Analog computation over the wireless channel: A proof of concept

In this work, a recently proposed wireless transmission scheme is verified to be feasible in practice. In contrast to standard message transfer, the scheme has the objective to reliably compute at a fusion center some function of the measurements taken by a set of spatially distributed sensor nodes (e.g., maximum temperature, mean pressure). The distinguishing property of the scheme is its ability to exploit the interference caused by concurrently transmitting sensor nodes in order to compute desired function-values much more efficiently than with standard interference-avoiding strategies. This means, at least in theory, a significantly reduced need for wireless resources and energy, which is a major concern in sensor network design. A first proof of concept should demonstrate that these performance gains can be realized in practice. Towards this end, the computation scheme has been implemented on self-developed software defined radio devices that establish an appropriate testbed. Based on this it turns out that in spite of practical impairments such as asynchronous channel-access, fading, and receiver noise, interference can indeed be beneficially exploited for analog function computation in sensor networks. The corresponding system design, however, fundamentally differs from that for standard message transfer.