机器人定位中稳健的自适应粒子滤波算法

Robust adaptive particle filter in robot localization

为了提高粒子滤波算法在机器人定位中的性能,在基本粒子滤波算法的基础上,引入概率回退的方法对机器人的初始状态进行估计,采用窗口滤波更新粒子集合,根据对机器人位置估计的情况动态更新粒子集合的大小,得到一种改进的粒子滤波算法——稳健的自适应粒子滤波算法。仿真结果表明:该算法充分利用了对机器人位置估计的有效信息,在显著提高算法稳健性的同时,降低了运算复杂度,较好地解决了机器人定位这一非线性非Gauss状态在线估计问题。

The performance of particle filters in robot localization is improved through the use of a robust adaptive particle filter. The novel algorithm introduces probability retrieval to initialize particle sets, uses multi-set resampling to update particle sets, and refreshes particle set sizes according to the estimation state. Extensive simulations show that the proposed algorithm is much more effective than simple particle filters for the improving robustness and the reducing computational complexity, and successfully solved the nonlinear, non-Gaussian state estimation problem of robot localization.