Neural Network-Based Adaptive Motion Control for a Mobile Robot with Unknown Longitudinal Slipping

Abstract

When the mobile robot performs certain motion tasks in complex environment, wheel slipping inevitably occurs due to the wet or icy road and other reasons, thus directly influences the motion control accuracy. To address unknown wheel longitudinal slipping problem for mobile robot, a RBF neural network approach based on whole model approximation is presented. The real-time data acquisition of inertial measure unit (IMU), encoders and other sensors is employed to get the mobile robot's position and orientation in the movement, which is applied to compensate the unknown bounds of the longitudinal slipping using the adaptive technique. Both the simulation and experimental results prove that the control scheme possesses good practical performance and realize the motion control with unknown longitudinal slipping. which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.

Smoothed-shortcut Q-Learning Algorithm for Optimal Robot Agent Path Planning

“…Dijkstra 算法采用遍历搜索模式 [2] ， 可以用于求 解一个顶点到其余各个顶点的最短路径问题，但该 算法的计算节点多，计算量大，导致寻路速度慢， 不适用于复杂环境 [3] 。HART 等在 1968 年提出的 A*算法是在 Dijkstra 算法的基础上加入了启发函 数 [4] ，根据代价值来评价扩展节点的优劣，进而找 出最佳的扩展节点，直至扩展到目标点。虽然 A* 算法减少了扩展节点的数量，提高了运算效率，但 是得到的路线存在转折点多，安全距离小等问题 [5] 。 段书用等在 A*算法中引入安全距离矩阵和优化启 发函数，减少了路径转折点，提高了路径安全性 [6] 。 JPS 算法引入定向搜索模式 [7] ，进一步优化了 A*算 法寻找后继节点的操作，减少了转折点和扩展节点 数量，提高了路径的平滑性和算法寻路的运算速 度 [8] 。 由 KHATIB 在 1986 年提出的人工势场法根据 目标点和障碍物所在位置，在地图中构建出引力场 和斥力场，通过势函数的下降方向规划出一条安全 路径 [9] 。然而，在复杂环境中所构建的势能场易出 现多个零势能点，导致路径无法到达目标点 [10] 。快 速随机树搜索算法是基于随机采样的路径规划算 法，根据搜索树上已有的节点不断拓展出新节点， 进而找到目标位置。该算法搜索能力强，搜索速度 快，但同样存在搜索精度低、路径平滑性差等问 题 [11][12] 。DORIGO 根据蚂蚁协同觅食的现象，提出 了蚁群算法，该算法鲁棒性好且易于与其他算法融 合，但存在收敛速度慢等问题 [13][14] 。神经网络算法 是通过对大量数据样本进行训练， 进而求出最优解。 但该算法需要花大量时间用于训练样本且对网络模 型要求较高 [15][16] 。上述算法适用于已知全局环境信 息的情况 [17] ，当静态障碍物信息未知时，则不适用。 由 WATKINS 提出的 Q-Learning 算法，不依赖环境 模型，不需要环境的先验知识，利用自身的奖惩机 制，通过试错便可找到一条从起点到终点的安全路 径 [18] ，而且 Q-Learning 算法鲁棒性好。因此，呈现 了很多基于 Q-Learning 算法的改进研究，如：LOW 等基于花朵授粉原理，改进了 Q-Learning 算法的初 始化过程，提高了算法的收敛速度 [19] ；MARTHI 将 奖惩函数进行分解，提高了 Agent 的学习效率 [20] ； BIANCHI 等 提 出 了 Heuristically Accelerated Q-Learning 算法通过引入可在线更新的启发函数来 指导 Agent 选择动作 [21] 。这些改进的 Q-Learning 算 法虽然提高了 Agent 的学习效率，但直接应用于路 径规划后，得到的路径存在转折点多，路径不平滑 等问题 [22][23] 。 综上所述，本文基于 Q-Learning 算法和改进的 Q-Learning 算法 [24][25][26] ，提出一种 Smoothed-Shortcut…”

Optimal trajectory tracking of tractor-trailer wheeled platforms taken into account wheel dynamics

“…According to their novel features, various practical industries like medical professions use these platforms [1][2][3]. For this reason, mobile robots, especially wheeled ones, have received vast attractions and treated by many robotic researchers [4][5][6]. In addition, in order to enhance the output and their effectiveness, researchers have enthusiastically endeavored to generate optimal paths for mobile robotic platforms.…”

Longitudinal Stability of Wheeled Mobile Robots - Degree of Stability

“…Usually, in similar studies, differential equations are used [1], [2], [4], [5] to represent the mathematical model. Instead, the D'Alembert's principle is used here, which considers an equilibrium system of forces, including inertial forces [6]; the accuracy of the calculations does not deteriorate.…”