About the event
Student: Nuzhat Yamin
Advisor: Ganapati Bhat
Degree: Electrical and Computer Engineering, Ph.D.
Dissertation Title: Uncertainty-Aware Energy Management of Internet of Things (IoT) Devices: Novel Models and Algorithms
Abstract: Low-power internet of things (IoT) devices have the potential to transform multiple fields, including healthcare, environmental monitoring, and digital agriculture. However, the operating life of these devices is severely constrained by their small batteries that require frequent recharging. Harvesting energy from ambient sources has emerged as an effective approach to prolong the lifetime of these devices. Consequently, energy harvesting necessitates the development of energy management approaches to manage the harvested energy effectively. The harvested energy must be carefully managed to ensure sufficient energy is available when ambient energy is scarce. Prediction of the energy available in the future can aid energy management algorithms in making better decisions about the allocation of the available energy. Once the harvested energy is predicted, the predictions must be used in an energy management algorithm to maximize the operating lifetime and application quality of service. However, this is a challenging problem due to two key reasons: 1) In a multi-sensor and harvester scenario, energy harvesting approaches typically assume that the placement of the energy harvesting device and sensors required for health monitoring are the same. However, this assumption does not hold for several real-world applications. For example, motion energy harvesting using piezoelectric sensors is limited to the knees and elbows. In contrast, a sensor for heart rate monitoring must be placed on the chest for optimal performance. 2) Using dynamic optimization methods to obtain the energy consumption in each decision interval is not suitable due to the high overhead of continuously executing dynamic optimization on low-power IoT devices. This dissertation proposal aims to address the above challenges by making the following contributions: 1) We provide a low-overhead algorithm to accurately provide future energy in outdoor solar energy harvesting as well as on-body energy harvesting using light, and motion. 2) We present an efficient algorithm to dynamically transfer and manage the energy in a multi-sensor, multi- harvester scenario. 3) Finally, we propose an imitation learning methodology to directly provide energy management decisions in IoT devices while eliminating the need to perform dynamic optimization, thus lowering the complexity of the energy management. We perform experiments on real-world energy and activity datasets to demonstrate the performance of our energy prediction and management algorithms. Building on this preliminary work, the dissertation proposes to perform work in the following directions:
• Ambient energy sources are highly stochastic in nature. Therefore, our first research direction will explore low-overhead and efficient approaches to accurately predict the uncertainty in future energy. Specifically, we will investigate mean-variance estimation, lower-upper bound, and conformal prediction methods.
• Energy management algorithms must also account for the uncertainty in future
energy to make effective decisions. Therefore, the dissertation will develop low-overhead algorithms to perform energy management while accounting for uncertainty in future energy.