Emulation of Human Walking by Biped Humanoid Robot with Heel-Contact and Toe-Off Motion

Abstract

The foot mechanism and pattern generation we propose for a biped humanoid robot achieves human-like heel-contact toe-off walking. Placing its heel and toes simultaneously on the ground while walking makes it difficult for a robot to take long strides and mimic human gaits. To solve this problem, we developed a biped foot with one passive toe joint based on human gaits analyzed by motion capture and propose pattern generation for human-like walking based on the zero moment point (ZMP) criterion. Using a genetic algorithm, we optimize walking parameters to generate continuous, smooth foot trajectories. Results of experiments in heel-contact toe-off walking in ground reaction force (GRF) using the biped humanoid robot WABIAN-2 demonstrated the similarity between proposed and human walking.