A novel low-complexity statistical discrimination technique for collision detection in wireless sensor networks based on short pilot periods transmission

Power consumption is the critical design goal for Wireless Sensor Networks (WSNs) that comprise large number of sensors which operate with limited power resources such as small batteries. Sensors usually follow various MAC protocols which allow accessing to wireless communication channels. However, sensors may access to wireless channels at the same time, potentially, leading to collisions among multiple nodes. This situation may exacerbate since the rate of collision may increase with increasing the number of sensors which have packets ready to transmit. In central nodes (receivers) of WSNs, it is necessary to decode every received packet regardless of the fact that the transmission may suffer from packets collision. In other words, central nodes need to expend a significant amount of energy and processing complexity in order to fully-decode a packet, only to discover the packet is illegible due to a collision. In this paper, we propose a suite of novel, yet simple and power-efficient technique to detect a collision at the receiver side without the need for full-decoding of the received packet. Our novel approach aims at detecting collision through fast examination of the signal statistics of a short snippet of the received packet via a relatively small number of computations over a small number of received IQ samples. Hence, operating directly at the output of the receiver´s analog-to-digital-converter (ADC) and eliminating the need to pass the corrupted packets through the entire demodulator/decoder line-up. We present a complexity and power-saving comparison between our novel scheme and a conventional full-decoding scheme (MAX-Log-MAP) to demonstrate the significant power and complexity saving advantage of our approach.