Research: Teaching Characters to Adapt to Complex Paths

"Humans are highly adept at walking in environments with foot placement constraints, including stepping-stone scenarios where the footstep locations are fully constrained. Finding good solutions to stepping-stone locomotion is a longstanding and fundamental challenge for animation and robotics," states the abstract. 

The team presented a novel approach to this difficult problem based on reinforcement learning. The paper emphasized the importance of a curriculum and broke down four possible curriculum choices compared to a non-curriculum approach. The researchers tested their approach using a simulated human character, a realistic bipedal robot simulation, and a monster character managing to generate robust motions for challenging stepping stone sequences and terrains.

You can find the full paper here.