Automobile driving support system evolved by Genetic Programming
Go Yakami, Ivan Tanev, Katsunori Shimohara, Shigeru Katagiri, Miho Ohsaki · 2016
We study a new approach, based on Genetic Programming, for generating automobile driving agents (driving rules). In a simulation environment, we develop agents for two tasks: lane departure recovery and risk avoidance lane change. The agents control a car by operating its front wheels using such observations as the distance between the car and the centerline. Our GP process generates an appropriate rule for front wheel operation. We experimentally demonstrate that our GP-based approach successfully generates an effective driving agent (rule).