A new metaheuristic based approach for the design of sensor networks

• A new metaheuristic based approach for the optimal design of sensor networks.
• Estimation of Distribution Algorithms is enhanced using local search.
• Strategic Oscillations within the context of Tabu Search are used for local search.
• Minimum-cost sensor network design problems of incremental size are solved.

The optimal design of sensor networks consists in selecting the type, number and location of sensors that provide the required quantity and quality of process information by optimizing an appropriate objective function. The problem is multimodal and involves many binary variables, therefore a huge combinatorial optimization problem results. In this work, the design is solved using a metaheuristic based approach. A strategy that combines the advantages of Tabu Search and Estimation of Distribution Algorithms is presented, which is able to solve high scale designs since it can be implemented to run in parallel. Application results of the methodology to the optimal selection of instruments for networks of incremental size are provided.