Planning Efficient Walking Gaits in Real-Time for Human Characters

Simulating humanoid motion has always been a challenging problem in computer animation. It has become an indispensable task in emerging virtual environment applications such as on-line games. Although there have been much research on this topic, most of them are only applicable to a specific terrain or only for a given set of footsteps. It is rare to have a system capable of automatically generating collision-free lower-body motion for a given terrain profile. In this paper, we have proposed such a system that can generate the motion plan for humanoid walking in real time. We use inverse kinematics to analyze the motion characteristics of human walking on various ground conditions in order to define the key frames of a walk cycle. We use Bezier curve to adjust the motion of a swing leg to avoid collisions with the ground or obstacles. In addition, the system is capable of generating legal and efficient footsteps for the next few steps according to an energy consumption model. According to observation and measured data, we designate appropriate timing for each phase between key frames to make the animation more realistic. Finally, we have succeeded in integrating such a system with a 3D browser and demonstrated that the planner can generate the locomotion motion for human characters in real time