INTELLIGENT CONTROL OF AUTONOMOUS SYSTEMS WILL BECOME A REALITY WHEN SUCH SYSTEMS CAN PERCEIVE SALIENT ASPECTS OF THEIR ENVIRONMENTS IN REAL-TIME.

INTELLIGENT CONTROL OF AUTONOMOUS SYSTEMS WILL BECOME A REALITY WHEN SUCH SYSTEMS CAN PERCEIVE SALIENT ASPECTS OF THEIR ENVIRONMENTS IN REAL-TIME. AUTOMATED VISUAL PERCEPTION, INVOLVING SPECIFIC SUBGOALS SUCH AS AUTOMATIC DETERMINATION OF DEPTH FROM MOTION STEREOPSIS, ARE CLEARLY BECOMING NECESSARY TECHNOLOGIES FOR IMPLEMENTATION OF INTELLIGENT CONTROL SYSTEMS. EXISTING COMPUTER SYSTEMS ARE NOT APPROPRIATE FOR DEDICATED REAL-TIME MOTION ANALYSIS. ADDITIONALLY, ALGORITHMS FOR THESE ANALYSES ARE NOT HEAVILY TESTED BECAUSE THEY CONSUME CONSIDERABLE RESOURCES ON EXISTING COMPUTERS. WE PROPOSE TO ANALYSE THREE PROMISING ALGORITHMS AS CANDIDATES FOR REAL-TIME MOTION ANALYSIS. A MOTION PROCESSING COMPUTER THAT IS CAPABLE OF EXECUTING THESE ALGORITHMS WILL BE DESIGNED AND IMPLEMENTED IN A BREADBOARD FASHION, WHERE ALL COMPONENTS ARE OPERATIONAL. THE FOLLOW-ON PROJECT (PHASE II) WILL CONSTRUCT A PROTOTYPE SYSTEM BASED ON DESIGN DATA DEVELOPED DURING PHASE I. THE EFFORT CONSISTS OF TWO CONCURRENT ACTIVITIES, ONE WHICH ANALYSES THE THREE ALGORITHMS IN DETAIL; THE OTHER WHICH DESIGNS AND CONSTRUCTS THE BREADBOARD COMPUTER. THE EQUIVALENT OF FOUR FULL TIME PEOPLE WILL WORK ON THE PROJECT FOR SIX MONTHS. ONE IS COVERED BY THE FUNDS OF THIS PROPOSAL, WHILE THE OTHER THREE WILL BE SUPPORTED BY INTERNAL VIC FUNDS.