Lattice-Based Analog Mappings for Low-Latency Wireless Sensor Networks

Abstract

We consider the transmission of spatially correlated analog information in a WSN through fading SIMO MAC with low latency requirements. A lattice-based analog JSCC approach is considered where vectors of consecutive source symbols are encoded at each sensor using an n-dimensional lattice and then transmitted to a multi-antenna central node. We derive a MMSE decoder that accounts for both the multi-dimensional structure of the encoding lattices and the spatial correlation. In addition, a sphere decoder is considered to simplify the required searches over the multi-dimensional lattices. Different lattice-based mappings are approached and the impact of their size and density on performance and latency is analyzed. Results show that, while meeting low-latency constraints, lattice-based analog JSCC provides performance gains and higher reliability with respect to the state-of-the-art JSCC schemes.