On-line Estimation of Implement Dynamics for Adaptive Steering Control of Farm Tractors

Abstract

An adaptive control technique for the control of a farm tractor during low levels of excitation and at low velocities is presented. Results of a set of system identification experiments are compared to previous tractor models. A cascaded controller is then designed for the feedback of steer angle, yaw rate, and lateral position baed on the new tractor model. An on-line analysis of the data is used to determine if enough excitation is available for adaptation. A cascaded Kalman Filter is presented to estimate the slope of the DC gain of the steer angle to yaw rate transfer function, MDC, with respect to velocity. An estimator also provides faster updates of position. From the on-line estimate of MDC, the controller gains are scheduled based on a lookup table of predetermined values that were calculated from system identification tests.

The sensitivity of the controller to model simplifications, incorrect velocities, and MDC estimate errors are investigated. The accuracy of the estimated MDC due to neglected dynamics and the rate of convergence is shown. A simulation is used to show the errors that can be induced in the position estimator by the GPS delay. The yaw rate estimator is designed for fixed point math using a square root covariance filter. Experimental and simulation results are provided which show the validity of the MDC estimate. Finally, experimental results which show that the accuracy changes little as a result of hitch loading and velocity are presented and discussed.