2/14/2013 Megan Kelly
Written by Megan Kelly
Kun Deng, Mechanical Engineering graduate student, received the Best Student Paper Award at the 48th IEEE Conference on Decision and Control in Shanghai, China in December.
Deng co-authored the award-winning paper, “A Simulation-Based Method for Aggregating Markov Chains”, with his adviser Professor Prashant Mehta and ECE Professor Sean Meyn. Deng, Mehta and Meyn are all associated with the Coordinated Science Laboratory (CSL).
A Markov chain – a sequence of dependent random variables – is named after the Russian mathematician Andrey Markov (1856-1922), who first described these models at the beginning of the 20th century. Today, Markov chain models are indispensible to many applications, including biology, computer science and engineering systems.
Meyn, who has authored a book on the subject, described how the group at Illinois makes use of Markov chains for modeling dynamic phenomena in buildings.
“Markov chains are used to describe occupancy evolution in buildings,” Meyn said. “Certain simplified models of thermal dynamics of a building can also be abstracted as large Markov chains.”
A fundamental problem in using Markov chain models in many of these applications is the large dimension of the state space. For example, Markov chains that arise in building applications typically have hundreds of thousands of states. Deng illustrated this complexity with the aid of a thermal model of a building.
“An individual room or a wall in the building typically requires a number of nodes to accurately represent the temperature,” Deng said. “The model for the entire building thus quickly explodes in complexity as the number of rooms increases.”
Deng said that the focus of his research thus is on the “reduction of abstract Markov chain models with very large state space.”
By combining ideas from dynamical systems, information theory and stochastic processes, Deng and colleagues have come up with a computationally efficient algorithm to simplify complex Markov chain models. A salient feature of the algorithm is that it can be implemented solely based on observations of the system of interest.
Although the research reported in the paper is mainly theoretical, Deng has since applied his algorithm to simplify thermal and occupancy models of large buildings. He uses the simplified models to describe macroscopic features of interest that can be used in monitoring and control applications in large buildings.
“Deng’s work can potentially be used to obtain energy-efficient control of heating, ventilation and air conditioning systems during normal building operation, and to synthesize efficient methods for evacuation of people from the building in the event of emergency,” said Deng’s adviser Mehta.
Deng, a third year graduate student, received his M.S. degree (2007) in the Department of Automotive and his B.S. degree (2005) in the Department of Automation from Tsinghua University in China. His research interests include robotics, building systems and energy-saving systems.