Implementation of Vision- Based Trajectory Control for Autonomous Vehicles.

Abstract

This paper demonstrates building, implementing, and developing a trajectory tracking control system based on computer vision for autonomous vehicles. The main goal of this system is to enforce the autonomous vehicle to be able to track road lane. This system includes a single digital camera, an embedded computer, and a microcontroller board. The digital camera is mounted at the top of the vehicle along its longitudinal axis. It captures a real-time sequence of images during vehicle motion. The captured images are then processed using Open-CV library for Python compiler over Linux operating system. These software packages are running on the embedded computer (Raspberry Pi 2) to obtain geometrical data of road lane. From this data, the observable errors can be determined. These errors are vehicle lateral offset and a heading error. Finally, a steering controller utilizes these errors in control law to compute the steering command. This command corrects offset and heading errors to ensure that the vehicle is in its way. The embedded computer then paths this command to Arduino microcontroller board to adjust the steering servomotor. The proposed implementation also demonstrates the integration between the embedded computer and microcontroller using Ethernet. During this work, a set of autonomous driving experiments is performed. Significant results are obtained that demonstrate the accuracy and robustness of the lane detection and control algorithms.