Abstract—One of the most common communication patternsin sensor networks is routing data to a base station, whilethe base station can be either static or mobile. Even instatic cases, a static spanning tree may not survive for along time due to failures of sensor nodes. In this paper, wepresent an adaptive spanning tree routing mechanism, usingreal-time reinforcement learning strategies. We demonstratevia simulation that without additional control packets fortree maintenance, adaptive spanning trees can maintain the“best” connectivity to the base station, in spite of nodefailures or mobility of the base station. And by using ageneral constraint-based routing specification, one can applythe same strategy to achieve load balancing and to controlnetwork congestion effectively in real time.

Index Terms—constraint-based routing, real-time reinforcementlearning, routing tree, wireless sensor networks

Cite: Ying Zhang and Qingfeng Huang, "A Learning-based Adaptive Routing Tree for Wireless Sensor Networks," Journal of Communications, vol. 1, no. 2, pp. 12-21, 2006.