A Simple Computational Model for the Estimation of Time-to-Collision

It is critical for animals to respond quickly and accurately to avoid looming objects. Recently computation of time-to-collision (TTC) with a looming object have received much attention. TTC can be specified by the optical variable tau ap thetas/thetas', where thetas is the visual angle subtended by the looming object and thetas' is the derivative of thetas over time. The neural mechanisms involved in the computation of tau have been described previously. In this article we describe a neural network model designed to explain the physiological response properties of those tau neurons. We found that 1/tau can be computed by using a weighted combination of 1/(thetas+B) and thetas where B is a constant. We then built a simple computational model by including an improved back propagation algorithm as a way of simulating the brain response patterns. The output of the model is consistent with the performance of the TTC computation.