| Summary: | 碩士 === 國立中正大學 === 電機工程研究所 === 101 === Building a model of humanoid robot with many dimensions, and applying this model to achieve the balance of robot behavior at the same time, it needs a lot of mathematical derivation. It is a difficult challenge for those who are non-professional in robot control. The purpose of the thesis is to obtain the gait of human by Kinect and let robot learn walk by imitating human’s gait.
Because of the structure difference between human and robot, how to map the gait from human to robot’s walking postures is a significant issue. Furthermore, we lack the information of human’s ankle, and it is a key data for walking. In order to solve this problem, we utilize reinforcement learning to adjust postures for each joint so the robot can walk stably and smoothly.
We design a learning structure to deal with the problem of stability on robot’s physical continuous motion. On the other hand, we use the dissimilarity value between two adjacent postures to obtain key postures to increase the efficiency of robot’s learning, and to train these key postures only. Because the stability is most important for adjusting robot’s gait, we set up force sensors on robot’s soles of feet to compute the center of press as the index of stability.
|
|---|
