Multi-agent collaborative service and distributed problem solving

In this paper, we show that collaborative services can be formulated into distributed constraint satisfaction problems. We introduce the notion of multi-agent collaborative service (MACS), and employ a distributed discrete Lagrange multipliers (DDLM) method to automatically handle an MACS task. The DDLM method is based on a distributed multi-agent system. The behaviors of agents are guided by predefined DDLM rules. In order to make it more efficient in achieving a solution state, we incorporate strategies for tuning the Lagrange multipliers. We validate the effectiveness of the DDLM method with benchmark SAT problems. Furthermore, we provide the mathematical properties of DDLM and present the corresponding DDLM algorithms.