RNR: Reverse & Replace Decoding for Collision Recovery in Wireless Sensor Networks

Interference between concurrent transmissions causes severe performance degradation in a wireless network. This paper addresses interference cancellation to enable simultaneous packet receptions at a node with a single radio in Wireless Sensor Networks (WSN). Interference cancellation is particularly important for WSN as most of its applications rely on convergecast where all the traffic in the network is delivered to a base station leading to a lot of packet collisions. Existing solutions for collision recovery make simplified assumptions such as the availability of one of the collided packets, repeated collisions of the same packets, and the ability to identify the collided packets before recovering them which do not hold for WSNs and most wireless networks. In this paper, we propose a novel collision recovery method called Reverse and Replace Decoding (RnR) for WSNs. RnR entails a physical-link layer design to exploit the raw samples of the colliding signals. It does not rely on the assumptions made in existing work, and can recover all packets from a single collision. To demonstrate its feasibility, we have implemented RnR using GNU Radio on USRP devices based on IEEE 802.15.4 network. Our experiments on a 6-node testbed demonstrate that RnR can successfully decode packets in 95% cases of collisions, and improves the correctly packet decoding rate up to 97.5% compared to standard decoders in the case of collisions. Also, our simulation based on GNU Radio simulator using 25 nodes shows that RnR achieves 4x higher throughput compared to the state-of-the-art collision recovery mechanisms.