| Author: | Yang, Yefeng |
| Title: | Deep reinforcement learning-based mobile robot path planning and control subject to model uncertainty and external disturbances |
| Advisors: | Wen, Chih-yung (AAE) Li, Boyang (AAE) Wang, Tianqi (AAE) |
| Degree: | Ph.D. |
| Year: | 2025 |
| Department: | Department of Aeronautical and Aviation Engineering |
| Pages: | xviii, 169 pages : color illustrations |
| Language: | English |
| Abstract: | Autonomous robots have attracted significant attention in recent years due to their wide-ranging applications in industry, environmental monitoring, and agriculture. However, many challenging issues must be addressed before robots can reliably execute specific tasks in complex environments. These challenges include mapping, self-localization, task allocation, planning, and control. Among these, trajectory planning and control are especially critical in design. This thesis introduces novel methodologies to address the challenges of trajectory planning and control. 1. To address the global and local planning challenges of autonomous robots, this thesis presents an enhanced Rapidly-exploring Random Tree (RRT) algorithm alongside a Deep Reinforcement Learning (DRL)-based obstacle avoidance method. Initially, a sampling-efficient RRT algorithm is proposed, which integrates the geometric properties of the environment to minimize the number of sampling nodes required in complex scenarios. Subsequently, in the local planning phase, the Network Decoupling (ND) technique is employed in the design of DRL to accelerate the training process and improve the efficacy of obstacle avoidance. 2. A DRL-based optimization framework is proposed to optimize the hyper-parameters in the traditional Recursive Fast Non-singular Terminal Sliding Mode Control(ler) (RFNTSMC). The Fast Non-singular Terminal Sliding Mode Control(ler) (FNTSMC) optimized by DRL achieves better performance in the quadrotor control problem. Thereafter, a fixed-time disturbance observer is utilized for compensating external disturbances and modeling uncertainty. The stability of the closed-loop learning-based control framework is guaranteed in a Lyapunov sense. 3. Building upon the second contribution, a distributed control framework is developed to stabilize a group of multi-quadrotors. First, a fully distributed FNTSMC and a distributed fixed-time disturbance observer are proposed to ensure the stability of the multi-quadrotor system. Subsequently, DRL techniques are employed to adaptively learn the near-optimal hyperparameters for the FNTSMCs. The stability of the framework is formally guaranteed using graph theory and Lyapunov stability analysis. 4. Building upon the second and third studies, and considering the impact of system convergence time characteristics on practical applications, we further propose an enhanced distributed predefined-time convergence control framework. This framework allows the upper bound of the convergence time for multi-agent systems to be explicitly specified. First, we derive a more generalized predefined-time stability criterion. Subsequently, we utilize this criterion to design a predefined-time stable observer. This observer simultaneously estimates external disturbances and their upper bounds, which enables the reduction of control gains, thereby mitigating chattering issues to a certain extent. Furthermore, we develop a distributed predefined-time controller to ensure that the entire quadrotor formation system achieves stability within the predefined time. Finally, extensive simulations and physical experiments are conducted to validate the effectiveness and superiority of the proposed control method. 5. Two simulation platforms are developed as auxiliary tools during the algorithm design process, and extensive physical experiments are conducted to validate the superiority and effectiveness of the proposed algorithms. |
| Rights: | All rights reserved |
| Access: | open access |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/13824