Real-time Physics-based Animation of a Humanoid Swimmer

Abstract

We present a real-time physics-based system for animating humanoid swimming. It includes a simplified fluid environment, a controlled virtual character, and a template for imitating different respiratory strategies. Swimmer actuation is carried out using a proportional derivative control strategy and a set of target poses. Balancing and interactive target tracking are achieved by generating additional torques in the controller feedback loop. Breathing patterns are modeled as a customizable piecewise linear function for torso density variation. After testing the system we find the optimal input parameters and investigate the effects of common breathing patterns on the performance of the swimmer. Our preliminary results suggest that buoyant properties using same breathing techniques might have slight effects on front-crawl and backstroke style performances, but are generally negligible.