Green-Waved traffic control is one of the most efficient strategies in allowing continuous traffic from major directions flow over multiple intersections to improve urban transportation efficiency. When the number of traffic lights scales up, traditional centralized control suffers a bottleneck in both communication and computation. Decentralized control is potentially inefficient when local traffic lights only gain very limited observations to the whole network. This paper proposes a decentralized, multi-agent based schema to adaptively control massive traffic lights, which promotes the effects of green-wave. The key is that agents use the prospection of local state one time ahead as evidence to support decisions. Noting that only the traffic from the adjacent intersections affect the next state of a given intersection, the study models the interactions as decentralized agents to cooperatively coordinate each intersection by using decision theoretical models. This paper presents the algorithm and simulation results to prove the feasibility of the approach to massive urban transportation system.