Application of linear preview control to modelling human steering control

Abstract

A well-known linear optimal preview controller has been investigated for potential application to modelling human steering control. The controller was modified to account for some characteristics of human control: limited understanding of the system being controlled, time delay in the control action, and a neuromuscular system with a finite bandwidth. The standard and modified controllers were compared in terms of their state and preview gains, and also in terms of their response during a lane-change manoeuvre. A technique for dealing with target paths that turn through more than 90° within the preview distance is introduced that overcomes a limitation of previous time domain implementations of linear optimal preview steering control. The results suggest some conditions necessary to identify models of human steering control. Identification of human time delay is likely to require non-previewed disturbances to the system. Identification of a driver's limited understanding of vehicle dynamics is likely to require the controlled vehicle to have non-simple dynamics. Work to identify the steering control models described in this paper from driving simulator experiments has been completed and will be reported in a later paper.