仿人机器人运动姿态的动态多目标优化研究

RESEARCH ON DYNAMIC MULTI-OBJECTIVE OPTIMIZATION OF MOTION POSTURE OF HUMANOID ROBOT

胡运强

11749153

硕士

机械工程

柯文德

2019-05-30

2019-06-28

哈尔滨工业大学

深圳

仿人机器人由于具备与人类似的形态和结构，可以协助或取代人类从事各种工作。近年来仿人机器人多应用于玩具、表演、教育等服务业领域，在实际生产作业中却难以得到大规模应用，其中一个原因是相比其它机器人，仿人机器人的结构更加复杂，性能难以得到高效合理的应用，导致工作效率低下和成本高昂。仿人机器人行走作业的运动姿态可分为步态规划和路径规划两部分，步态规划决定行走作业的速度、稳定性和电机成本等性能，路径规划则决定行走路径长度和路径安全性等性能。这些性能间互相制约，无法将全部性能最优化，是一个多目标优化问题。为了在平衡各项性能的基础上实现总体性能最优，本文提出一种改进动态多目标优化算法，对仿人机器人行走过程中的各项性能进行优化。研究内容分为以下三个部分：（1）提出了一种改进动态多目标优化算法，并通过测试实验验证算法的适应性、实时性和可靠性。（2）将仿人机器人步态前进的速度、行走的稳定性与电机成本性能归纳为目标函数；将目标函数代入改进动态多目标优化算法中，实现仿人机器人的步态规划优化，并进行实验验证。（3）基于人工势场法进行机器人多目标路径规划，解决了传统人工势场法在复杂障碍物情况下易陷入零势场点、且无法保证路径最优的问题；通过实验实现最短路径与最安全路径间的优化平衡，完成仿人机器人的路径规划。

Humanoid robots can assist or replace humans in various jobs because of their similar shape and structure. In recent years, humanoid robots are mostly used in toys, performances, education and other service industries, but it is difficult to get large-scale application in actual production operations. One of the reasons is that compared with other robots, the structure of humanoid robots is more complex, its performance is difficult to get efficient and reasonable application, resulting in low efficiency and high cost. The motion posture of humanoid robot walking can be divided into gait planning and path planning. Gait planning determines the speed, stability and motor cost of the walking operation, and path planning determines the length and safety of the walking path. These performances restrict each other and cannot optimize all performances. This is a multi-objective optimization problem. In order to achieve performance optimization on the basis of balancing various performances, an improved dynamic multi-objective optimization algorithm is proposed to optimize the performance of humanoid robots during walking operation. The research contents are divided into three parts: (1) An improved dynamic multi-objective optimization algorithm is proposed, and its adaptability, real-time performance and reliability are verified by test experiments. (2) The walking speed, walking stability and motor cost performance of the humanoid robot are summarized as objective functions, and the objective function is substituted into the improved dynamic multi-objective optimization algorithm to realize the optimization of gait planning of the humanoid robot , which is verified by experiments. (3) Multi-objective path planning of robot based on artificial potential field (APF) method can solve the problem that traditional APF method is easy to fall into zero potential field under complex obstacles and can not guarantee the optimal path. The optimal balance between the shortest path and the safest path is achieved through experiments, and the path planning of humanoid robot is completed.

仿人机器人 动态多目标优化 步态规划 路径规划

humanoid robot dynamic multi-objective optimization gait planning path planning

中文

联合培养

南方科技大学

胡运强. 仿人机器人运动姿态的动态多目标优化研究[D]. 深圳. 哈尔滨工业大学,2019.