Abu Bakar

Protean: An Energy-Efficient and Heterogeneous Platform for Adaptive and Hardware-Accelerated Battery-free Computing

Abu Bakar, Rishabh Goel, Jasper de Winkel, Jason Huang, Saad Ahmed, Bashima Islam, Przemysław Pawełczak, Kasım Sinan Yıldırım, Josiah Hester

ACM Conference on Embedded Networked Sensor Systems (SenSys'22)

Protean is an inference-capable battery-free platform designed for energy-aware run-time scalability. It consists of three main components: 1) SuperSensor, a modular plug-and-play platform that provides hardware support for intermittent computing with a reconfigurable energy storage circuit, a 32-bit ARM-based microcontroller with a convolutional neural network accelerator, etc, 2) Chameleon, an energy-aware adaptive runtime that provides intermittency-proof execution of machine learning tasks across heterogeneous processing elements, and 3)Metamorph, a code generator tool that automates the conversion of ML models to intermittent safe programs.

DOI and video coming soon!

Adaptive Intelligence for Batteryless Sensors Using Software-Accelerated Tsetlin Machines

Abu Bakar, Tousif Rahman, Rishad Shafik, Fahim Kawsar, Alessandro Montanari

ACM Conference on Embedded Networked Sensor Systems (SenSys'22)

In this work, we propose Lite-TM, a framework that addresses the shortcomings of the Tsetlin Machines (TM) and makes them an energy-efficient inference engine for batteryless sensors. We exploit the inherent redundancies observed in the TM architecture and trained TM models with two novel compression techniques, respectively, that reduce the model size by 99%. These techniques significantly reduce the memory footprint and speed up the run-time execution by dynamically scaling the computational complexity based on available energy.

DOI and video coming soon!

Logic-based Intelligence for Batteryless Sensors

Abu Bakar, Tousif Rahman, Alessandro Montanari, Jie Lei, Rishad Shafik, Fahim Kawsar

International Workshop on Mobile Computing Systems and Applications (HotMobile'22)

In this work, we explore a logic-based inference algorithm, Tsetlin Machine (TM), for making batteryless sensors intelligent. Because of constrained memory, energy and compute resources, using TM models as is in real-world applications is not possible. We propose a lossless compression scheme based on run-length encoding and show that it can compress the model by up to 99%. This translates into a lower memory footprint and better energy efficiency (up to 4.9x) compared to the original Tsetlin Machine algorithm, and provides promising trade offs when compared against state-of-the-art binary neural networks.

FaceBit: Smart Face Masks Platform

Alex Curtiss, Blaine Rothrock, Abu Bakar, Nivedita Arora, Jason Huang, Zachary Englhardt, Aaron-Patrick Empedrado, Chixiang Wang, Saad Ahmed, Yang Zhang, Nabil Alshurafa, Josiah Hester

Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies (IMWUT/UbiComp'22)

FaceBit is an unobtrusive, intelligent platform that sticks with masks and collects various physiological signals. It monitors heart rate without skin contact via ballistocardiography, respiration rate via temperature changes, and mask-fit and wear-time from pressure signals, all on-device with an energy-efficient runtime system. It’s basically a FitBit for your face! FaceBit can harvest energy from breathing, motion, or sunlight to supplement its tiny primary cell battery that alone delivers a battery lifetime of 11 days or more.

Fast Company 2022 Innovation by Design Award—Finalist in the Students category

REHASH: A Flexible, Developer Focused, Heuristic Adaptation Platform for Intermittently Powered Computing

Abu Bakar, Alexander G. Ross, Kasım Sinan Yıldırım, Josiah Hester

Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies (IMWUT/UbiComp'21)

REHASH is the first energy-aware adaptive runtime system for battery-free systems. It uses novel lightweight signals stemming from the intermittent execution of programs and combines them in simple heuristic functions to predict energy availability which helps applications dynamically adjust their complexity at run-time. A simulation tool, REHASH-explorer, helps developers choose from the proposed signals and design heuristic functions that best suit their application needs.

SIGMOBILE GetMobile Research Highlight 2022

BFree: Enabling Battery-free Sensor Prototyping with Python

Vito Kortbeek, Abu Bakar, Stefany Cruz, Kasım Sinan Yıldırım, Przemysław Pawełczak, Josiah Hester

Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies (IMWUT/UbiComp'21)

BFree allows the development of battery-free applications, to novices and hobbyists, using the Python (leveraging AdaFruit’s CircuitPython ecosystem) programming language and widely available hobbyist maker platforms. BFree provides energy harvesting hardware and a power failure resilient version of Python, with durable libraries that enable common coding practice and off-the-shelf sensors. This work allows makers to engage with a useful, long-term, and environmentally responsible future of ubiquitous computing.

Time-sensitive Intermittent Computing Meets Legacy Software

Vito Kortbeek, Kasım Sinan Yıldırım, Abu Bakar, Jacob Sorber, Przemysław Pawełczak, Josiah Hester

International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS'20)

TICS is a checkpoint-based runtime system for battery-free intermittently-powered devices that guarantees forward progress, data consistency, and correctness. TICS provides simple programming abstractions for handling the passing of time through intermittent failures, and uses this to make decisions about when data can be used or thrown away. It also provides predictable checkpoint sizes by keeping checkpoint and restore times small, enabling numerous existing embedded applications to run intermittently.