In this paper we consider policies for cooperative, decentralized traffic management among a number of autonomous mobile agents. The conflict resolution problem is addressed considering realistic restrictions on possible maneuvers. We formulate this problem as one in Mixed Integer Linear Programming (MILP). The method, which proves successful in a centralized implementation with a large number of cooperating agents, is also extended to a decentralized setting. Conditions for the existence of conflict avoidance maneuvers for a system of 5 autonomous agents with a transitive information structure are provided, along with the explicit policy to be applied by each agent.