Loading...

Safe Path Planning for Cooperative Mobile Robots Based on Deep Reinforcement Learning

Kazemi Tameh, Ehsan | 2023

81 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 56279 (08)
  4. University: Sharif University of Technology
  5. Department: Mechanical Engineering
  6. Advisor(s): Khodaygan, Saeed
  7. Abstract:
  8. Nowadays, with the remarkable development of the robotics industry, there is an increasing demand for mobile robots. Mobile robots can be deployed individually or in groups for various tasks such as autonomous warehouses, search and rescue operations, firefighting operations, and maintenance and repairs. It is evident that performing certain tasks, such as moving large and long objects or firefighting operations, is more efficient when robots are deployed cooperatively, and in some cases, these tasks cannot be accomplished by a single robot alone. Therefore, in recent years, the issue of path planning for cooperative robots has received significant attention. By cooperation, we mean that robots work as a team to achieve a goal, with a focus on leader-follower robots. Although previous works have addressed the path planning problem for leader-follower robots using classical control methods or a combination of control and deep reinforcement learning methods, there was a gap for a method that could solve the path planning solely based on deep reinforcement learning methods. This is because control-based or hybrid control and deep reinforcement learning methods lack the ability to generalize and adapt to different environments and unforeseen scenarios. Hence, they are only applicable in known environments and face difficulties in unknown environments. To address this issue, this thesis proposes a method to perform safe path planning for both leader and follower robots solely based on deep reinforcement learning algorithms. To achieve this, the state space, action space, and reward function for each robot were designed according to the problem's requirements. In the next stage, an improved algorithm called I-MATD3 was developed based on deep reinforcement learning methods to guide the robots in safe routing towards the target point. The objective of safe path planning is to minimize the probability of collision with obstacles in the environment. Then, a curriculum learning approach was utilized to train the robots in complex environment routing. This means that two environments, a simple and a complex one, were designed in the Gazebo simulator, and the robots first learned routing in the simple environment and then were transferred to the complex environment to enhance their capabilities in performing tasks. Finally, the results of the I-MATD3 algorithm were compared with the MATD3 algorithm, and the performance of the proposed algorithm was evaluated
  9. Keywords:
  10. Deep Reinforcement Learning ; MATD3 Algorithm ; Cooperative Mobile Robotics ; Robot Pathplanning ; Leader-Follower Robots ; Path Planning

 Digital Object List

 Bookmark

...see more