Research

Current Research Areas

My current research interests are principally in the following areas:

• Mobile And Wearable Computing (activity and gesture recognition, IoT-based sensing, wireless energy harvesting, novel wearable interfaces) More...
• Urban And Socio-Physical Analytics and Smart City Services (spatiotemporal data mining, urban event and anomaly detection) More...
• Urban Services and Applications (Mobile Crowdsourcing, Smart Venues and Testbeds) More...
• IoT-Enhanced Smart Spaces (Smart Manufacturing, Occupancy-Aware Smart Building Systems) Less...

This is a more recent, active body of work that is looking at new technologies to use IoT devices for energy-efficient, real-time, fine-grained tracking of physical world context. Selected ongoing activities include:

· Smart Equipment Monitoring: We have been developing new ways to combine wearable and IoT devices to sense the condition of workers and machines in future smart factories. As an example of early work, we have shown [paper] how a low sampling-rate camera (e.g., mounted on a worker’s smartglass) can be combined with intelligent strobing of light sources to remotely detect the vibration frequencies of different machinery. In ongoing work, we are expanding this concept to capture a wider variety of factory context, by incorporating additional sensors (e.g., short-range radar).

· Smart Building Energy Management: Taking advantage of the LiveLabs experimental testbed deployed on the SMU campus, we are developing new ways to improve the energy efficiency of smart buildings. A key idea is to use various forms of sensing (e.g., WiFi, cameras) to accurately capture the human occupancy in different parts of a building. We then use machine learning (ML) to predict both the future occupancy of such spaces and the right set of operating parameters (e.g. HVAC settings, LED light intensity) that minimize energy consumption while satisfying comfort-related constraints. A key area of our research is in developing new forms of cheap IoT sensing to provide such fine-grained occupancy sensing.

· Edge Analytics and Distributed Machine Learning: For many future IoT-based applications (such as augmented reality & robotic manufacturing), it is important to compute analytics insights at the network “edge”, on resource-constrained embedded devices. Indeed, as sensor deployments grow in both scale and sophistication, the current approach of “centralized, cloud-based sense-making” (where most analytics happen on raw data that are shipped to a cloud infrastructure) may not always be feasible for multiple reasons: connectivity, privacy and latency. Our work is developing new techniques for such edge-based analytics that rely on (i) the coordinated operation of an ecosystem of heterogeneous IoT devices and (ii) the ability to execute machine-learning algorithms in a distributed fashion.