Author:
Toshihiro Matsui
Affiliation:
Nagoya Institute of Technology, Gokiso-cho Showa-ku Nagoya Aichi 466-8555, Japan
Keyword(s):
Distributed Constraint Optimization, Message Loss, Publish and Subscribe, Multiagent System.
Abstract:
Distributed constraint optimization problems have been studied as the bases of decentralized resource allocation and decision making on multiagent systems. The studies address constraint optimization problems distributed on multiple agents and decentralized solution methods. A number of types of solution methods based on different optimization techniques have been proposed, and various applications have been investigated, including resource allocation for sensor networks and power grids, and collaboration for meeting scheduling and disaster responses. One issue with implementation is the low communication quality in some cases of actual devices. A recent study addressed the delay and loss of messages in local search methods. On the other hand, opportunities exist for considering further practical implementation techniques. For example, in real-time systems, a solution process might be required to complete an optimization cycle even if several agents do not respond temporally. In this
paper, we investigate fundamental implementation techniques toward distributed constraint optimization methods on a message-oriented communication platform based on the publish/subscribe model, which is generally employed for IoT devices and robot systems. In our first study, we address the deterministic local search method on the low QoS settings of communication environments with message loss, where the solution method must continue with temporally missing agents. We experimentally evaluate the influence of several approaches to handle temporally non-responding agents in the executing solution processes.
(More)