UConn Cookie Information

 

Data Analytics and Fusion	Intelligent systems are envisioned to have the capability of analyzing quasi-stationary sensor data (e.g., time series, images and videos) for achieving comprehensive real-time situation awareness. This data could be generated from heterogeneous sensing devices that observe different types of spatio-temporal evolving processes (e.g., dynamically changing ocean bathymetry, damage evolution on 3-D structural surfaces, growth of anomalies in biological tissues, development of microscopic faults in electromechanical equipment, unfolding of new features as a robot discovers the map of its surroundings, intruders entering in a sensor network, etc.).   Typically, these intelligent systems are part of a larger complex system, for example a networked-control system such as an aircraft engine control system, smart grid, robotic swarm, and distributed sensor network, and they are equipped with myriads of sensing devices. Consequently, large volumes of heterogeneous data are generated that need to be inferred via development of robust, scalable, and computationally efficient tools of optimal sensor selection, data reduction, visualization, fusion, and classification. In this regard, the research at LINKS is focused on developing novel methods of data analytics and information fusion using multidisciplinary concepts derived from machine learning, automata theory, and symbolic dynamics.
Networked Robotic Systems	Networked robotic systems offer several advantages such as: i) task sharing and coordination, ii) reduced cost in design and deployment of individual entities, iii) robustness to failures of a few individuals, and iv) distributed operation for wide area coverage. These complex systems are engineered to achieve a macroscopic (i.e., global) objective by requiring minimal operational complexity at the microscopic (i.e., individual entity) level. Biological Science offers several examples of Systems of Systems (e.g., shoal of fish, flock of birds, and ant colonies) that exhibit emergent group behavior. Thus, a key challenge in engineering sciences is to learn the principles of emergent behavior from these biological systems, and to apply them for control of human-engineered systems, that could then be used for different distributed operations (e.g., cooperative search, rescue, and target-tracking). The objective is to generate simple rules, which govern the operational behavior of individual entities and their interactions, to yield the desired global performance. Click here to view videos of some real-world experiments.
Intelligent Sensor Networks	Distributed sensor networks are deployed for urban area monitoring as well as for Intelligence, Surveillance & Reconnaissance (ISR) operations (e.g., underwater sensor networks). The objective of these sensor networks is to observe and interpret gradually (or rapidly) evolving spatio-temporal phenomenon of the environment in which they are deployed. It is necessary to impart intelligence to these sensor networks such that they can make distributed decisions in the event of anomalies, abrupt environmental changes, and detection of targets. Typical control objectives include maximization of coverage area, minimization of energy consumption, improving the target detection probability, and accommodation of failures. It is also important that the local decisions are consistent with the global objectives. In order to make local decisions, the sensor nodes must have the capabilities of distributed learning and real-time pattern discovery.   Furthermore, since the sensor nodes are usually randomly deployed, energy constraints must be considered to maximize the lifetime of the network because battery replacement can be infeasible and time consuming. This requires proper scheduling protocols to optimize the energy consumption. In this regard, the research at LINKS is focused on distributed learning as well as optimization & control aspects of sensor networks.
Fault Diagnosis and Prognosis	Gradual development of anomalies (i.e., deviations from the nominal condition) may alter the quasi-static behavior of human-engineered complex systems (e.g., avionic systems, sensor networks, aircraft air management system, engine control systems, building HVAC systems, energy systems, smart grids, fuel cells, hybrid electric vehicles, autonomous systems, and power plants). While the complexity remains largely hidden during the normal phase of operation of a complex system, it becomes acutely conspicuous when contributing to rare cascading and catastrophic failures. In particular, a major goal for safe and reliable operation of such complex systems is early diagnosis and prognosis of incipient faults and mitigation of fault propagation for prevention of catastrophic failures without any significant compromise of quality and performance. From these perspectives, the goal of future generation human-engineered complex systems is augmentation of the basic infrastructure with an automated diagnostic and prognostic toolbox for real-time information assessment and condition monitoring that provides sufficiently advanced warnings of emerging faults. In this regard, the research at LINKS is focused on integrating the model-based, knowledge-based and data-driven approaches into a unified PHM architecture based on multidisciplinary concepts derived from Information Theory, Symbolic Dynamics, and Machine Perception.