Title:To Send or Not to Send: An Optimal Stopping Approach to Network Coding in Multi-hop Wireless Networks

Abstract:Network coding is all about combining a variety of packets and forwarding as much packets as possible in each transmission operation. The network coding technique improves the throughput efficiency of multi-hop wireless networks by taking advantage of the broadcast nature of wireless channels. However, there are some scenarios where the coding cannot be exploited due to the stochastic nature of the packet arrival process in the network. In these cases, the coding node faces two critical choices: forwarding the packet towards the destination without coding, thereby sacrificing the advantage of network coding, or, waiting for a while until a coding opportunity arises for the packets. Current research works have addressed this challenge for the case of a simple and restricted scheme called reverse carpooling where it is assumed that two flows with opposite directions arrive at the coding node. In this paper the issue is explored in a general sense based on the COPE architecture requiring no assumption about flows in multi-hop wireless networks. In particular, we address this sequential decision making problem by using the solid framework of optimal stopping theory, and derive the optimal stopping rule for the coding node to choose the optimal action to take, i.e. to wait for more coding opportunity or to stop immediately (and send packet). Our simulation results validate the effectiveness of the derived optimal stopping rule and show that the proposed scheme outperforms existing methods in terms of network throughput and energy consumption.