHAPTER · en

Stable and robust path planning and movement in ground mobile robots require a combination of accuracy and low latency in their state estimation. However, state estimation algorithms must provide these qualities under the computational and power constraints of embedded hardware. Simultaneous localization and mapping (SLAM) algorithms are the best choices for state estimation in these scenarios, in addition to their ability to operate without external localization from motion capture or global positioning systems. Moreover, a single-camera setup is the most common solution for robotic platforms, which reduces our domain of interest to the specific SLAM algorithms type—monocular SLAM. Yet, it is still not clear from the existing literature, which monocular SLAM algorithms perform well under the accuracy, latency, and computational constraints of a ground mobile robot with onboard state estimation. This paper evaluates an array of the most recent publicly available monocular SLAM methods: ORB-SLAM2, DSO, and LDSO. The evaluation considers the pose estimation accuracy (alignment error, absolute trajectory error, and relative pose error) while processing the TUM Mono and EuRoC datasets on the specific hardware platform with a balanced amount of computational resources and power consumption. We present our complete results as a benchmark for the research community.

Person-Following Algorithm Based on Laser Range Finder and Monocular Camera Data Fusion for a Wheeled Autonomous Mobile Robot

2020 · CHAPTER · en

Reliable human following is one of the key capabilities of service and personal assisting robots. This paper presents a novel person tracking and following approach for autonomous mobile robots that are equipped with a 2D laser rangefinder (LRF) and a monocular camera. The proposed method does not impose restrictions on a person’s clothes, does not require a head or an upper body to be within a camera field of view and is suitable for low height indoor robots as well. The algorithm is based on a metric that takes into an account parameters obtained directly from LRF and monocular camera data. The algorithm was implemented and tested in the Gazebo simulator. Next, it was integrated into a control system of the TIAGo Base mobile robot and successfully validated in university environment experiments with real people. In addition, this paper proposes a new criterion of algorithm performance estimation, which is a function of false positives number and traveled distances by a person and by a robot. Further this criterion is used to compare performance of the proposed method with the Multiple Instance Learning (MIL) tracker in simulated and in real world environments.

Comparison of ROS-based monocular visual SLAM methods: DSO, LDSO, ORB-SLAM2 & DynaSLAM

2020 · CHAPTER · en