An optimal scheduling of pick-place operations of a robot-vision-tracking system by using neural networks and rule-based systems

Scope of study. In this research program, a scheduling of pick-place operations of a robot-vision tracking system which handles multi-type and multi-size objects presented randomly on a moving conveyer has been identified and formulated as a real-time repetitive optimization problem. A mathematical model for calculating robot processing times in a constrained environment is constructed, and the strategy to deal with constraint violation situations is also proposed.;Findings and conclusions. The real-time repetitive optimization problem can be viewed as an associative mapping problem. A scheme which uses both the back-propagation and Hamming networks is proposed to implement the mapping in an attempt to overcome difficulties raised by using traditional methods. To improve the training time, an alternative neural network, a Modified ARTMAP, which is based on the ARTMAP (Adaptive Resonance Theory Networks) introduced by Carpenter, et al. has been investigated to solve the optimal scheduling problem. The main modification is that a matching check mechanism is added to the system such that the Modified ARTMAP can cope with situations appearing in the optimal scheduling problem, whereas the original ARTMAP cannot, without using the complement coding technique. The repetitive optimization problem may also be viewed as an on-line decision making problem. A rule-based system can be constructed to conduct the on-line decision making directly if some rules are available. The discrimination net approach is designed for cases where all needed rules can be obtained; the partial search approach is designed for cases where only some of the rules are available. The heuristic approach is proposed for the cases of variable object patterns. Both neural network and rule-based system approaches are implemented, and the experimental results have shown that time savings of up to 21% with only 4 different objects are possible, over first-come, first-served schemes currently used in industry.