-
Decentralized Multi-Robot Formation Control Using Reinforcement Learning
Authors:
Juraj Obradovic,
Marko Krizmancic,
Stjepan Bogdan
Abstract:
This paper presents a decentralized leader-follower multi-robot formation control based on a reinforcement learning (RL) algorithm applied to a swarm of small educational Sphero robots. Since the basic Q-learning method is known to require large memory resources for Q-tables, this work implements the Double Deep Q-Network (DDQN) algorithm, which has achieved excellent results in many robotic probl…
▽ More
This paper presents a decentralized leader-follower multi-robot formation control based on a reinforcement learning (RL) algorithm applied to a swarm of small educational Sphero robots. Since the basic Q-learning method is known to require large memory resources for Q-tables, this work implements the Double Deep Q-Network (DDQN) algorithm, which has achieved excellent results in many robotic problems. To enhance the system behavior, we trained two different DDQN models, one for reaching the formation and the other for maintaining it. The models use a discrete set of robot motions (actions) to adapt the continuous nonlinear system to the discrete nature of RL. The presented approach has been tested in simulation and real experiments which show that the multi-robot system can achieve and maintain a stable formation without the need for complex mathematical models and nonlinear control laws.
△ Less
Submitted 26 June, 2023;
originally announced June 2023.
-
A Search Strategy and Vessel Detection in Maritime Environment Using Fixed-Wing UAVs
Authors:
Marijana Peti,
Ana Milas,
Natko Kraševac,
Marko Križmančić,
Ivan Lončar,
Nikola Mišković,
Stjepan Bogdan
Abstract:
In this paper, we address the problem of autonomous search and vessel detection in an unknown GNSS-denied maritime environment with fixed-wing UAVs. The main challenge in such environments with limited localization, communication range, and the total number of UAVs and sensors is to implement an appropriate search strategy so that a target vessel can be detected as soon as possible. Thus we presen…
▽ More
In this paper, we address the problem of autonomous search and vessel detection in an unknown GNSS-denied maritime environment with fixed-wing UAVs. The main challenge in such environments with limited localization, communication range, and the total number of UAVs and sensors is to implement an appropriate search strategy so that a target vessel can be detected as soon as possible. Thus we present informed and non-informed methods used to search the environment. The informed method relies on an obtained probabilistic map, while the non-informed method navigates the UAVs along predefined paths computed with respect to the environment. The vessel detection method is trained on synthetic data collected in the simulator with data annotation tools. Comparative experiments in simulation have shown that our combination of sensors, search methods and a vessel detection algorithm leads to a successful search for the target vessel in such challenging environments.
△ Less
Submitted 22 June, 2023;
originally announced June 2023.
-
Cooperative Aerial-Ground Multi-Robot System for Automated Construction Tasks
Authors:
Marko Krizmancic,
Barbara Arbanas,
Tamara Petrovic,
Frano Petric,
Stjepan Bogdan
Abstract:
In this paper, we study a cooperative aerial-ground robotic team and its application to the task of automated construction. We propose a solution for planning and coordinating the mission of constructing a wall with a predefined structure for a heterogeneous system consisting of one mobile robot and up to three unmanned aerial vehicles. The wall consists of bricks of various weights and sizes, som…
▽ More
In this paper, we study a cooperative aerial-ground robotic team and its application to the task of automated construction. We propose a solution for planning and coordinating the mission of constructing a wall with a predefined structure for a heterogeneous system consisting of one mobile robot and up to three unmanned aerial vehicles. The wall consists of bricks of various weights and sizes, some of which need to be transported using multiple robots simultaneously. To that end, we use hierarchical task representation to specify interrelationships between mission subtasks and employ effective scheduling and coordination mechanism, inspired by Generalized Partial Global Planning. We evaluate the performance of the method under different optimization criteria and validate the solution in the realistic Gazebo simulation environment.
△ Less
Submitted 29 April, 2022;
originally announced April 2022.
