基于随机重启的机器人高斯过程运动规划  

Gaussian process motion planning for robots based on random restarts

作　　者：袁绪清 魏媛媛 王耀力[1] 常青[1] 付世沫 YUAN Xuqing;WEI Yuanyuan;WANG Yaoli;CHANG Qing;FU Shimo(College of Optoelectronics,Taiyuan University of Technology,Jinzhong 030600,China;Taiyuan Water Supply Design and Research Institute Co.,Ltd.,Taiyuan 030024,China)

机构地区：[1]太原理工大学电子信息与光学工程学院,晋中030600 [2]太原供水设计研究院有限公司,太原030024

出　　处：《现代制造工程》2024年第12期54-60,86,共8页Modern Manufacturing Engineering

基　　金：山西省重点研发项目(201903D321003);太原供水设计研究院有限公司项目(RH2000005391);山西省自然科学基金项目(201801D121141)。

摘　　要：针对高斯路径运动规划(GPMP2)算法应用于移动机器人时,在复杂障碍物环境中易陷入局部最优和避障性能不佳的问题,提出一种基于随机重启和避障改进的(GPMP2-SROAI)方法。首先,引入协变哈密尔顿优化(CHOMP)算法中的随机重启机制对轨迹施加扰动,使其跳出局部最优,提高轨迹优化的效率和鲁棒性;随后,引入基于障碍函数的模型预测控制(MPC-CBF)方法,在优化过程中通过预测机器人的运动范围以避免碰撞。仿真结果表明,改进后的规划成功率达到了92.6%,较GPMP2提高了24.9%,路径最短概率提高了12.5%,平均平滑度提高了4.8%,与主流算法进行对比也取得了更好的轨迹规划质量,轨迹更加平滑且避障效果更佳。In complex obstacle environments,mobile robots using the Gaussian Path Motion Planning(GPMP2) algorithm suffer from the problems of falling into local optimums and poor obstacle avoidance performance,a method based on stochastic restart and obstacle avoidance improvement was proposed Gaussian Process Motion Planning with Stochastic Restart and Obstacle Avoidance Improvement(GPMP2-SROAI).Firstly,the random restart mechanism in the Covariant Hamiltonian Optimisation for Motion Planning(CHOMP) was introduced to apply perturbations to the trajectory to jump out of the local optimum and improve the efficiency and robustness of the trajectory optimization.Then,a Model Predictive Control with Control Barrier Function(MPC-CBF) approach based on the barrier function was introduced to avoid collisions by predicting the range of motion of the robot during the optimisation process.Simulation results show that,the improved algorithm achieves a success rate of 92.6 % in path planning,which is 24.9 % higher than that of GPMP2,12.5 % higher than the path shortest probability,and 4.8 % higher than the average smoothing degree,and also achieves a better quality of trajectory planning compared with the mainstream algorithms,with smoother trajectories and better obstacle avoidance effects.

关 键 词：路径规划 随机重启 机器人避障 因子图优化 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]