Adaptive Control of Multimedia Deliveries in Smart Homes
Originally from Xi'an, China, Zelin is currently undertaking a research project with the Marie Curie Early Stage Research Group, connected to the Institute for Digital Technologies.
- BSc in Electronic Engineering, Xidian University, China, during 2009-2013
- MSc in Environmental Since in Xidian University during 2015-2015. During this period, Zelin was on a double-master program with LU, Loughborough, in Digital Communication System, 2014-2015
- Now studying for a PhD facilitated by MC Cloudscreens, IDT, LU London, since 2015
PhD research description
The Marie Curie Cloudscreens project aims to establish an adaptive control framework in a smart home environment, so that by learning the home environment as well as user’s activities, it can adjust the delivered services to user’s preferences via changes on the related device settings. To achieve it, the following must be considered:
1. Why is adaptive control important in smart homes?
2. How can we design and employ an effective adaptive control method?
3. Where does the input data come from?
Smart home refers to an indoor living environment where different kinds of devices can associate accordingly to provide services that satisfy the end user. The change in user’s condition from time to time would influence the settings of those devices which provide necessary information on user’s preferences. Home adaptive control therefore is essential in achieving smart control to indoor devices, by recognising specific patterns in those devices according to user’s current activity and history of behaviours.
The project proposes a framework for adaptive control where reinforcement learning algorithms will be used for learning and decision making. The change in the home devices and user habits requires that the learning algorithm shall be able to adapt to new conditions and change its decision policy accordingly. Reinforcement learning is a method whereby a new decision for the adaptive control is influenced by both the existing devices’ states and user’s reaction. Instead of simply performing a rule-based, one-step action, the learning feedback in reinforcement learning can better deal with complicated and new situations, making the adaptive control smart that in turn better responds to the user.
Awards, grants or scholarships received
- MSc (distinction) in Digital Communication Systems, Loughborough University, 2015
- Marie Curie Cloudscreens grant, Loughborough University London, 2015- 2018
Papers, publications and articles
For more information about this project, please see the project website here.