From transient computing to transient systems: overcoming application challenges in energy harvesting sensor systems

Abstract

Energy harvesting is a potential solution to power sensor systems, avoiding periodic battery replacements. Nevertheless, these sources usually incorporate supercapacitors to cope with energy intermittency caused by temporal variation in environmental conditions. Therefore, they do not solve the problem of dealing with the size and cost of energy storage. A relatively new concept termed transient computing, aims to remove big storage and enable systems to operate safely when powered from highly-variable energy harvesting sources. However, certain elementary functions of typical sensor systems, such as working with external peripherals or keeping track of time, represent important challenges in systems that operate transiently.

This thesis highlights the challenges that are fundamental to enable transiently-powered system applications and the proposed approaches to address them. This involves a quantitative evaluation of four state-of-the-art approaches to transient computing. The comparison was performed in a system powered by synthesized signals of different harvester sources. Their performances were used to identify the scenarios where one approach outperforms others, and thus, aid designers to choose the most suitable.

In order to retain the peripheral state in transiently-powered sensor systems, a generic middleware was proposed. This approach allows an application to keep the coherency between the processing unit and the state of external peripherals from one power cycle to another. The proposed middleware was tested in a transient sensor system with multiple peripherals and was able to successfully operate with I2C and SPI protocols, causing a time overhead of only 0.82% during the complete execution of the sensing application.

A novel framework to design transient systems is also presented, including a strategy for keeping track of time. The viability of the framework was proven by designing and implementing a step counter. The experimental validation demonstrated that the step counter is able to calculate step rate, metabolic equivalent and activity time, as well as encrypt and wirelessly transmit data, reducing the required capacitance by up to 60%.

More information

Identifiers

ORCID for Alexander Weddell: orcid.org/0000-0002-6763-5460

Catalogue record

Export record