Double-resampling based Monte Carlo localization for mobile robot

The computational efficiency of Monte Carlo localization (MCL) for mobile robots mainly depends on the weight updating of particles. A double-resampling method which adapts the sample size in MCL is presented in this paper. The first resampling with fixed sample size mitigates the weight degeneracy and improves the diversity of particles for prediction. The second sparse resampling reduces the number of particles for updating using a particle merging technique based on rational distribution of spatial particles. Decreasing the weight updating computation and enhancing the prediction capability of particles, the double-resampling method improves the efficiency of the filtering while guarantees the accuracy of the estimation. Simulation and experiment results show that the double-resampling approach can adapt the sample size efficiently and that the double-resampling based MCL for mobile robot is highly efficient and robust.