Direct Sensorimotor Control for Low-Cost Mobile Tracking Applications

IEEE Journal Publications

The following abstract is from a journal publication titled, "Direct Sensorimotor Control for Low-Cost Mobile Tracking Applications," which has been accepted to, "IEEE Transactions on Industrial Electronics," and will be published in August of 2000.

Due to copyright restrictions I can not post the entire paper on the Internet.

Abstract

A biologically inspired system for tracking objects in a visual scene is presented. The uniqueness of the system is in the absence of a microcontroller to convert sensory information to tracking decisions, reducing power, size, weight and cost of the overall system. The system consists of a mobile vehicle outfitted with a custom analog VLSI architecture for encoding the position of an object of interest in the vehicles's field of view. Once determined, the object of interest retains hysteresis proportional to its size and intensity to limit the potential for distraction by other objects in the sensing environment. The encoded position of the object of interest is directly converted to a series of motor control signals to drive the vehicle in the direction of the object. The motor drive signals are pulse width modulated to control the speed and direction of travel induced by two DC motors via a conventional differential steering arrangement. Neural oscillators are used to drive the DC motors to provide a compact, single-chip system for tracking bright objects. The nature of the system is sufficiently modular so that it can be adapted relatively easily to tracking other features of visual objects and even to objects representative of other sensing modalities.

The system described here is one of the first efforts to fully integrate and apply aVLSI sensorimotor control to a mobile vehicle and to analyze the complete system from a control systems' perspective. The system described here has the advantages of aVLSI integration in its small size (.011 mm2 elements), low power (0.3 mW per element), and fast system response time (1.5 msec from sensory input to motor response).