The Mobile, Pervasive and Sensor Computing Group's research focus mainly on the intelligent decision making process from multi-modal information sources such as wireless sensor networks, mobile phones etc. Some information on the ongoing research projects can be found on this page. Please see our publication page and individual project page for more details.

Project Title: An Energy-Efficient Quality Adaptive Framework for Multi-Modal Sensor Context Recognition
Description: Energy-efficient determination of an individual's context (both physiological and activity) is an important technical challenge for pervasive computing environments. Given the expected availability of multiple sensors, context determination may be viewed as an estimation problem over multiple sensor data streams. In this project, we are developing a formal and practically applicable model to capture the tradeoff between the accuracy of context estimation and the communication overheads of sensing. In particular, we are proposing the use of tolerance ranges to reduce an individual sensor's reporting frequency, while ensuring acceptable accuracy of the derived context. In our vision, applications specify their minimally acceptable value for a Quality-of Inference (QoINF) metric. We are working on an optimization technique allowing a Context Service to compute both the best set of sensors, and their associated tolerance values, that satisfy the QoINF target at minimum communication cost. This approach has been validated using SunSPOT and Shimmer sensors testbed. Please see our publication list for more details on this project.

Project Title: Smart Plug-based Appliance Energy Profiling and Prediction Portal for Green Buildings
Description: New emerging “smart plugs” embed a micro-controller and low-power communication device that allows to monitor the power consumption of individual devices (e.g., microwave, coffee machine, laptop etc) plugged into the power sockets, and communicate such power consumption information over a wireless network to a central monitoring station. Such devices could lead to substantial savings of energy and money by enabling Internet-based monitoring and real-time control of the behavior of individual appliances. This project will use real-life microcontroller kits (e.g., ACME Plugs from Moteware) and real-life building measurement data to explore whether such measurement-based monitoring can be used to

1. Profile individual devices— using NILM data analytics algorithm on the time-series of power consumption traces to infer the type of plugged-in device (e.g., distinguish between a laptop and a coffeemaker), thereby building a dynamic catalog of the types and number of devices connected by a consumer.
2. Predict the power consumption of individual rooms— using the past history of the power consumption of individual devices to create predictive inferences of the usage patterns for individual devices (e.g., learn that the individual switches on a dehumidifier for ~3 hrs every Sunday). Besides investigating analytics algorithms for such power profiling and prediction, the project can integrate this into a Web-based portal that visualizes these analytic insights.

Project Title: Improving the Physiotherapy Treatment with Fine Grained Information and Tele-consultation
Description: Physiotherapy treatment includes physical examination of joints’ ranges of motion, muscle length, and muscle power. Given assessments of needs, physiotherapists advise caregivers and family members on appropriate exercise regimens, however there is no way to ensure that these regimens are correctly followed or even carried out on a regular basis. Current practice requires frequent manual examinations, coupled with expensive equipment for assessment and therapy. This approach requires the physical proximity and dedicated attention of a highly trained physiotherapist, potentially making insufficient use of his skills and expertise. Furthermore it necessitates transportation of the patient (who is often elderly) from his home or nursing home to the hospital for each visit or for the therapist to perform services in the home. Given the increasing availability of pervasive computing technologies, patients implicitly find themselves immersed in environments capable of supporting complicated physiotherapy routines in a transparent and natural way that encourages the patient’s interaction with both the physical and digital environments. Briefly, our vision of pervasive computing assisted physiotherapy is one in which a patient can follow a prescribed exercise and physiotherapy regimen at anytime and in any place while being monitored and guided by digitally augmented physical objects embedded in their natural spaces. Therapists can be provided detailed dynamic and adaptive regimens and can monitor their patients’ progresses and capabilities at a very fine grain too. In order to realize this, a holistic physical therapy management system for tele-rehabilitation needs to be designed and developed by leveraging on multimodal sensor networks, intelligent motion analysis, collaborative user interactions, centralized management and communication networks. In this project we will present an early attempt of a systematic design as well as an initial implementation and evaluation of the proposed remote physical therapy management system. Our plan is to develop a prototype system which can give the freedom to the elderly in this ageing society to do their normal physiotherapy exercise from their home at their convenience. We believe in this highly connected society this innovation may add a tremendous value to the healthcare industry.

Project Title: An Activity Recognition Framework for Multiple People in Smart Home Environment
Description: Coming soon! Check out my PhD dissertation for some early work on this topic.

Project Title: Mobile Phone and Sensor based Transfer Learning Model for Activity Recognition
Description: Coming soon!

Project Title: Supporting Multi-Fidelity-Aware Concurrent Applications in Dynamic Sensor Networks
Description: Most existing research in wireless sensor networks focuses on optimally running a single application on top of a tailor-made and deployed network. However, as sensors become an integral part of our environments, we posit that sensor networks will be increasingly viewed as platforms that will be used to run several user applications simultaneously. Concurrently executing applications requires the network’s resources to be shared across applications so as to make the best long-term utilization of the constrained devices and communications network. This resource sharing entails tradeoffs in the fidelity offered to individual applications. Fidelity is an application-dependent concept that can denote a variety of operational measures including communication latency, data quality, and redundancy. In this project, we are investigating a principled way to define the fidelity associated with an application given a particular allocation of the available resources. Given a desired set of applications to deploy, we are also exploring the impact of sharing resources among the applications in terms of fidelity degradation to individual applications. Please see our publication list for more details.