An Auction-based Approach for Decentralized Multi-Project Scheduling (Doctoral Consortium)

Many scheduling problems in industry practice are considered as Decentralized Resource-Constrained Multi-Project Scheduling Problem (DRCMPSP). Given its characteristics, the problem must be solved in a decentralized manner while respecting information privacy of projects. However, existing approaches fail in satisfying these requirements, and encounter difficulties in dealing with large cases. To this end, we propose a novel approach based on multi-unit combinatorial auction with greedy resource allocation strategy. A bid modification step is also incorporated to provide chances for project agents to improve utilities. This approach does not require private project information, and scales well to large cases. Experiments show that our approach can obtain better solutions compared to state-of-the-art approaches.