A hybrid approach for real-time sequencing and scheduling

Abstract

Abstract An approach for real-time sequencing and scheduling is described which integrates artificial neural networks, real-time simulation, genetic algorithms, and a trace-driven knowledge acquisition technique. This approach will be used to solve both single machine sequencing and multi-machine scheduling problems. Single-performance, artificial neural nets are used to quickly generate a small set of candidate sequencing or scheduling rules from some larger set of heuristics. A more detailed evaluation of these candidates is carried out using realtime simulation. This evaluation is necessary to generate a ranking that specifics how each rule performs against all of the performance measures. A further reduction of these candidates can be achieved from this ranking. Genetic algorithms are applied to this remaining set of rules to generate a single 'best' schedule. To capture the knowledge contained in that schedule, a trace-driven knowledge acquisition technique is used. Then, the derived rule is added to the original set of heuristics for future use. In this paper, we describe how this integrated approach is envisioned to work, and provide an example.