Realistic collision avoidance of upper limbs based on neuroscience models

When articulated figures interact in a 3D environment, collisions are highly likely and must often be avoided. We present a method automatically producing realistic collision-free animation of the upper arms. Based on the latest models of collision avoidance provided by neuroscience, our method allows realistic interpolation of keyframes at interactive speed. In order to validate our scheme we compared computer generated motions with motions performed by a sample of ten humans. These motions were defined by start and final postures and by an obstacle which had to be passed. In each case the generated positions are the same as those chosen by 30% of real humans, we therefore consider our method provides realistic motions. Moreover, the collision-free paths are automatically generated in a few seconds. Hence, our method can be very beneficial to animators by reducing the level of detail needed to define motions of articulated figures. It can also be used for the automatic generation of realistic animations for virtual reality applications.