ADJUSTABLE AUTONOMY FOR HAZARDOUS ROBOTIC OPERATIONS

RECENT RESEARCH WITHIN ADS HAS DISCOVERED HOW TO APPLY ARTIFICIAL INTELLIGENCE TECHNOLOGY TO PLANNING AND CONTROLLING THE OPERATIONS OF ROBOTS AND OTHER PROCESSES IN UNCOOPERATIVE ENVIRONMENTS. THIS WORK HAS BRIDGED THE GULF BETWEEN SYMBOLIC REASONING AND NUMERIC SERVOING, AND ALSO THE GULF BETWEEN ADVANCE PLANNING AND ADAPTIVE REACTION. ASA BY-PRODUCT OF THIS WORK ADS HAS DISCOVERED HOW TO BUILD ANINTELLIGENT CONTROL MODULE THAT INTERLEAVES REASONING AND ACTING DURING COMPLEX ACTIVITIES, MUCH AS HUMANS DO. THIS PROJECT WILL EXPLOIT THE HUMANE-NESS OF OUR EARLIER WORK, TOALLOW HUMAN SUPERVISION AND CONTROL OF DECISIONS MADE BY ROBOTS. THE KEY TECHNICAL INNOVATION IS AN INTEGRATION OF CONSTRAINTPOSTING WITH METALEVEL REASONING. THIS WILL ALLOW---FOR PHASE II--- SOFTWARE CAPABLE OF PLANNING AND EXECUTING COMPLEX ROBOTIC TASKS IN EITHER A FULLY AUTONOMOUS OR A FULLY DEPENDENT MANNER, OR WITH ANY DEGREE OF AUTONOMY IN BETWEEN. THE DEGREE OF AUTONOMY PERMITTED FOR EACH TASK WILL BE ADJUSTABLE OVER TIME, AS THE ROBOT DEMONSTRATES ITS TRUSTWORTHINESS TO HUMAN OBSERVERS. FOR PHASE I, NASA WILL BRIEF ADS ON THE SPACE STATION TASKS TO BENEFIT FROM ADJUSTABLE AUTONOMY, AFTER WHICH ADS WILL DESIGN THE PHASE II SOFTWARE, ANALYZE THE DESIGN'S LIMITATIONS, AND BUILD A PROOF-OF-CONCEPT DEMO.