Bionic Self-learning of Two-wheeled Robot Based on Skinner's Operant Conditioning

Abstract

Aiming at the problem about the posture balance control of two-wheeled self-balancing mobile robot, a learning algorithm that it is made up of BP neural network and eligibility traces based on the operant conditioning theory is put forward as a learning mechanism of the two-wheeled robot. The algorithm utilizes the characters of eligibility traces about quicker learning speed, higher reliability, ability in resolving effect about delay, and combines with BP neural network, so that the two-wheeled robot can obtain the movement balance skills of controlling like a human or animal by interacting, studying and training with unknown environmental, and realize the posture balance control of the two-wheeled robot. Finally, some simulation experiments have been done in the undisturbed and disturbed environment respectively. The simulation results show that a learning mechanism of the complex learning algorithm consisted by BP neural network and eligibility traces based on Skinnerpsilas operant conditioning theory can realize the posture balance control to the two-wheeled robot, and reveal the better dynamic character and robustness. And it also has the higher research significance in theory and the application value in project.