Robot Development Intrinsic Motivation Research Of The Algorithm

Dominated by driving internal factor which is the strong desire of explore and knowledge of new things, people obtain new knowledge and experience, this driving mechanism is called internal motivation in psychology and its formation process is also a kind of developments process. Biology motivation mechanism is brought into robot system, by the frame of reinforcement learning of internal motivation and robot’s independent exploring and learning of environment, robot’s control of point balance,straight line balance and turning balance is realized, and two-wheel robot’s independent cognitive development is realized. The following is main research of this paper:Firstly, research background of development and internal motivation of cognitive development robot is summarized, internal motivation and reinforcement learning algorithm is analyzed based on the summary. By reinforcement learning based on internal motivation, learning framework of internal motivation reinforcement learning algorithm and definition of reinforcement learning internal motivation is prompted.Secondly, mathematical modeling of dynamics and kinematics has been done separately for two-wheel robot. Reinforcement learning algorithm driven by internal motivation which is based on FRBF(Fuzzy Radial Basis Function) network is proposed and simulated in two-wheel robot model. The analysis of simulation result confirms feasibility of this method.Finally, to improve convergence rate of algorithm, existing reinforcement learning algorithm problems are summarized, reward function, exploration strategy and state space of reinforcement learning is improved. Combining with the improving measure,hierarchical reinforcement learning algorithm of fuzzy space driven by internal motivation is proposed and simulated in balance control of two-wheel robot. It is known from the test result that convergence can be accomplished in shorter time by this algorithm.