Computer and Information Sciences
University of Delaware
|
|
| Home |
| Curriculum Vitae |
| Research |
| Publications |
| Presentations |
| Bookmarks |
| Search |
|
|
|
RESEARCH |
||
| I am or have been involved with the following research projects: | ||
This research deals with integrating Plug-in Electric Drive Vehicles (EDVs) into the electricity grid. The impetus for this research stems from two trends:
These two trends complement each other --- when parked and plugged into the electricity grid, these EDVs can be used as a large distributed battery that can be used to provide storage and regulate electric power on the grid --- a concept known as Vehicle-To-Grid power or V2G power. At the same time, the EDV owners would be paid for the power services that they provide and this money can be used to offset or partially subsidize the high cost of the batteries in these EDVs. Vehicles with power management and controls capable of this are called Grid Integrated Vehicles (GIVs). To effectively use these EDVs as storage resources, the grid operators require a certain minimum power capacity, which cannot be provided by an individual vehicle. Hence, a group of EDVs need to come together and form a coalition that can provide the required capacity to the grid. In my research, I am formally modeling the coalition formation problem for EDVs and trying to answer some of the open research questions that this problem poses --- for example, how much capacity can a coalition of EDVs report to the grid operators, which vehicles within the coalition should be used to service the power requests, and how can the money be fairly distributed amongst the coalition participants? I am developing algorithms and techniques that address these questions and allow us to evaluate the effect of adding a large group of electric vehicles to the power grid. Publications:
|
||
Multiagent systems are increasingly being used to solve a wide variety of problems in a range of applications such as distributed sensing, information retrieval, workflow and business process management, air traffic control and spacecraft control, amongst others. Each of these systems have to be designed at two levels: the micro-architecture level, which involves the design of the individual agents and the macro-architecture level which involves the design of the agents' organizational structure. In our research, we are primarily concerned with the agents' macro-architecture. At the macro-architecture level, the multiagent designer is concerned with issues such as the number of agents needed to solve the problem, the assignment of tasks and resources to the agents and the coordination mechanisms being used. The design of the agents' macro-architecture is complicated by the fact that there is no best way to organize and all ways of organizing are not equally effective. Instead the optimal organizational structure depends on the problem at hand and the environmental conditions under which the problem needs to be solved. In some cases, the environmental conditions may not be known a priori, at design time, in which case the multiagent designer does not know how to come up with the optimal organizational structure. In other cases, the environmental conditions may change requiring a redesign of the agents' macro-architecture. These are just a few of the hurdles confronting the macro-architecture designer. In our research, we intend to simplify the macro-architectural design by passing on some of the macro-architectural design responsibilities to the agents themselves. That is, instead of manually designing the macro-architecture of our multiagent system at design time, we intend to allow the agents to come up with their own organizational structure at run time. This Organizational Self Design (OSD) will allow the organizational structure to adapt to changing environmental conditions and differences in the problems being solved. Our approach to OSD involves not only coming up with an optimal organizational structure, consisting of roles, role assignments and interaction patterns (for some definition of optimal), but also coming up with the optimal number of agents required to solve the problem. Towards this end, first and foremost, we intend to investigate whether the agents can come up with an optimal organizational structure to solve and coordinate repetitive instances of the same problem (that is problem instances with the same task structure.) This will involve coming up with an optimal set of agents, assigning subtasks to individual agents and imposing a coordination structure on the agents. Next we intend to see how the agents may adapt the organizational structure to incorporate changes in the environmental constraints on the task structure. These changes may include changes in the task arrival rate, changes in the task deadlines, changes in the quality desired or changes in the uncertainty underlying the task itself. Finally, we intend to see how small changes in the problem's task structure can incorporated into the organizational structure without requiring the significant overhead involved with redesigning the organizational structure from scratch. Publications:
|
||
The Network Time Protocol (NTP) is used in the Internet to synchronize computer clocks to each other and to the national standard time (or coordinated universal time -- UTC). Proper time synchronization is needed for a wide range of applications such as air traffic control, distributed databases and DNS servers. In its current implementation, NTP provides accuracies generally in the range of a millisecond or two on LANs and a few tens of milliseconds on global WANs. It can be argued that NTP is the the longest running, continuously operating, ubiquitously available protocol on the Internet. For more information, please refer to the NTP Protocol Page. As a part of the NTP project, I am currently working on implementing a multi-server discrete event simulator for the project. I have also written a test plan that implementors of NTP can use to test for conformance with the protocol. |
||
The UD Genome project aims to build a central repository of annotated gene information for different organisms using various data mining techniques. To build this repository, expressed sequence tags (a short cDNA sequence that is a part of an expressed gene) are collected from the NCBI EST database and are combined into contigs (multiple ESTs that contain overlapping sequences) using the programs Phrap and Phred. The contig data is then annotated with BLAST homology information, Gene Ontology Terms, Domain information, Pathway Information, etc collected from different databases/sources on the Internet. Since, the ESTs and other databases are constantly being updated and since the data in these distributed databases is often overlapping, software agents are used to check for updates and curate/combine the updated information with the preexisting information. All the curated and annotated information is available to the public through a web interface. As a part of the UDGenome project, I was involved in writing the agents that build the contigs from the ESTs and in setting up a Distributed Annotation System (DAS) server which allowed our contigs to be displayed in the Ensembl Genome Browser. More recently, I have been working on the comparison of homologous pathways in different organisms in order to determine the differences in the proteins/domains present in these pathways. Publications:
|
||
Name classification is an integral part of name entity extraction -- the process of identifying a named entity (a noun/object) and determining its type (a protein, a chemical, etc). Name entity extraction is usually forms the first stage of any information extraction program. For example, biologists may be interested in finding all the abstracts that deal with a certain kind of interaction between biological entities (such as phosphorylation). This will require the extraction of the names from the abstracts, the determination of the types of the named entities and subsequently the determination of the entities that participate in the interaction that we are interested in. In our research, we used various machine learning techniques to determine the information sources that may help in the classification of named entities. To this end, we investigated the use of both name internal sources (such as prefixes and suffixes) and name-external sources (such as the context or surrounding word in which the name appeared.) Publications:
|