Robust on-line adaptive footplant detection and enforcement for locomotion

摘要: A common problem in virtual character computer animation concerns the preservation of basic foot-floor constraint (or footplant), consisting detecting it before enforcing it. This paper describes a system capable generating motion while continuously preserving footplants for real-time, dynamically evolving context. introduces detection method that improves classical techniques by adaptively selecting threshold values according to type and quality. The are then enforced using numerical inverse kinematics solver. As opposed previous approaches, we define footplant attaching two effectors whose position at beginning can be modified, order place foot on ground, example. However, corrected posture is needed starts ensure smoothness between unconstrained constrained states. We, therefore, present new approach based anticipation, which computes postures advance, time-evolving parameters, such as locomotion speed type. We illustrate our on-line with modified patterns, demonstrate its ability correct artifacts, sliding, change modify from straight curved walk motion.