Autonomous vehicle parallel parking design using function fitting approaches

Abstract

One of the most fundamental problems in the development of an intelligent highway system or an autonomous mobile robot system for factory use is to find the necessary input control variables for smooth and safe movement of the vehicle, or robot, between any two configurations. In this paper it is demonstrated that this problem can be converted into one of finding a fitting function which satisfies the boundary conditions. Three curves, a quintic polynomial, a cubic polynomial and a triangular function are developed to perform the parallel transfer manoeuvre which forms the basis of several important manoeuvres such as reverse parking, moving off, negotiating a stationary obstacle, overtaking a moving vehicle, and changing lane. A detailed discussion of the effect of the vehicle's steering angle limit on the feasibility of these manoeuvres is presented. Simulation results using three typical vehicles, a long commercial vehicle, an ordinary car, and a small laboratory robot, travelling along three curves are also presented and discussed. Based on the comparative study, some suggestions for further work are made. Compared with other methods, this approach is simple and provides excellent simulation of human driver techniques. The paper concludes with a focused discussion about the integration of these techniques with satellite based GPS systems for automated vehicle guidance on highways.