论文部分内容阅读
为解决在速度层上无穷范数最小化模型中可能出现的不连续点问题,提出一种基于双判据方法的二次型优化模型.冗余机器人运动规划与控制模型可以统一各种关节物理极限,如关节变量极限与关节速度极限.同时该模型又可以最终转化为一个标准的二次规划问题.为了实时求解该二次规划问题,提出一种基于线性变分不等式(LVI)的原对偶神经网络.该神经网络作为实时求解器具有简单的分段线性结构和较高的计算效率.计算机对PUMA560机器手臂的模拟仿真表明,该方案具有灵活性和有效性.
In order to solve the problem of discontinuous points that may appear in the infinite norm minimization model of velocity layer, a quadratic optimization model based on double criterion method is proposed. The redundant robot motion planning and control model can unify all kinds of joint physics Such as joint variable limit and joint velocity limit, and the model can be finally transformed into a standard quadratic programming problem.In order to solve the quadratic programming problem in real time, an original duality based on linear variational inequality (LVI) Neural network as a real-time solver with a simple piecewise linear structure and a high computational efficiency.Computer simulation of the PUMA560 robot arm shows that the proposed scheme is flexible and effective.