Learning vision algorithms for real mobile robots with genetic programming

We present a genetic programming system to evolve vision based obstacle avoidance algorithms. In order to develop autonomous behavior in a mobile robot, our purpose is to design automatically an obstacle avoidance controller adapted to the current context. We first record short sequences where we manually guide the robot to move away from the walls. This set of recorded video images and commands is our learning base. Genetic programming is used as a supervised learning system to generate algorithms that exhibit this corridor centering behavior. We show that the generated algorithms are efficient in the corridor that was used to build the learning base, and that they generalize to some extent when the robot is placed in a visually different corridor. More, the evolution process has produced algorithms that go past a limitation of our system, that is the lack of adequate edge extraction primitives. This is a good indication of the ability of this method to find efficient solutions for different kinds of environments.