인공지능을 이용한 휴머노이드 로봇의 자세 최적화

Optimization of Posture for Humanoid Robot Using Artificial Intelligent

This research deals with posture optimization and generalization using genetic algorithm and neural network, development of real humanoid robot platform, and application of proposed optimization method for real robot. A humanoid-robot motion pushing an object is considered. When the robot takes a motion to push an object, the palm and the sole are assumed to be fixed at the object and the ground, respectively, and sense the reaction force from the object and the ground. The reaction force results in change of torques of the joints. This study finds an optimized posture using a genetic algorithm such that torques are evenly distributed over the all joints. Then, a number of different optimized postures are generated from varying the reaction forces at the palm and the sole. The data is to be used as training patterns of MLP(Multi-Layer Perceptron) neural network with BP(Back Propagation) learning algorithm. Using the trained neural network, humanoid robot can find the optimal posture at different reaction forces in real time.