We present a novel framework for animating human characters performing fast visually guided tasks, such as catching a ball. The main idea is to consider the coordinated dynamics of sensing and movement. Based on experimental evidence about such behaviors, we propose a generative model that constructs interception behavior online, using discrete submovements directed by uncertain visual estimates of target movement. An important aspect of this framework is that eye movements are included as well, and play a central role in coordinating movements of the head, hand, and body. We show that this framework efficiently generates plausible movements and generalizes well to novel scenarios.

We describe and test a non-linear control algorithm inspired by the behavior of motor neurons in humans and other animals during extremely fast saccadic eye movements. The algorithm is implemented on a robotic eye, which includes a stiff camera cable, similar to the optic nerve, which adds a complicated non-linear stiffness to the plant. For high speed movement, our "pulse-step" controller operates open-loop using an internal model of the eye plant learned from past measurements. We show that the controller approaches the performance seen in the human eye, producing fast movements with little overshoot. Interestingly, the controller reproduces the main sequence relationship observed in animal eye movements.