Adaptive state estimation for 4-wheel steerable industrial vehicles

Addresses the multi-sensor data fusion problem in the navigation of a four-wheel steerable industrial vehicle with substantial load variations. An adaptive estimation approach based on the extended Kalman filter is used to realise the multi-model filtering. The vehicle plant is represented using a modified kinematic model to effectively describe the side-slip which causes the vehicle to deviate from its ideal course. In view of the large mass variations and wheels' deflections, a method to constantly calibrate the odometry encoder's resolution is proposed to maintain an accurate position report even with long dead-reckoning distance. The position measurements from a landmark-based local reference system are fused with the odometry measurements to provide an optimal estimate of the vehicle's states. The filter performance is evaluated at different speeds and loading patterns using data obtained from field trials.