移动机器人的概率定位方法研究进展

Progress of Probabilistic Localization Methods in Mobile Robots

综述了近几年来流行的移动机器人基于概率定位的各种方法,对它们的性能进行了分析比较,所有这些方法都应用贝叶斯规则作为理论基础.首先,介绍了位置跟踪广泛应用的卡尔曼滤波方法和在全局定位方面取得一定成功的马尔可夫定位方法.然后,介绍了计算效率更高的粒子滤波定位方法,即蒙特卡洛法,以及最近自适应采样的粒子滤波方法,它比简单的粒子滤波效率更高.最后,对概率定位方法的关键技术进行了分析,并探讨了未来的发展趋势.

This paper overviews some popular mobile robot probabilistic localization methods in recent years, analyzes and compares the performances of these methods. All of these methods employ the Bayesian rule as a fundamental theory. Firstly, we introduce the Kalman filter which is extensively used in position tracking, and the Markov localization method which has made many successes in global localization. Secondly, the Monte Carlo method is presented, which uses a particle filter technique and are more efficient computationally. The most recently used adaptive sampling methods are also introduced, and they have demonstrated much better results than the simple particle filter approaches. At last, the key technologies of probabilistic localization methods are analyzed, and the trends of research in the future are discussed.