Map-based localization for mobile robots in high-occluded and dynamic environments

Purpose – The purpose of this paper is to propose a localizability-based particle filtering localization algorithm for mobile robots to maintain localization accuracy in the high-occluded and dynamic environments with moving people. Design/methodology/approach – First, the localizability of mobile robots is defined to evaluate the influences of both the dynamic obstacles and prior-map on localization. Second, based on the classical two-sensor track fusion algorithm, the odometer-based proposal distribution function (PDF) is corrected, taking account of the localizability. Then, the corrected PDF is introduced into the classical PF with “roulette” re-sampling. Finally, the robot pose is estimated according to all the particles. Findings – The experimental results show that, first, it is necessary to consider the influence of the prior-map during the localization in the high-occluded and dynamic environments. Second, the proposed algorithm can maintain an accurate and robust robot pose in the high-occluded and dynamic environments. Third, its real timing is acceptable. Research limitations/implications – When the odometer error and occlusion caused by the dynamic obstacles are both serious, the proposed algorithm also has a probability evolving into the kidnap problem. But fortunately, such serious situations are not common in practice. Practical implications – To check the ability of real application, we have implemented the proposed algorithm in the campus cafeteria and metro station using an intelligent wheelchair. To better help the elderly and disabled people during their daily lives, the proposed algorithm will be tested in a social welfare home in the future. Original/value – The localizability of mobile robots is defined to evaluate the influences of both the dynamic obstacles and prior-map on localization. Based on the localizability, the odometer-based PDF is corrected properly.