Planning and Synthesizing Superhero Motions

摘要

This paper presents an approach to planning and synthesizing collision-free motions of characters with extreme physical capabilities, or superheroes, using a human motion database. The framework utilizes the author's previous work on momentum-based motion editing, where the user can scale the momentum of a motion capture sequence to make more or less dynamics motions, while maintaining the physical plausibility and original motion style. In our new planning framework, we use a motion graph that contains all possible motion transitions to list the candidate motion segments at each planning step. The planner then computes the momentum scale that should be applied to the original motion segment in order to make it collision-free. Experimental results demonstrate that the planning algorithm can plan and synthesize motions for navigating through a challenging environment, using a relatively small motion capture data set.