Learning and Exploring Motor Skills with Spacetime Bounds

Equipping characters with diverse motor skills is the current bottleneck of physics-based character animation. We propose a Deep Reinforcement Learning (DRL) framework that enables to learn and explore from reference motions. The key insight use loose space-time constraints, termed spacetime bounds, limit search space in an early termination fashion. As we only rely on specify our learning more robust respect low quality references. Moreover, bounds are hard constraints improve challenging motion segments, which can be ignored by imitation-only learning. compare method state-of-the-art tracking-based DRL methods. also show how guide style exploration within proposed