Visual Obstacle Avoidance Using Genetic Programming: First Results

Abstract

Genetic Programming is used to create a reactive obstacle avoidance system for an autonomous mobile robot. The evolved programs take a black and white camera image as input and estimate the location of the lowest non- ground pixel in a given column. Traditional computer vision operators such as Sobel gradient magnitude, median filters and the Moravec interest operator are combined arbitrarily. Five memory locations can also be read or written to. The first evolved program is now controlling the robot. When constructing a system, engineers typically practice iterative design, namely instantiating a design, evaluating it, and then modifying it in light of the evaluation. In the current work Genetic Programming can be seen as automating this process by iteratively improve the architecture of the system in fundamental, previously unplanned ways. The system described here successfully navigates in the hallways outside the lab.