This paper addresses target point search methods for course following systems. A central concept in the development of the control algorithms for such systems is that of target point selection. For a given driving situation, target points constitute spatial objectives that the control algorithm strives to realize. The results presented in this paper are based on experiments made with a recently developed new driver model [Schaefer and Schuller, 1999]. The model establishes control in two steps: geometric dynamic planning and plan-to-action mapping. The separation into these two units allows one to investigate the process of target point search independently.

Target point search is conducted for the guidance of a vehicle’s c.g., i.e. a system thats trajectory can be assumed to have ‘differentiable’ curvature profile. The concepts introduced here, however, may easily be generalized to any system whose state transition (i.e. trajectory) may be described locally by instantaneous circles and that has to follow an abstract nominal path in the state space.

A so called situational driving motivation ‘SDM’ is formulated that defines a clear guideline for geometric dynamic planning based on an time isolated situation. A number of different search methods including Preview Point Search, End of Sight Search, Deviation Dependent Preview Point, and a so-called Nestle Point Search, are investigated. The results are evaluated on the basis of the vehicle’s ability to go around a course with a minimum lateral deviation from the nominal course. The results show that the Nestle Curve Search method provides the best performance.

This content is only available via PDF.
You do not currently have access to this content.