Sensor fusion for four-wheel steerable industrial vehicles

This thesis addresses the multi-sensor data fusion problem in the tracking of a bi-directional, moderate speed, four-wheel steerable industrial vehicle with substantial load variations. The main contribution lies in the development of an adaptive estimator based on the extended Kalman filter with two-tier, side-slips compensation. The proposed fusion algorithm is robust and offers sufficiently accurate position reports for long dead-reckoning distance. The temporary position measurements from an absolute landmark-based reference system are fused with the periodically available odometry measurements to provide an optimal estimate of the vehicle's states. The vehicle plant is represented using a modified kinematic model to effectively describe the slip bias that causes the vehicle to deviate from its course due to unbalanced loading and tyre conditions. In addition, the substantial side-slips at the wheels during wheels' steer are compensated to improve the integrity of the odometry measurements. The processing of redundant measurements further improves system robustness against noisy observations. To adapt to tyre wear and deflections under varying loads, the odometry encoder's resolution is constantly calibrated to maintain an accurate position estimate. The filter's performance is evaluated at different speeds, loading patterns and maneuvers using data obtained from field trials. Statistical tests are carried out to verify the filter's consistency. In addition, a prototype of the proposed magnetic sensor configuration is developed and its detection performance is analysed.Master of Engineerin