TY - JOUR
T1 - A Multi-Agent Learning Approach for the Job Shop Scheduling Problem
AU - Martinez Jimenez, Yailen
AU - Nowe, Ann
PY - 2009
Y1 - 2009
N2 - In this paper, a Multi-agent learning approach for the JSSP is proposed, a Q-Learning algorithm is implemented assigning an agent per machine, which will decide which job will be processed next out of a set of currently waiting jobs at the corresponding resource. The behavior of the algorithm is studied using three different cost functions, and on different benchmarks problems from the OR-Library. The approach is compared with other heuristic approaches such as ACO. The results show that our approach yields solutions as good as better than the ACO approach.
AB - In this paper, a Multi-agent learning approach for the JSSP is proposed, a Q-Learning algorithm is implemented assigning an agent per machine, which will decide which job will be processed next out of a set of currently waiting jobs at the corresponding resource. The behavior of the algorithm is studied using three different cost functions, and on different benchmarks problems from the OR-Library. The approach is compared with other heuristic approaches such as ACO. The results show that our approach yields solutions as good as better than the ACO approach.
KW - Reinforcement Learning
KW - Job Shop Scheduling
M3 - Editorial
SP - 259
JO - Book of Abstracts of the 23rd European Conference on Operational Research
JF - Book of Abstracts of the 23rd European Conference on Operational Research
ER -