On October 1st, Fabian Castaño has defended his PhD thesis with success. His presentation lasted for 45 minutes and has been followed by 1h of hard questions where Fabian was able to answer with precision. The jury deliberated for a short time and concluded that Fabian deserved the title of PhD. Please call him "Doctor Fabian Castaño" from now on ;-)
Title: Decomposition-based approaches for the design of energy efficient wireless sensor networks
Abstract: Energy is a major concern in wireless sensor networks (WSN). These devices are typically battery operated and provided with a limited amount of energy. As a consequence, the time during which sensors can monitor the interesting phenomena and communicate through wireless signals might be limited because of (sometimes) irreplaceable batteries. Additionally, it is very common for WSN to be used in remote or hostile environments which possibly makes necessary a random placement strategy (by using an airplane, a drone or a helicopter). Hence, the sensors location is not known a priori and approaches to efficiently use the energy are needed to answer to network topologies only known after sensors deployment. This thesis explores the use of column generation to efficiently use the energy in WSN. It is shown that column generation can be used as a general framework to tackle different problems in WSN design. Several versions of the problem and models for the operation of the WNS are adapted to be solved through column generation. These approaches take advantage of the natural way that column generation offers to consider different features of the WSN operation. Additionally, some computational improvements are proposed to keep the column generation method operating as an efficient exact approach. Hybrid strategies combining column generation with (meta)heuristic and exact approaches are considered and evaluated. The computational experiments demonstrate the efficiency of the proposed approaches and provide practitioners on WSN research with strategies to compute upper bounds to evaluate heuristic centralized and decentralized approaches. Finally, some future directions of research are provided based on the performance and adaptability of column generation to consider more sophisticated models and characteristics newly introduced in sensor devices.
You can access the slides of the presentation here. You can also download his manuscript.