A Selective Erasure Channel Model for Packet Collision Channels within Wireless Sensor Networks

Wireless sensor networks typically exhibit a collision channel environment. Error correction over collision channels has been frequently treated by investigating packet erasure codes over binary erasure channels (BEC). Modeling collision channels as BEC overlooks the possibility to salvage valuable information from collided bits. Moreover, in often times, sensor network transmissions are sporadic such that packet erasure codes might not be suitable. In this paper, a selective erasure channel (SEC) model is used to characterize the collision channel. Each bit is classified as an erasure or a bit to be detected in accordance with an erasure threshold applied to the corresponding received signal level. Optimal erasure threshold that maximizes the channel capacity is derived. It is shown that the SEC model supports superior capacities which ultimately approach those of BEC at inferior signal-to-interference ratios. Finally, convolutional and turbo codes are tested as candidates for FEC within wireless sensor networks assuming a star topology. Simulation results suggest that it might be worth to adopt turbo codes for such networks given they significantly outperform convolutional codes even within the scope of short packet lengths that feature wireless sensor networks.