A vision navigation algorithm based on linear lane model

Abstract

A model intelligent vehicle system (JUTIV-II) is introduced in the paper. The visual system has an omni-direction camera, which enables the view field to be changed with the speed of the vehicle, more information about the lane to be obtained and disturbances in the view field to be reduced. A linear lane model is applied to match the shape of the lane in the view field. A vision-based navigation algorithm-human imitating steering control algorithm-is proposed based on a linear lane model. An artificial neural transfer function is applied in the model to imitate the human driving. The model shows good results under a wide range of parameter input. The direction error angle ranges from -90/spl deg/ to +90/spl deg/ and the departure error distance ranges from -/spl infin/ to +/spl infin/. Experiment results are given in the paper.