THIS PAPER DESCRIBES THE DEVELOPMENT OF A MICROCOMPUTER-BASED RECOGNITION SYSTEM, USING A VISION SYSTEM AND A ROBOT, TO LOAD MULTIPLE SIZE BOXES. THE PAPER PRESENTS AN ALGORITHM THAT DETERMINES THE LOCATION AND ORIENTATION OF EACH BOX ON A MULTIPLE LAYER PALLET. THE DESIGN OF THE SOFTWARE SYSTEM, CAPABLE OF DETECTING LOCATION AND/OR ORIENTATION ERRORS, IS DESCRIBED IN DETAIL. THREE MAJOR PERFORMANCE MEASURES, WHICH HAVE NOT BEEN CONSIDERED IN EXISTING ALGORITHMS, ARE INCORPORATED IN THE PROPOSED ALGORITHM: STABILITY CRITERIA, RANDOM INCOMING SEQUENCE AND RESOURCE UTILIZATION, AS WELL AS PALLET AND ROBOT UTILIZATION.