Information-directed routing in sensor networks using real-time reinforcement learning
In this paper, we demonstrate the benefits of using real-time reinforcement learning for information-directed routing that jointly optimizes for maximal information gain and minimal communication cost. We show that: (1) for tracking targets, the learning algorithm, compared to the previous greedy algorithm, is able to route around the sensor hole after a couple of packets and to track the moving target effectively; and (2) for querying targets, the learning algorithm performs well in both situations with known or unknown destinations.
Zhang, Y. ; Liu, J. J. ; Zhao, F. Information-directed routing in sensor networks using real-time reinforcement learning. In Combinatorial optimization in communication networks, edited by M. X. Cheng, Y. Li, and D-Z. Du. Springer; 2006; 259-288.