基于粒子滤波的自动装配定位方法  

Automatic assembly location method based on particle filter

作　　者：高巍[1] 邵晓东[1] 刘焕玲[1] 

机构地区：[1]西安电子科技大学电子装备结构设计教育部重点实验室,陕西西安710071

出　　处：《计算机集成制造系统》2014年第7期1615-1624,共10页Computer Integrated Manufacturing Systems

基　　金：陕西省自然科学基础研究计划资助项目(2014JZ016)~~

摘　　要：为了在虚拟环境中对零件模型进行快速自动定位并仿真实际装配过程,将传统的粒子滤波算法与典型的马尔科夫蒙特卡洛方法—MH抽样算法相结合,并应用于虚拟装配。利用随机样本描述零件位姿的概率分布,根据重要性函数对零件进行位姿采样。通过调节各采样粒子权值的大小对发生干涉的零件进行位姿重采样,模拟实际装配中零件位姿的概率分布,并以样本的加权计算结果对零件位姿进行估计。分析了采样粒子数、零件外形复杂程度等因素对方法性能和装配效率的影响。该方法已经用于自主开发的基于自然交互的虚拟设计平台,实例表明它可以自动精确地完成装配引导。To realize the fast automatic assembly positioning of virtual part model and realistic simulation of assembly process, the traditional particle filter algorithm was combined with typical Markov Chain Monte Carlo （MCMC） method-Metropolis Hastings （MH） sampling method to applied to virtual assembly. Random samples were used to describe the probability distribution of parts position, and the parts were sampled according to the importance func- tion. Through adjusting the weights of particles and performing poses resampling for the collided parts, the proba- bility distributions of actual part pose were simulated and the assembling part pose was estimated by the weighted calculation results of the samples. The influences of the factors such as sampling number and parts shapes complexity on the performance of this method and assembly efficiency were discussed. This algorithm had been applied to a self-developed virtual assembly prototype system, and the application result showed that the proposed algorithm could complete assembly navigation automatically and precisely.

关 键 词：虚拟装配 粒子滤波 自动装配定位 马尔科夫链蒙特卡洛方法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]