A greedy topology design to accelerate consensus in broadcast wireless sensor networks

• Reaching a distributed average consensus is an important topic for Wireless Sensor Networks (WSNs).
• Improving the convergence speed of such consensus algorithms translates into energy saving for the sensor nodes.
• The problem of accelerating consensus in broadcast WSNs can be translated into a range assignment (RA) problem.
• RA is an NP-hard problem.
• Two different greedy techniques are proposed, which outperform state-of-the-art and meta-heuristic solutions.

We present techniques to improve convergence speed of distributed average consensus algorithms in wireless sensor networks by means of topology design. A broadcast network is assumed, so that only the transmit power of each node can be independently controlled, rather than each individual link. Starting with a maximally connected configuration in which all nodes transmit at full power, the proposed methods successively reduce the transmit power of a chosen node in order to remove one and only one link; nodes are greedily selected either in order to yield fastest convergence at each step, or if they have the largest degree in the network. These greedy schemes provide a good complexity–performance tradeoff with respect to full-blown global search methods. As a side benefit, improving the convergence speed also results in savings in energy consumption with respect to the maximally connected setting.