FOVEAL MACHINE VISION FOR MOBILE ROBOTS USING AGENT-BASED GAZE CONTROL

AN ATTENTION MECHANISM CAN INFLUENCE SYSTEM BEHAVIOR AND PERFORMANCE ONLY TO THE EXTENT IN WHICH SYSTEM RESOURCES ARE ALLOCATABLE. A KEY ROBOT VISUAL SYSTEM RESOURCE IS ACUITY, AND YET THIS IS FIXED IN CONVENTIONAL UNIFORM ACUITY MACHINE VISION; REGIONS WITH LITTLE OR NO RELEVANCE TO THE ROBOT'S MISSION ARE SAMPLED AT THE SAME RESOLUTION AS KEY FEATURES, OCCUPYING VALUABLE SIGNAL BANDWIDTH AND COMPUTATIONAL RESOURCES, AND INCREASING SYSTEM LATENCIES. FOVEAL ACTIVE VISION FEATURES IMAGING SENSORS AND SIGNAL PROCESSING WITH GRADED ACUITY, AND CONTEXT SENSITIVE SENSOR GAZE CONTROL, ANALOGOUS TO THAT PREVALENT THROUGHOUT VERTEBRATE VISION. FOVEAL SYSTEMS OPERATE MORE EFFICIENTLY BECAUSE RESOLUTION IS TREATED AS A DYNAMICALLY ALLOCATABLE RESOURCE. WIDE FIELD-OF-VIEW, LOCALIZED HIGH ACUITY, AND FAST FRAME RATES ARE ACHIEVED WHILE MINIMIZING IRRELEVANT SENSOR DATA. THE MATURATION OF FOVEAL TECHNOLOGY IS HAMPERED BY THE NEED FOR FAST AND INTELLIGENT GAZE CONTROL. THIS PROJECT PROTOTYPES A MOBILE ROBOT FOVEAL VISION SYSTEM WHICH RELIES ON AN AGENT-BASED, GROUNDED, LAYERED ARCHITECTURE WITH INTEGRATED REASONING FOR GAZE CONTROL. THIS ARCHITECTURE FACILITATES THE DEVELOPMENT OF SUCCESSFUL ACTIVE VISION BEHAVIORS (SENSOR POINTING AND PLATFORM MOTION). LEARNING TECHNIQUES WHICH ACCELERATE ATTENTIVE ACTIVE VISION BEHAVIORS BY EVOLVING THEM INTO PREATTENTIVE RESPONSES ARE BEING INVESTIGATED.