An adaptive sampling-based strategy for mobile robot motion and path planning

Ph.D.Mobile robot motion and path planning is a basic research topic in robotics, which aims to plan a feasible trajectory or collision-free motions for the mobile robot for some specific tasks among a number of static or moving obstacles. Sampling-based algorithms are very popular due to their ability to fast and efficiently explore the state space. However, they suffer from several disadvantages such as slow convergence to the optimal solution, initial solution sensitivity and other problems. In order to find a high-quality initial solution and accelerate the convergence to the optimal solution, we propose an adaptive sampling-based strategy for mobile robot motion and path planning.First, we propose the Tropistic RRT*, an efficient planning algorithm via adaptively restricted sampling space, which is inspired by the tropism of plants. As most plant shoots exhibit positive phototropism and grow towards the light, we make the planner sample nodes towards the goal position as much as possible by adaptively restricting the sampling space. The experimental results demonstrate that the Tropositc RRT* performs better com- pared with the state-of-the-art algorithms. Then we introduce another novel algorithm, an extension of the Tropistic RRT*, on finding a high-quality initial solution for the RRTs algorithms in 2D environments because most existing RRTs algorithms are sensitive to the initial solution. We use the Gaussian Mixture Model (GMM) to represent the sampling function and the parameters of the GMM are automatically adjusted in the sampling process. This method is applicable to any RRTs algorithms and helps to improve the performance of the underlying algorithm. The experimental results reveal that our algorithm can achieve convincing performance in different 2D scenarios.Secondly, to accelerate the convergence speed of the sampling-based algorithm to the optimal solution, we propose the EB-RRT to achieve real-time motion planning for the mobile robot in the human robot coexisting environment. By combining the advantages of RRTs algorithms fast exploring the dynamic environment and Elastic Band method efficiently optimizing the initial solution, our proposed algorithm can quickly find the homotopy optimal trajectory compared with the state-of-the-art algorithms.Third, we apply the motion and path planning strategy to the robotic autonomous luggage trolley collection at airports. To solve this problem, we formulate it as a Multiple Traveling Salesman Problem (MTSP). Different from the conventional methods, we firstly divide the MTSP into a number of independent TSPs using Minimum Spanning Tree (MSP) algorithm. Then we calculate the feasible trajectory for each mobile robot to collect the luggage trolleys using a motion planning algorithm by taking into account the nonholonomic constraints. Finally, we implement the EB-RRT algorithm to achieve real-time motion planning in the collection process. The experimental results reveal that our algorithm performs better in terms of the time cost and the control cost compared with the conventional methods.移动机器人运动与路径规划是机器人学中的一个基础课题，它的目标是为移动机器人在静态或动态障碍物环境中规划出可行的路径和动作。基于采样的规划算法非常流行，因为它可以高效快速地搜索整个状态空间。然而，它也有一些缺点，包括收敛到最优解的速度慢、对初始解敏感和一些其他问题。为了找到一个高质量的初始解并且加快收敛到最优解的速度，我们提出了一种用于移动机器人运动与路径规划的自适应采样策略。首先，受到植物向光性的启发，我们提出了Tropistic RRT*，一种自适应限制采样空间的高效规划算法。就像大多数植物表现出向光性并且朝着光生长一样，我们的规划器通过自适应地调整采样空间，使得整个采样过程尽可能地趋向于目标点。实验结果表明相比于当前最好的规划算法，我们的算法性能更加优越。然后，作为Tropistic RRT* 算法的扩展，我们提出了另一种用于在二维环境下为RRT 算法生成高质量初始解的算法。我们使用高斯混合模型来表示采样函数，而且高斯混合模型的参数可以在采样过程中自动调整。这个方法可以应用到任何基于RRT 的规划算法中去帮助它们提高本身的性能。实验结果显示我们的算法可以在不同的二维环境中展现出令人信服的性能。第二，为了加快采样算法收敛到最优解的速度，我们提出了EB-RRT 实现在人机共存环境中的移动机器人实时运动规划。通过把RRT 算法快速搜索动态环境的优势和Elastic Band 方法高效优化初始解的优势结合在一起，我们提出的算法可以快速地找到同族最优解。第三，我们把运动与路径规划的策略应用到了机器人自动收集机场行李车中。为了解决机器人自动收集机场行李车的问题，我们首先把它规划成一个MTSP 问题。不同于传统的解决方法，我们首先利用MSP 算法把MTSP 分解成一个个独立的TSP。然后我们利用基于非完整性约束的运动规划算法为每一个机器人计算可行路径去收集相应的行李车。最后，我们使用EB-RRT 算法去实时优化机器人在收集行李车过程中的运动轨迹。实验结果表明相比于传统的解决方案，我们算法的时间成本和控制成本更少。Wang, Jiankun.Thesis Ph.D. Chinese University of Hong Kong 2019.Includes bibliographical references (leaves 109-117).Abstracts also in Chinese.Title from PDF title page (viewed on 7, May 2021)