Localisation system for an autonomous mobile platform

In this article, we present the design and development of a localisation system for a mobile autonomous platform. The purpose of this system is to endow this vehicle with the capability of maintaining a position estimate in real time without artificial aids in a structured environment. Each position estimate is generated from the previous one with information obtained from the on board sensors as well as from the world model. Process and observation models are built in such a way that take into account the physical restrictions of the vehicle and its ultra-sonic sonars. All the stages of the extended Kalman filtering (EKF) process applied to the vehicle of interest are described. To increase the performance of the sonar data real time acquisition, a method based on entropy choosing the sonar observation in order to minimise the uncertainty of the position estimate is used. An adaptive process of matching the sonar observation decreasing the possibility of loss of the vehicle is also chosen. Satisfactory results showing the good performance of this localisation system are presented