Abstract: Observation information reduction is a hot area of the uncertainty planning research in recent years,but these researches concentrate on the single agent’s environment,and the planning problem related to observation information reduction in the multi-agent domain is lack of researching.Confronted with the planning problem in the multi-agent domain,this paper designed an ORMAP algorithm which can find a collaborative planning in the nondeterministic multi-agent domain．At first,the ORMAP algorithm layers all states in the problem domain according to the model-based hierarchical states thought in order to avoid the conflicts between different agents．Then,it searches the collaborative planning solution with the method of the backtracking prior to minimum cost,meanwhile reduces the observation information．At last,a cooperative planning solution can be obtained and it is the one which needs least amount of observation information in all cooperative planning solution to the problem domain,so that it reaches the point.Finally,the experiment shows the efficiency of this algorithm is higher after considering the constraints of the observation information reduction.

Key words: Multi-agent,Intelligent planning,Uncertainty planning,Observation information reduction,Hierarchical state