Wireless Sensor Networks (WSNs) are prone to failures. To be robust to failures, the network topology should provide alternative routes to the sinks so when failures occur the routing protocol can still offer reliable delivery. We define
-CRC, a new centrality index which measures a node’s importance to connectivity and efficient delivery in the network. We then use this centrality index to concentrate on the most important nodes, providing alternative paths around the nodes with high centrality. Varying
-CRC allows us to trade off cost for robustness. We introduce GRASP-ABP, a local search algorithm for initial robust topology design. We evaluate the algorithm empirically in terms of the number of additional nodes it suggests and its runtime. We then evaluate the robustness of the designs against node failures in simulation, and we demonstrate that the centrality-based GRASP-ABP’s designs are able to offer reliable delivery, comparable to competitor algorithms, but with fewer additional relays and faster runtime.