Controlling Humanoid Robots by Means of Genetic Programming
published: Oct. 20, 2009, recorded: September 2009, views: 6320
Report a problem or upload filesIf you have found a problem with this lecture or would like to send us extra material, articles, exercises, etc., please use our ticket system to describe your request and upload the data.
Enter your e-mail into the 'Cc' field, and we will keep you updated with your request's status.
We show the real-world applications of EC (evolutionary computation) to robotics, which is called "evolutionary robotics". Machine Learning techniques can be applied to a robot in order to achieve a task for it if the appropriate actions are not predetermined. In such a situation, the robot can learn the appropriate actions by using trial-and-error in a real environment. GP (Genetic Programming) can generate programs to control a robot directly, and many studies have been done showing this. GA (Genetic Algorithms) in combination with neural networks (NN) can also be used to control robots. Regardless of the method used, the evaluation of real robots requires a significant amount of time partly due to their complex mechanical actions. Moreover, evaluations have to be repeated over several generations for many individuals in both GP and GA. Therefore, in most studies, the learning is conducted in simulation, and the acquired results are applied to real robots. To solve these difficulties, we propose an integrated technique of genetic programming and reinforcement learning (RL) to enable a real robot to adapt its actions in a real environment. Our technique does not require a precise simulator because learning is achieved through the real robot. In addition, our technique makes it possible for real robots to learn effective actions. Based on this proposed technique, we evolve common programs using GP, which are applicable to various types of robots. Using this evolved program, we execute reinforcement learning in a real robot. With our method, the robot can adapt to its own operational characteristics and learn effective actions. The effectiveness of our proposed approach is demonstrated by performing experiments with real humanoid robots.
Link this pageWould you like to put a link to this lecture on your homepage?
Go ahead! Copy the HTML snippet !