Session Name: | On All Fours: Creating Realistic Quadruped Motion |
Speaker(s): | Tobias Karlsson |
Company Name(s): | Microsoft |
Track / Format: | AI Summit |
Overview: | Simple pathfinding and pathfollowing has long been a well-understood problem; however, as soon as we step outside our normal comfort zone of bipedal locomotion, many of the shortcuts and much conventional wisdom that has been created and relied on are no longer valid. The most fundamental limitation that is almost universally imposed on a character's locomotion is to assume that a bounding circle describes the character's footprint in the pathfinding world. While this works well for humans and other bipedal upright creatures, it does not work for a quadruped character (e.g. a horse or dog) with an oblong footprint required to navigate an obstacle-rich environment with passages narrower than the character's length, but wider than its shoulder width. This lecture will explain a solution to this problem that does not require the use of more advanced motion planning algorithms or solving hard problems like "the piano mover's problem". |