| |
| Research |
| Grants |
| Students |
| Publications |
| Books |
| Of Interest |
| Teaching |
| Home |
| Dr. Mili's Research | |
Dr. Mili's research activities center around three inter-connected themes: 1. Distributed computing: algorithm development, optimization, adaptability, performance (see projects 1, 2, 3). 2. System validation: Program verification, formal methods in computing, fault-tolerance and system security (see projects 3, 5). 3. Knowledge modeling: data and knowledge modeling, semantic consistency, semantic reconciliation between multiple sources (see project 4). |
|
| __________________________________________________________________________ | |
Completed and Current Research Projects: |
|
1: Query Optimization in Sensor Networks: We use an information-theory approach to minimize power usage in sensor networks. Power is saved through selective querying where the selection of nodes chosen to participate is based primarily on the mutual-information between the query and the node. We have also used a scouting approach whereby a minimal coverage is provided using static nodes. These static nodes are supplemented with mobile nodes. We have developed algorithms for the mobile nodes to make the decisions of when to move, where to move, and how far to go. 2: Adaptive Distributed Computing: When approximate results are acceptable, large gains can be made in the performance of distributed systems. The difficulty is that the accuracy of the results will often 1. depend on the dynamic environment, 2. is not directly accessible. In this project, we develop an approach in which the distributed system monitors the level of activity of its environment and adjusts its resource usage accordingly. The premise of our approach is that high accuracy requires fewer resources when the environment is in a quiescent state and more resources when it is in a highly dynamic state. 3: Intrusion Prevention and Detection in Sensor Networks: Security is an increasingly critical issue in all networked systems. Security in sensor networks presents unique challenges and unique opportunities. In this project we use a mix of bio-inspired methods and economical-based approaches to the prevention and detection of intrusion in sensor networks. 4: Semantic discovery for Sensor-Web-Enablement: The OGC's Sensor Web Enablement (SWE) standards enable developers to make all types of sensors, transducers and sensor data repositories discoverable, accessible and useable via the Web. In this project, we research approaches to deducing semantics of sensor networks using existing or readily available meta-data. 5: Computer Program Construction: Formal Methods are used in Computer Science to create precise models of software artifacts and enable us to reason about them and prove properties of interest. Earlier work focused on formalizing the process of creating a computer program from its specification by modeling the specification as relations and the creation process as a set of relational transformations that preserve essential properties (correctness) while imparting concreteness and order of operations. This work was funded by NSF and resulted in two books and a number of journal and conference papers. The discipline of establishing program properties through formal proofs is now a relatively mature field in Computer Science when the programs are traditional sequential programs. By contrast, little prior work exists on the formal specification and verification of Artificial Intelligence programs and Cognitive Models. We have developed a modular specification language specifically adapted to the needs of cognitive modeling.
|
|