GMR: Geographic Multicast Routing for Wireless Sensor Networks

We present geographic multicast routing (GMR), a new multicast routing protocol for wireless sensor networks. GMR manages to preserve the good properties of previous geographic unicast routing schemes while being able to efficiently deliver multicast data messages to multiple destinations. It is a fully-localized algorithm (only needs information provided by neighbors) and it does not require any type of flooding throughout the network. Each node propagating a multicast data message needs to select a subset of its neighbors as relay nodes towards destinations. GMR optimizes cost over progress ratio. The cost is equal to the number of selected neighbors, while progress is the overall reduction of the remaining distances to destinations. That is, the difference between distance from current node to destinations and distance from selected nodes to destinations. Such neighbor selection achieves a good trade-off between the cost of the multicast tree and the effectiveness of the data distribution. Our cost-aware neighbor selection is based on a greedy set merging scheme achieving a O(Dn min(D, n)3) computation time, where n is the number of neighbors of current node and D is the number of destinations. This is superior to the exponential computational complexity of an existing solution (PBM) which tests all possible subsets of neighbours, and to an alternative solution that we considered, tests all the set partitions of destinations. Delivery to all destinations is guaranteed by applying face routing when no neighbor provides advance toward certain destinations. Our simulation results show that GMR outperforms previous multicast routing schemes in terms of cost of the trees and computation time over a variety of networking scenarios. In addition, GMR does not depend on the use of any parameter, while the closest competing protocol has one parameter and remains inferior for all values of that parameter