Autonomous Mobile Robot Localization Based on Multisensor Fusion Approach

The objective of this paper is to develop a multisensor fusion approach based on particle filter for autonomous mobile robot localization. Sensors used here are ultrasonic sensor and received signal strength (RSS) which is a byproduct of ZigBee communication device. Both two sensors are inexpensive compared with other range sensors. Ultrasonic sensor dominates the localization system in this paper. Transition and rotation error caused by dead-reckoning can be eliminated by using particle filter. Particle filter uses point mass and weight to present probabilistic density. It can be used in state-space which is suitable for estimation state or pose of a mobile robot. Due to beam number of ultrasonic sensor is much less than laser range finder, particle filter will be easy to produce an ambiguous answer. Such as a robot maybe estimated in two similar places in a map. RSS localization system is achieved by transform RSS to distance. If a robot can get its distance relation by receiving up to three signals from different beacons. The position of a robot can be estimated by triangular relation equation. Although the accuracy of RSS localization system is lower than ultrasonic localization system, it can still improve the drawback of ultrasonic system. RSS localization can be used to limit where new particle be generated. By increasing particle weight nearby position estimated by RSS localization, an ambiguous answer estimated by particles can be avoided. Furthermore, convergence time of finding true robot pose can be reduced.