Vehicle localization system using offboard range sensor network

Most mobile vehicle localization approaches rely on onboard sensors. However, whenever the installation of sensors onboard the vehicle is unfeasible, an alternative solution is to install them in the environment. One such environment is the ITER nuclear fusion reactor, where all maintenance operations have to be performed by remote handling, due to the radiation levels. This paper addresses the problem of vehicle localization in a structured environment, using a network of laser range finder sensors. The approach taken is based on: (1) the optimization of the sensor placement in the environment, aiming at the maximization of the area covered by the sensors and the redundancy of the sensor network, and (2) a probabilistic approach for vehicle localization. Two localization methods were evaluated: Extended Kalman Filter and Particle Filter. These two methods are compared, with respect to localization performance and robustness, both in simulation and using a real vehicle in a mock-up scenario.