The goal of qualityofservice (QoS) multicast routing is to establish a multicast tree which satisfies certain constraints on bandwidth, delay and other metrics. The network state information maintained at every node is often imprecise in a dynamic environment because of nonnegligible propagation delay of state messages, periodic updates due to overhead concern, and hierarchical state aggregation. The existing QoS multicast routing algorithms do not provide satisfactory performance with imprecise state information. We propose a distributed QoS multicast routing scheme based on traffic lights, called QMRI algorithm, which can probe multiple feasible tree branches, and select the optimal or nearoptimal branch through the UR or TL mode for constructing a multicast tree with QoS guarantees if it exists. The scheme is designed to work with imprecise state information. The proposed algorithm considers not only the QoS requirements but also the cost optimality of the multicast tree. The correctness proof and the complexity analysis about the QMRI algorithm are also given. In addition, we develop NS2 so that it is able to simulate the imprecise network state information. Extensive simulations show that our algorithm achieves high calladmission ratio and lowcost multicast trees with modest message overhead. The goal of qualityofservice (QoS) multicast routing is to establish a multicast tree which meets certain constraints on bandwidth, delay and other metrics. The network state information maintained at every node is often imprecise in a dynamic environment because of nonnegligible propagation delay of state messages The existing QoS multicast routing algorithms do not provide satisfactory performance with imprecise state information. We propose a distributed QoS multicast routing scheme based on traffic lights, called QMRI algorithm, which can probe multiple The proposed algorithm considers only the QoS requirements but also the cost optimality of the multicast tree. The correc tness proof and the complexity analysis about the QMRI algorithm are also given. Extend simulations show that our algorithm achieves high calmission ratio and lowcost multicast trees with modest message overhead.