An Adaptive Control Model for Lateral Path Following with Closed-Loop Handling Simulations 971061

For the purposes of modeling vehicle response in lateral lane deviation maneuvers, a path follower that utilizes adaptive control has been developed and implemented into vehicle simulation software. The path follower utilizes input-output, model reference adaptive control so that no a priori knowledge of the vehicle system is required.

The reference model response is generated by a second order linear dynamic model and represents the desired lateral lane deviation trajectory of the vehicle under the conditions of a single maneuver. The controller utilizes any steering degree of freedom of the vehicle model as the input. Therefore different vehicle model structures can be used without modification of the controller. With its adaptive structure, the vehicle model can be made to follow the desired trajectory by minimizing the error between its trajectory and the trajectory response of the reference model. Adaptation of the controller gains occurs over a period of time, and when they stabilize the vehicle simulation model responds similarly to the model reference response. Furthermore, the controller will adapt to each vehicle as well as to the particular maneuver of interest.

The controller has been implemented into 3 D.O.F functional vehicle models for initial evaluation and then into the simulation package ADAMS^™ for application purposes. With ADAMS^™ models that have more than 200 degrees of freedom, the controller has been used to simulate lateral lane deviation maneuvers.

This paper demonstrates the formulation and usage of the path follower for lateral deviation maneuvers (i.e lane changes and slalom). Various vehicle models of interest have been exercised using this controller, and the resulting adaptive control responses demonstrate that the controller will adapt to each vehicle. Furthermore, the adaptation process for each vehicle will provide a different set of steady state control gains indicating that the controller was required to drive the vehicle model differently to obtain the desired response. This allows for evaluation of the vehicle system response alone because the controller does not consider torque and proprioceptive feedbacks in its control structure. This effectively removes driving style from the analysis and provides a more consistent means to evaluate closed loop maneuvers for vehicle response.

On-vehicle measurements have been performed to serve as a means of comparison with modeled simluation responses of the controller/vehicle system and to establish reference model trajectory responses for the controller. A specialized experimental procedure for the lane change maneuver has been prescribed so that on-vehicle measurements will be consistent for a given driver used in the testing. Comparisons between on-vehicle measuremets and the modeled controller responses indicate that the controller does not initiate extremely uncharacteristic control inputs for the given handling situations.