基于CMA-ES算法的面向任务的机器人结构优化

Task-Oriented Robot Structure Optimization Based on CMA-ES Algorithm

为了解决指定任务的机器人结构设计问题,将任务要求和相关性能指标纳入到机器人结构设计过程中,提出了一种面向任务的机器人结构(DH参数)优化设计方法。首先,将指定任务描述为笛卡尔空间中一系列位姿点,姿态用四元素表示;其次,根据机器人运动学特性,推导出任务可达性约束条件,用各向同性指标、关节角度指标、尺寸指标构建目标函数,将机器人DH参数和到达各任务点时的关节值作为设计变量,从而建立了机器人结构优化设计的数学模型,并进一步利用罚函数法处理可达性约束;最后,根据优化问题具有非线性、不可导、高维度和连续值优化的特性,选用CMA-ES(协方差矩阵自适应进化策略)进行机器人结构优化设计问题的求解。实例仿真通过模拟一个位姿不断变化的复杂作业场景,验证了该方法的有效性。

In order to solve the problem of task-specific robot structure design,a task-oriented optimization method of robot structure(DH parameter)are proposed,in which task requirements and related performance indexes are incorporated into the process of robot structure design.Firstly,the assigned task is described as a series of pose points in Cartesian space,and the posture is represented by unit quaternion.Secondly,according to the kinematics characteristics of the robot,the constraints of task reachability are derived and the objective function are constructed by isotropic index,joint angle index and dimension index.Simultaneously,the mathematical model of optimal design of robot structure is established which the DH parameters of the robot and the joint values at each task point are taken as design variables,and the penalty function method is further used to deal with the reachability constraints;Finally,according to the optimization problem has characteristics of nonlinear,non-derivable,high dimension,CMA-ES(covariance matrix adaptive evolution strategy)is used to solve the robot structure optimization design problem.The case study is given to illustrate the effectiveness of the proposed method by simulating a design scenarios with changing pose.