Research On Path Planning For Bionic Development Model Robot Based On Intrinsic Motivation

In view of the problem of slow speed,difficult whiteboard learning,dimension disasters in reinforcement learning,artificial potential field algorithm and other algorithm which are applied to avoidance obstacle and plan optimal path for two-wheeled robot,inspired by intrinsic motivation of bionics,an extreme learning machine has been proposed.Intrinsic motivation mechanism is combined with Q learning algorithm,artificial potential field algorithm and hierarchical reinforcement learning algorithm respectively to solve the above problems.The main results are as follows:1)Due to reinforcement learning is inefficient and self-developing in obstacle avoidance and path planning of robots,an Q-learning algorithm based on intrinsic motivation has been proposed.The robot can make decisions through motive like humans,and the robot can develop independently and complete the task of path planning well.2)In view of the problem that whiteboard learning leaded to low overall learning efficiency in the early stage of path planning for IM-Q algorithm,a Q learning algorithm based on intrinsic motivation and gravitational field has been proposed.Firstly,the gravitational field was used to simulate the environment to provide a priori knowledge for the algorithm,and then IM-Q learning algorithm was used to train the algorithm,which improved the learning speed and the overall path planning of the robot.3)In most cases,there are moving obstacles.Therefore,a hierarchical reinforcement learning algorithm based on intrinsic motivation was proposed to further enhance ability of obstacle avoidance and path planning.Using the principle of hierarchy,the algorithm divided reinforcement learning algorithm into different layers,thus simplify the structure and solving efficiency of the algorithm,so that the obstacle avoidance and path planning of robot in the unknown complex environment can be realized as well as before.The learning algorithm based on intrinsic motivation provided a novel idea for autonomous learning and path planning of robots in the future.Figure 41;Table 3;Reference 53...