Abstract
We added a new mechanism for the Distributed Constraint Optimization Problem (DCOP) algo- rithm to the RRS simulator and confirmed its effectiveness through a case of improving Binary Max-Sum, one of the DCOP algorithms. However, the agents can communicate only once within each simulation step in the RRS that adopts synchronous simulation. It isn’t easy to apply the DCOP algorithm that requires communication repeatedly to derive a solution. Therefore, we introduce a pseudo-communication module of the RRS system to use the DCOP algorithm on the RRS simulator in this paper. This module provides the function of processing many message-passings among agents in each simulation step. RMASBench, which is a similar study, does not work under the agent’s partial situational awareness and the commu- nication range limitation. However, our module allows an evaluation of the DCOP algorithm under such situations. In addition, we apply Binary Max-Sum ADVP, which is an improvement of Binary Max-Sum, to the RRS using our module. Finally, we confirm the effectiveness of the module through a study of Binary Max-Sum ADVP on RRS. As a result, we confirmed the effectiveness of our pseudo-communication module through the study that the Binary Max-Sum ADVP converged to a better solution than the Binary Max-Sum. It implied that the extended RRS system worked well for the DCOP algorithm.
