摘要
In distributed constraint optimization, agents executing BnB-ADOPT(+) react eagerly to cost changes: they send non-redundant COST messages to their parents as soon as they receive new messages. We have observed that a lazier reaction (not sending COST messages until a condition is met) substantially decrements the number of messages sent and causes only a small variation in ENCCCs. This approach combines nicely with soft arc consistency maintenance during search. We provide experimental evidence of the benefits of this approach on several benchmarks.
- 出版日期2015-4
- 单位香港中文大学