Biologically inspired analog IC for visual collision detection

We have designed and tested a single-chip analog VLSI sensor that detects imminent collisions by measuring radially expanding optic flow. The design of the chip is based on a model proposed to explain leg-extension behavior in flies during landing approaches. We evaluated a detailed version of this model in simulation using a library of 50 test movies taken through a fisheye lens. The algorithm was evaluated on its ability to distinguish movies ending in collisions from movies in which no collision occurred. This biologically inspired algorithm is capable of 94% correct performance in this task using an ultra-low-resolution (132-pixel) image as input. A new elementary motion detector (EMD) circuit was developed to measure optic flow on a CMOS focal-plane sensor. This EMD circuit models the bandpass nature of large monopolar cells (LMCs) immediately postsynaptic to photoreceptors in the fly visual system as well as a saturating multiplication operation proposed for Reichart-type motion detectors. A 16 x 16 array of two-dimensional motion detectors was fabricated in a standard 0.5µm CMOS process. The chip consumes 140 µW of power from a 5 V supply. With the addition of wide-angle optics, the sensor is able to detect collisions 100-400 ms before impact in complex, real-world scenes. Index Terms-CMOS imager, collision detection, Gilbert multiplier, insect vision, neuromorphic systems, optic flow, smart sensor.