Agent-based MST Construction

Minimum-weight spanning tree (MST) is one of the fundamental and well-studied problems in distributed computing. In this paper, we initiate the study of constructing MST using mobile agents (aka robots). Suppose n agents are positioned initially arbitrarily on the nodes of a connected, undirected, arbitrary, anonymous, port-labeled, weighted n-node, m-edge graph G of diameter D and maximum degree Δ. The agents relocate themselves autonomously and compute an MST of G such that exactly one agent positions on a node and tracks in its memory which of its adjacent edges belong to the MST. The objective is to minimize time and memory requirements. Following the literature, we consider the synchronous setting in which each agent performs its operations synchronously with others and hence time can be measured in rounds. We first establish a generic result: if n and Δ are known a priori and memory per agent is as much as node memory in the message-passing model (of distributed computing), agents can simulate any O(T)-round deterministic algorithm for any problem in the message-passing model to the agent model in O(Δ T log n+nlog^2n) rounds. As a corollary, MST can be constructed in the agent model in O(max{Δ√(n)log n log^*n, Δ D log n,nlog^2n}) rounds simulating the celebrated O(√(n)log^*n +D)-round GKP algorithm for MST in the message-passing model. We then establish that, without knowing any graph parameter a priori, there exists a deterministic algorithm to construct MST in the agent model in O(m+nlog n) rounds with O(n log n) bits memory at each agent. The presented algorithm needs to overcome highly non-trivial challenges on how to synchronize agents in computing MST as they may initially be positioned arbitrarily on the graph nodes.