Toward a path co-processor for automated vehicle control

The paper describes an embedded controller as the basis for high performance navigation tasks. The approach solves Laplace's equation on a discrete grid representing the roadway map. The resulting harmonic potential is free of local minima and a greedy descent minimizes the probability of collision with posted obstacles. In this paper, the authors (1) introduce the important properties of harmonic potential functions for vehicle control applications, (2) describe a numerical procedure to compute harmonic functions on a grid, (3) express nonholonomic constraints as connectivity in the grid, (4) formulate an energy-referenced control scheme to manage vehicle dynamics, and (5) provide estimates for the performance on a four processor DSP architecture.