Encoded Sensing for Energy Efficient Wireless Sensor Networks

Energy efficient communication is a fundamental design problem in wireless networks significantly affecting network performance and the lifetime of wireless sensor networks (WSNs). We introduce encoded sensing-an approach for collaborative encoding and transmission of sensors data-that drastically reduces communication energy expenditure in WSN. Encoded sensing exploits the inherent spatial structure in sensed data to adaptively partition a WSN into groups of sensor nodes, so that nodes in each group sense highly correlated values. Each group encodes all individual measurements sensed by its nodes at time t into a single binary sparse codeword via novel minimum distance combinatorial encoding local algorithm. When the codeword's Hamming weight equals w, a subset of w nodes in the group cooperatively transmits a single binary symbol each. Upon receiving the w bits, the sink has enough information to decode a measurement estimate, which is within a small error from each of the group nodes' individual measurements. The error is bounded and guaranteed to satisfy a priori QoS accuracy requirements. We compare encoded sensing to non-cooperative state-of-the-art transmission protocols and demonstrate at least a factor of two in energy savings, without significant loss of measurement quality. Encoded sensing achieves at least 80% the energy savings of theoretically optimal cooperative transmission distributed beamforming architectures. We show by simulations and theoretical derivations that as the size of a node group grows the performance of encoded sensing converges to the optimal transmission energy efficiency.