Abstract
The performance of a vehicle is intimately linked with the decisions and actions taken by its driver. This justify the importance of a good driver model for use in computer simulations to predict vehicle performance, as well as for automatic driving.
Drivers have an internal working model of the inverse dynamics of the vehicle (as a product of years of practise of trial and error) and are able to monitor, assess and process a large set of cue, such as road geometry weather conditions and vehicle dynamics. Schematically, a driver interacts with the vehicle by means of planned manoeuvres that are decided on the basis of the above environment-vehicle model, and according to a desired goal (which may be minimum lap time or other criterions).
The aim of this paper is the definition of a framework for the development of a new class of virtual drivers, based on the Optimal Manoeuvre Method, as the manoeuvre planning block.
The basic idea of the Optimal Manoeuvre Method is the finding of the minimum time solution (or other optimal criterions) between given endpoints on a given track. From a control point of view this is a two point boundary value optimal control problem with trajectory constraints. It allows to find the best way to move the vehicle between the two points. Further details about this method can be found in papers [2], [1].
This paper first presents the state of the art in the field of driver models, then outlines our solutions main features. Briefly, at fixed time intervals, the virtual driver retrieves track information and evaluates the actual vehicle state. These are input to the manoeuvre planning block - which is based on the Optimal Manoeuvre Method - and the outcome is a planned manoeuvre which satisfy the optimal criterion used. This manoeuvre is fed to the execution block to drive the vehicle along the seen path. Because the dynamics of the real vehicle is different form that of the model used for planning, some differences develops between the actual and the planned paths. These are corrected when the manoeuvre is re-planned at the subsequent time interval. A first driver model based on the explained idea is implemented and used to drive a motorcycle multi-body model with 11 degrees of freedom along a U and S shaped curves to show the working of this method.