An Enhanced Adaptive Monte Carlo Localization for Service Robots in Dynamic and Featureless Environments

The Adaptive Monte Carlo Localization (AMCL) is a common technique for mobile robot localization problem. However, AMCL performs poorly on localization when robot navigates to a featureless environment. To address this issue, an enhanced AMCL is proposed through using the information from laser scan points to improve the preciseness and robustness of the localization problem for service robots. The proposed new method first matches the laser scan points with a pre-built grid map by an iterative closest point (ICP) algorithm and then designs a Localization Confidence Estimation (LCE) method to evaluate the localization credibility of ICP and AMCL respectively. Finally, the ICPs with high LCE scores are selected to inject particle swarms in the form of particles with adaptive amounts to optimize the next step of the AMCL estimation process. With the improved method, AMCL's particle swarm can quickly converge to the correct position after several iterations. Experimental results show that the proposed algorithm outperforms the original AMCL in respect of accuracy and robustness even in dynamic environments.