This paper addresses the challenge of decentralized task allocation within heterogeneous multi-agent systems operating under communication constraints. We introduce a novel framework that integrates graph neural networks (GNNs) with centralized training and execution (CTDE) paradigm, further enhanced by tailored Proximal Policy Optimization (PPO) algorithm for deep reinforcement learning (MARL). Our approach enables unmanned aerial vehicles (UAVs) ground (UGVs) to dynamically allocate tasks...

With the growing demand for efficient logistics and warehouse management, unmanned aerial vehicles (UAVs) are emerging as a valuable complement to automated guided (AGVs). UAVs enhance efficiency by navigating dense environments operating at varying altitudes. However, their limited flight time, battery life, payload capacity necessitate supporting ground station. To address these challenges, we propose HetSwarm, heterogeneous multi-robot system that combines UAV mobile robot collaborative...

Autonomous drone navigation in dynamic environments remains a critical challenge, especially when dealing with unpredictable scenarios including fast-moving objects rapidly changing goal positions. While traditional planners and classical optimisation methods have been extensively used to address this problem, they often face real-time, changes that ultimately leads sub-optimal performance terms of adaptiveness real-time decision making. In work, we propose novel motion planner, AgilePilot,...

The ability to perform locomotion in various terrains is critical for legged robots. However, the robot has have a better understanding of surface it walking on robust different terrains. Animals and humans are able recognize with help tactile sensation their feet. Although, foot robots not been much explored. This paper presents research novel quadruped DogTouch sensing feet (TSF). TSF allows recognition textures utilizing sensor convolutional neural network (CNN). experimental results show...

Autonomous drone landing on dynamic platforms presents formidable challenges due to unpredictable velocities and aerodynamic disturbances. This study introduces an advanced Deep Reinforcement Learning (DRL) agent, Lander.AI, designed navigate land 3D moving in the presence of disturbance sudden velocity changes, thereby enhancing autonomy safety. Lander.AI is trained gym-pybullet-drone simulation, environment that mirrors real-world complexities, including wind disturbance, ensure agent's...

Reinforcement learning (RL) methods have been actively applied in the field of robotics, allowing system itself to find a solution for task otherwise requiring complex decision-making algorithm. In this paper, we present novel RL-based Tic-tac-toe scenario, i.e. SwarmPlay, where each playing component is presented by an individual drone that has its own mobility and swarm intelligence win against human player. Thus, combination challenging strategy human-drone collaboration aims make games...

Nowadays, design and development of legged quadruped robots is a quite active area scientific research. In fact, the have become popular due to their capabilities adapt harsh terrains diverse environmental conditions in comparison other mobile robots. With higher demand for robot experiments, more researches engineers need an affordable quick way locomotion algorithm development. this paper, we present new open source HyperDog platform, which features 12 RC servo motors, onboard NVIDIA...

We propose a novel human-swarm interaction system, allowing the user to directly control swarm of drones in complex environment through trajectory drawing with hand gesture interface based on DNN-based recognition. The developed CV-based system allows behavior without additional devices human gestures and motions real-time, providing convenient tools change swarm's shape formation. two types were proposed implemented adjust hierarchy: free-form generation control. experimental results...

The continuous monitoring by drone swarms remains a challenging problem due to the lack of power supply and inability drones land on uneven surfaces. Heterogeneous can support longer inspections, however, their capabilities are limited mobility wheeled legged robots in cluttered environment.In this paper, we propose novel concept SwarmGear for autonomous inspection. It leverages heterogeneous swarm that investigates environment leader-follower formation. leader is able rough terrain traverse...

The automation of the car charging process is motivated by rapid development technologies for self-driving cars and increasing importance ecological transportation units. Automation this requires implementation Computer Vision (CV) techniques. However, it remains challenging to precisely position charger plug autonomously due sensitivity CV algorithms lighting weather conditions. We introduce a novel robotic operation system based on hand gesture recognition through teleconferencing...

Current methods for 3D reconstruction and environmental mapping frequently face challenges in achieving high precision, highlighting the need practical effective solutions. In response to this issue, our study introduces FlyNeRF, a system integrating Neural Radiance Fields (NeRF) with drone-based data acquisition high-quality reconstruction. Utilizing unmanned aerial vehicles (UAV) capturing images corresponding spatial coordinates, obtained is subsequently used initial NeRF-based of...

The paper proposes a novel concept of docking drones to make this process as safe and fast possible. idea behind the project is that robot with soft gripper grasps drone in midair. human operator navigates robotic arm ML-based gesture recognition interface. 3-finger hand fingers equipped touch sensors, making it possible achieve catching avoid inadvertent damage drone's propellers motors. Additionally, featured unique color-based force estimation technology based on computer vision (CV)...

The aspects of remote sensing data processing for natural and anthropogenic objects detection identification are examined. research results made by hyperspectral image Samara region given.

The teleoperation of robotic systems in medical applications requires stable and convenient visual feedback for the operator. most accessible approach to delivering information from remote area is using cameras transmit a video stream environment. However, such are sensitive camera resolution, limited viewpoints, cluttered environment bringing additional mental demands human paper proposes novel system based on an augmented virtual (VE). region-based convolutional neural network (R-CNN)...

Robots able to run, fly, and grasp have a high potential solve wide scope of tasks navigate in complex environments. Several mechatronic designs such robots with adaptive morphologies are emerging. However, the task landing on an uneven surface, traversing rough terrain, manipulating objects still presents challenges.This paper introduces design novel rotor UAV MorphoGear morphogenetic gear includes description robot’s mechanics, electronics, control architecture, as well walking behavior...

We propose a novel concept of augmented reality (AR) human-drone interaction driven by RL-based swarm behavior to achieve intuitive and immersive control formation unmanned aerial vehicles. The DroneARchery system developed us allows the user quickly deploy drones, generating flight paths simulating archery. haptic interface LinkGlide delivers tactile stimulus bowstring tension forearm increase precision aiming. released drones dynamically avoids collisions between each other, drone...

Teleoperation of robotic systems for precise and delicate object grasping requires high-fidelity haptic feedback to obtain comprehensive real-time information about the grasp. In such cases, most common approach is use kinesthetic feedback. However, a single contact point insufficient detect dynamically changing shape soft objects. This paper proposes novel telemanipulation system that provides cutaneous stimuli user's hand achieve accurate liquid dispensing by dexterously manipulating...

Today there is a high variety of haptic devices capable providing tactile feedback. Although most existing designs are aimed at realistic simulation the surface properties, their capabilities limited in attempts displaying shape and position virtual objects.

To achieve high fidelity haptic rendering of soft objects in a mobility virtual environment, we propose novel display DandelionTouch. The tactile actuators are delivered to the fingertips user by swarm drones. Users DandelionTouch capable experiencing feedback large space that is not limited device's working area. Importantly, they will experience muscle fatigue during long interactions with objects. Hand tracking and control algorithm allow guiding hand motions avoid collisions inside...

This paper proposes a novel concept of hybrid tactile display with multistimulus feedback, allowing the real-time experience position, shape, and texture virtual object. The key technology TeslaMirror is that we can deliver sensation object parameters (pressure, vibration, electrotactile feedback) without any wearable haptic devices. We developed full digital twin 6 DOF UR robot in reality (VR) environment, adaptive surface simulation control real-time. preliminary user study was conducted...