Distributed Routing Protocol Based on Biologically-Inspired Attractor Selection with Active Stochastic Exploration and a Short-Term Memory

Distributed routing algorithms that require only local information of a network attract much attention recently in order to realize fault-tolerant and low-overhead routing. Lack of global knowledge, however, often makes it difficult to promptly respond to traffic changes on paths out of the local scope. In order to overcome the difficulty, we need to accommodate distributed algorithms with exploration mechanisms of nonlocal information of a network. Here, based on the biologically inspired attractor selection model, we propose a distributed routing protocol with stochastic and a periodic information exploration. In order to avoid excess flapping caused by the exploration, the model has a short-term memory and automatically returns, if necessary, to the memorized state after explorations. Using numerical simulation, we confirm that the proposed mechanism successfully balances rapid exploration and stable routing. We show that response time to traffic change on nonlocal paths is actually reduced in the proposed protocol.