The objective of this project is the dynamic management of sensors and actuators in networked systems with applications to minimizing contamination in drinking water networks. A defining feature of modern infrastructures is the prevalence and abundance of real-time sensing and actuation devices. The increased integration of smart communities through Internet-enabled devices will achieve superior system-level infrastructure performance and reliability. Power grids, water systems, and transportation networks share billions of sensors and actuators amongst them. Although the exponential increase in the number of sensing and actuating devices offers an abundance of societal merits, the real-time management of these devices becomes a daunting task for system stakeholders. The dynamic deployment of sensors and actuators, actuators implementing optimal actions, and sensors selectively reporting time- and space-critical data, will lead to significant socio-economic gains. The specific themes addressed by this project are curbing contamination levels in water distribution networks and reducing energy consumption in power grids.
The research goal of this project is to create fundamental scientific methods that guide networked systems stakeholders in the adaptive selection of the most reliable sensors and actuators--amid the inevitable topologically evolution and uncertainty in these systems. The low- or high-frequency topological evolution is a natural consequence of the physical changes in networked systems. For example, the addition of nodes and links in networks causes a change in topology. Related prior work focused on problems of scheduling or one-time placement of sensors and actuators for mostly linear systems. In contrast, this research investigates methods for adaptively selecting sensing and actuating devices as network conditions change, in addition to considering a wide range of control-theoretic metrics. Such an approach significantly enhances the resilience of networked systems and infrastructures to changes in topology, in addition to ensuring robustness against uncertainty. The investigated research also has important impacts on quality control of contamination-free water distribution networks that leverage high-end mobile water sensors traversing pipes and tanks, while acquiring data through wireless communications every few seconds. Exploiting the slow time-scales of water networks, optimal and sub-optimal online algorithms are developed based on semidefinite and mixed-integer programming. These algorithms capitalize on the inherent sparsity of networked systems to obtain the optimal timing and location of decontaminant injections, acting as actuators, while simultaneously sampling data from mobile water sensors. This can ultimately guarantee a minimal level of contamination in drink water networks.
