Abstract

Although communication is generally considered to dominate over processing cost in distributed systems, the problem of communication cost in multiagent planning has not been sufficiently addressed. One method for reducing both communication cost and planning time is the use of social laws. Social laws, however, can be too restrictive, limiting soundness. Flexible social laws can enable multiagent systems to reap the benefits of reduced communication cost and planning time (except in the worst case), without limiting soundness (although optimality may be degraded). By analyzing the performance, we show that this model can make multiagent planning exponentially more efficient without limiting its applicability. 1

Citations