Animation of quadrupedal animals and perceptual evaluation of their gaits

Abstract

Computer generated animals have become a common feature of today's digitised society, often found in animated films. In computer games highly realistic animals are simulated in real time. High quality characters and their animations are usually generated by hand by skilled artists, but there has also been automatic generation of animation for multi-legged creatures in games like Spore TM (Electronic Arts). In both cases the process is very challenging - in hand animation the timing and movement of the motion is difficult to represent correctly, and in physically based simulations it is equally difficult to set up the correct functions that deal with timing of the motion, and the physical rules associated with a character’s motion. In some cases it is possible to reuse a motion of one character and apply it to that of another. For example in Evan Almighty (Universal Pictures) different motions of one monkey were recorded and applied to virtual monkeys of different species in the film. This process makes the animation task easier as it allows for the reuse of motions that have been created or motion-captured. The main focus of this work has been on realistic quadruped animation driven by knowledge of human perception of animal gaits. Towards this end we have built various tools and tested different methods for the creation, capture and replication of animal motion. We propose a study of how sensitive humans are to different gaits of quadrupeds, and whether it is possible to use the captured motion of one and apply it to a model of another. We also investigate which animations look best for animating flocks and herds of animals.