Probabilistic Localization of Service Robot by Sensor Fusion

In recently years, various localization methods using two or more sensors have been proposed. Though it is still difficult to cope with the situation that the sensor data are polluted. In this paper, we proposed a probabilistic localization of service robot using laser scanner and indoor GPS system for treating sensor polluted situations. This scheme is a sample based algorithm as an application of Monte Carlo localization. In prediction phase, a number of samples are scattered by referring to the odometry data and position data of indoor GPS system. And those are evaluated by map-matching method using laser sensor and distance-error-matching using indoor GPS system in update phase. We are able to get more robust localization results by using these processes, even though the floor condition is deteriorated and there are many or huge obstacles. And it is possible to cope with severe errors which are caused when the robot is blocked by unexpected obstacles for a long time. However, adding another sensor and manipulating its data makes the computational consumption become larger. For overcoming these shorts, we had improved sensor model. Experimental results demonstrate the evaluation of the robustness of our algorithm fusing range image sensor and indoor GPS system data