Anytime Heuristic for Determining Agent Paths That Minimize Maximum Latency in a Sparse DTN

Mobile data relay agents (or ferries) have the potential to build wireless backbones over which sparsely located, unconnected nodes can communicate in a delay-tolerant fashion. Paths taken by agents determine the quality of such delay tolerant networks (DTNs). We consider the problem of determining agent paths that minimize the maximum pair wise latency in the network. We introduce the concept of Bounded Edge-Count Diametric Latency Minimizing Steiner trees (BECDLMST) where each edge represents an agent path and non-leaf nodes represent rendezvous points. Using ideas from particle swarm optimization, we propose an anytime algorithm to generate a near optimal BECDLMST. Beginning with a simple Minimum Diameter Steiner Tree (MDST), we apply an anytime heuristic and iteratively evolve the tree to produce a network structure that minimizes the maximum latency. Network latency results are presented for varying problem sizes in terms of number of nodes and number of agents.