CS Postdoc Profiles

Research Profile

I work with Dr. Yugyung Lee at the UMKC Distributed Intelligent Computing Lab in University of Missouri – Kansas City, USA. My research area broadly covers the field of Service Oriented Architecture based systems. This includes the World Wide Web Cloud, pervasive environments and other agent-based systems. I am mostly interested in modeling complex systems based on Event-driven SOA (ED-SOA) principles. In our works we have argued that SOA based systems can become much more flexible and intelligent (with respect to situation awareness, complex event processing, self-organization and healing, etc) if they are designed on ED-SOA principles. However, such modeling becomes extremely difficult when the system is open (i.e. no a priori knowledge base can be assumed), dynamic (i.e. changes in the system state are frequent), and non-deterministic (i.e. the changes in system state are hard to predict with certainty). Currently, I am working on the problem of service composition in non-deterministic and dynamic SOA based systems. We have observed that a service composition problem can be translated into an event handling problem when the system is modeled on ED-SOA principles. Thus, instead of a task based or goal-based approach, the user tasks or goals are viewed as user events. Such user events invoke certain interpretations for a set of services (or rather agents hosting these services). The services can also be viewed as events (or device events) that are triggered as a response to an initial (user) event so as to satisfy a set of desired target events. The coordination between the services is maintained by a global authority (an intelligent decision maker) or a group of global authorities. However, the model is distinctively different from current ED-SOA approaches. This is because the model does not need an ESB based event publication/subscription framework where services need to be mapped (subscribed) to published events a priori and then react to the events that are pushed into the ESB. Rather, we propose an agent cognition theory (termed notability theory) based on which automated reasoning can be done by the system itself to understand whether a particular user event can be handled or not at a particular point of time. In this way the system becomes more flexible in adapting to completely new events or events that are similar to previously known events in some aspects. In essence, the model has the potential to become fully proactive (in contrast to push-based reactive models) and intelligent in its mode of working.
Apart from service composition there are other issues that are highly interesting and very much related to the problem at hand. One of these issues that I am also studying is the analysis of social behavior of agents in distributed SOA-like systems. Agents normally work in three modes: (a) autonomous, (b) cooperative, and (c) collaborative. However, such behavioral patterns are mostly localized concerning a subset of the agents constituting the system. Some agents seem to cooperate/collaborate very easily without getting negatively affected by other agents. They form stable societies. However, some cooperation/collaboration cannot lead to a stable society and we have to mathematically analyze: (a) the cause of instability, (b) the degree of instability, and (c) the effect of instability. This sort of analysis will help us to understand the dynamism in the system better and hence, can lead us to a possible instability resolution methodology by re-adjusting the social structure of the agents. In order to continue with my research I am currently involved in developing an agent-based simulation tool (called UbiSim) that can be used to create real-world SOA – based systems. The backbone system of UbiSim is MetaGlue (agent simulation framework by MIT AI Lab). Through UbiSim we can create interesting interactive (virtual) situations and test service composition, situation recognition, conflict/deadlock resolution, social stability assessment and resolution, etc under those situations. We can also validate existing SOA based systems and create a virtual learning environment for such systems where new events and necessary reactions can be understood and learnt over time.

Contact Information

E-Mail: EMAIL OBFUSCATED

Phone: 352 – 278 – 6010

Categories Posted To:

AI / Machine Learning / Robotics / Vision, Databases / Information Retrieval / Data Mining, Mobile / Ubiquitous / Embedded Computing, Social Computing / Social Informatics, Software Engineering