|Stochastic Systems Group|
Trackability and Process Query Systems
The concept of generic process tracking is formalized as to identify a group of processes from a large volume of observations made by a sensor network. A real world phenomenon is considered as a process that generates a temporal sequence of internal or hidden states, which is only partially observable. Hidden Markov model (HMM) is a very useful technology to model processes with discrete states. However, the sensitivity of inference performance to disturbances in HMM parameters could be unacceptably high. We'll focus on a rigorous theory of trackability that investigates robustness of using HMMs to solve tracking problems. The parametric factor of trackability for HMMs is analyzed under the framework of Shannon's theory. The structural effect, on the other hand, is studied base on its non-parametric counterpart, the non-deterministic finite automata (NFA).
We are developing a generic framework to solve multi-model and multi-process tracking problem in complex situation and environment, which we call "Process Query Systems". PQS has been applied successfully in many applications including computer and network security, physical object tracking, chemical plume detection and autonomic computing systems. The trackability theory is also used to improve the tracking performance of PQS.
Problems with this site should be emailed to firstname.lastname@example.org