Relay node placement for performance enhancement with uncertain demand: A robust optimization approach

The relay node placement problem in wireless sensor networks strongly depends on the data traffic patterns, which can be dynamically changing and not known with precision in advance. In our work, locations of the relay nodes need to be defined with the aim of improving network performance. While in current literature a uniform and constant traffic generation is usually assumed, we propose a robust optimization methodology to deal with the uncertainty in nodes' traffic generation. Solving to optimality the resulting robust optimization problem would be extremely expensive in computational terms, therefore we propose a distinctive robustness measure aiming at dramatically reducing computational costs without sacrificing robustness. The resulting optimization problem is tackled using a bio-inspired meta-heuristic, namely a genetic algorithm. The approach is validated through extensive simulation experiments. Results show that the performance degradation of the robust solutions is minimized in comparison with the solutions obtained ignoring the uncertainty.