The 2021 Paul Panayotatos Endowed Scholarship in Sustainable Energy has been awarded to ECE graduate students Tahiya Chowdhury and Murtadha Aldeer.
The Paul Panayotatos Endowed Scholarship was established in memory of Professor Paul Panayotatos who served for 30 years as a professor in the Department of Electrical and Computer Engineering. The Scholarship is awarded to graduate students demonstrating academic excellence and pursuing an advanced degree in the sustainable energy areas including renewable energy, energy efficiency, energy conversion or a related area.
Tahiya is a Ph.D. student working with Prof. Jorge Ortiz and a member of the Cyber-Physical Intelligence Lab. She received an MS in Computer Engineering from Rutgers University and a BS in Electrical Engineering from Bangladesh University of Engineering and Technology. Her current research focuses on developing deep learning-based models to analyze occupants’ behavior in buildings. Using the internet-of-things and machine learning, her work explores time series segmentation, occupancy detection, and activity recognition using limited data for dynamically changing physical environments.
According to the Department of Energy, 39% of total US energy consumption is used for heating, ventilation, air conditioning (HVAC), and lighting in residential and commercial buildings. The ubiquitous sensing ability of the internet-of-things enables occupancy and activity detection applications to sense occupants’ presence and can be used for adaptive building energy optimization. Deep learning -- a machine learning paradigm that involves learning the physical world events by training multi-layer neural networks -- powers such occupant-driven HVAC control systems when given access to large computing resources and massive datasets. However, the physical environment is dynamically changing over time. This increases the amount of data and computational cost required to learn new, evolving data. There is a growing concern over the amount of power required for training deep models in long-term deployment and its carbon footprint on Earth. Light-weight, computationally inexpensive models that require fewer examples can be trained faster, and consume less memory and power. Such energy-efficient models are particularly useful for low-power devices used in applications such as activity recognition and occupancy detection. Tahiya’s research focuses on developing deep learning models that are high-performing for evolving data through adaptive strategies tailored for energy-efficient training. This can also bring down building energy consumption by reducing unnecessary heating, cooling, air conditioning, and lighting, and thus curb its effect on climate change.
Murtadha Aldeer is a Ph.D. student working with Professor Richard Martin and Professor Jorge Ortiz in the Department of Electrical and Computer Engineering. He is a member of WINALB and Prof. Jorge Ortiz' lab – the Cyber-Physical Intelligence lab – where researchers focus on designing new low-energy systems and machine learning techniques on embedded sensors for connected health applications. One of the main application domains is in smart pill bottles for medication adherence monitoring and user identification.
For patients managing chronic diseases, medication non-adherence is one of the leading factors determining outcomes. In many cases, medication adherence can make the difference between life and death. It is estimated that almost 50% of patients with chronic illnesses do not adequately follow their prescribed medication-intake schedules. That translates to almost 125,000 deaths and about 25% of hospitalizations each year. Also, 16-27% hospitalization and death in nursing homes can be attributed to medication errors, with over 75% of patients experiencing medication and inappropriate medication events. Indeed, these issues have driven a recent wave of Cyber-Physical Systems (CPS) research, including commercial development of new pill bottles that monitor when a pill is extracted.
Murtadha's PhD research focuses on low-energy sensing and communication for smart pill bottles. Such pill bottles have an array of embedded sensors and are connected to a network. A smart pill bottle, for example, can help medical practitioners observe how well patients comply with their medication schedules.
Murtadha’s current work has even used sensors to identify who is taking the pill and when, through their interaction with the bottle. The system, is an in-bottle patient discrimination system that is low-energy and unobtrusive. It uses PIP-Tag (a wireless transmitter designed at WINLAB), accelerometers, and load cells. The system, called PatientSense, also uses computationally lightweight machine learning models for feature construction, and patient classification.
Congratulations to Tahiya and Murtadha!