A multi-agent system for distributed multi-project scheduling with two-stage decomposition.

Area of Science:

• Operations Research
• Artificial Intelligence
• Project Management

Background:

• The distributed resource constrained multi-project scheduling problem (DRCMPSP) presents significant challenges in optimizing resource allocation and minimizing project completion times.
• Existing decentralized methods often struggle with complex instances and resource conflicts, leading to suboptimal schedules.

Purpose of the Study:

• To propose a novel two-stage decomposition approach for DRCMPSP using a multi-agent system (MAS).
• To minimize individual project makespans and the overall total tardiness cost (TTC) in multi-project environments.

Main Methods:

• Stage one utilizes a forward-backward hybrid genetic algorithm (FBHGA) for generating initial local project schedules.
• Stage two employs a sequential game-based negotiation mechanism to resolve global resource conflicts.

Main Results:

• FBHGA effectively produces high-quality local project schedules.
• The proposed approach successfully handles complex multi-project instances with varying resource conflicts.
• The method demonstrates superior performance in reducing TTC compared to existing decentralized methods, particularly for large-scale and highly conflicting instances.

Conclusions:

• The novel two-stage MAS approach offers an effective solution for DRCMPSP.
• The combination of FBHGA and game-based negotiation provides a robust method for optimizing multi-project scheduling under resource constraints.