机器人无迹微分动态规划算法研究  

Research on Unscented Kalman Dynamic Programming Algorithm of Robot

作　　者：刘伟 马利强 马彪 陈雪辉 黄磊 LIU Wei;MA Liqiang;MA Biao;CHEN Xuehui;HUANG Lei(School of Mechanical and Electrical Engineering,Anhui Jianzhu University,Hefei 230601,China)

机构地区：[1]安徽建筑大学机械与电气工程学院,安徽合肥230601

出　　处：《安徽建筑大学学报》2021年第4期71-76,共6页Journal of Anhui Jianzhu University

基　　金：安徽建筑大学引进人才及博士启动基金项目(2018QD16);安徽省高校协同创新项目(GXXT-2019-036)。

摘　　要：针对机器人非线性系统轨迹规划中微分动态规划算法由于动力学导数计算导致的实时性差与梯度下降慢问题,采用微分动态规划算法与无迹卡尔曼思想相结合的方式,以采样与差分方式代替动力学导数计算,建立无迹微分动态规划算法。将无迹微分动态规划与微分动态规划在非线性的倒立摆模型上进行模拟仿真对比。实验结果表明,系统参数相同时,无迹微分动态规划算法在保证良好的二阶收敛性和相同控制效果的前提下,既能减少迭代次数,又对成本压缩更敏感,且梯度下降更快,同时缩短算法整体的运行时间。The differential dynamic programming algorithm in the trajectory planning of the robot's nonlinear system due to the calculation of the dynamic derivative leads to poor real-time performance and slow gradient descent.Trackless differential dynamic programming algorithm adds Unscented Kalman thought to differential dynamic programming,using sampling and difference instead of dynamic derivative calculation.The traceless differential dynamic programming and differential dynamic programming are compared on the nonlinear inverted pendulum model.The experimental results show that when the system parameters are the same,the traceless differential dynamic programming algorithm can ensure good second-order convergence and control effect.The traceless differential dynamic programming algorithm can not only reduce the number of iterations and the overall running time of the algorithm,but also has a better sensitivity of cost compression and a faster gradient descent.

关 键 词：轨迹规划 无迹微分动态规划 仿真对比 二阶收敛 时间缩短 

分 类 号：TP242.6[自动化与计算机技术—检测技术与自动化装置]