Battery-powered technologies are critical to a multitude of processes and workflows in society, from infrastructure and environmental management to personal healthcare and pollution monitoring. But maintaining and replacing batteries requires human involvement, which can be inefficient and costly. Taking batteries out of the equation and harvesting energy from sources in the surrounding environment, such as light or vibrations, offers a promising alternative. There is just one catch: such sources are often uncontrollable and unpredictable. Researchers at Newcastle University say the solution is systems that are resilient to power intermittency. The team has been working with Arm to develop a chip that enables state retention for battery-less power.
We spoke to researchers Domenico Balsamo and Osama Bin Tariq to find out more.
Osama Bin Tariq, Research Associate in Embedded Systems Design, Newcastle University:
“We are working on systems that do not have to use batteries. If we have systems where a battery bank or any battery source needs human intervention, they tend to be expensive in terms of maintenance. What we want is a system where we do not need such interventions. Imagine a sensor embedded in a bridge that has a specific purpose to monitor the health of the infrastructure. The life of the bridge, or any such infrastructure, can be 30 years, 50 years. What we want is to use the energy resources around us to power that system instead.”
Domenico Balsamo, Senior Lecturer (Associate Professor) in Microsystems Design, School of Engineering, Newcastle University:
“The way you harvest energy strongly depends on the type of application and the surrounding environment. For example, we can harvest energy through thermoelectric generators in smart water sensors. They tell us the difference in temperature between water pipes and the environment. You can harvest energy from a difference of just a few degrees, but this process cannot be fully controlled, because we can’t control the temperature difference. So, the amount of energy harvested is small.
Another example: you can harvest vibrations in a stadium, while there is a big event, concert or match. But again, the amount of energy cannot be directly controlled. But if you have an intermittent system, the small amounts of energy produced can still be used to do some useful computations.
The research we’re conducting is creating systems that can be resilient to power intermittency. This is what we need when we harvest energy from these uncontrollable and unpredictable sources.
“Our approach was: how can we enhance the active time of the system? How can we achieve more?”
With energy harvesting systems, the challenge is the type and size of energy storage. There are some situations when this can be controlled, and you can rely on enough energy to power your system. You put your system in low power mode. Accumulate enough energy to do something meaningful, then go to low power mode or sleep mode again. This is great when you have enough energy that you can invest. Unfortunately, this isn’t always the case.
We had to push the boundaries and address this problem. We pushed one step further and we said: how can we enable systems with a zero-power state in which the zero-power becomes part of the system? Not just the low power but zero-power. The systems can just stay there until enough energy is available again.”
Osama Bin Tariq:
“With intermittent computing, we can cut the computation into smaller parts. Whenever we have enough power available, we can make use of that to compute that little bit. If the power isn’t available anymore, we just retain the state of our system. Whether it was sensing on the transmission side or in the compute state. Whatever the state, we retain it. Once enough power is available again, we start from the same point where we left off before.
This is where working with Arm was helpful. We prototype and validate our custom peripheral hardware design using Arm IP (through Arm Academic Access) on an FPGA platform.
Our approach was: how can we enhance the active time of the system? How can we achieve more? We try to rely as little as possible on the software. Each time you add more software, it will consume more and more compute cycles. So, we have an additional peripheral.
This is where working with Arm was helpful. We prototype and validate our custom peripheral hardware design using Arm IP (through Arm Academic Access) on an FPGA platform. The hardware unit keeps track of what has been changed, what was on the memory side. When we don’t have enough power, we can take the compute state of the CPU and store it to the non-volatile memory. All these tasks have been offloaded from the CPU and we can handle it through the hardware that we’ve designed.”
“We would like to extend this work from the core to the system, to the network.”
Domenico Balsamo:
“We are enabling the capability of retaining the computational state without changing the architecture significantly, but by adding components that track changes. The moment we detect that a power failure is going to happen we just intervene. We retain the state that’s on the volatile elements to the non-volatile memories of our system. We do this in a selective manner. We don't do it blindly, copying everything from volatile elements to non-volatile elements. That won't be efficient. What we do is track what has changed during the computation, and therefore do it in an efficient way.
We have worked closely with Arm in the past, on different types of architectures. In that case the overall aim was enabling some form of runtime management for energy efficiency with multicore systems. Now with this research, we’re pushing more towards these intermittently powered systems.
We would like to extend this work from the core to the system, to the network. Working on intermittent computing systems, at a different level of abstraction. Covering all aspects from single code up to the entire framework or the entire infrastructure, is the long-term plan.”
Domenico Balsamo is a Senior Lecturer (Associate Professor) in Microsystems Design at the School of Engineering, Newcastle University. Osama Bin Tariq is a Research Associate in Embedded Systems Design.
Arm offers free access to a wide range of commercially-proven Arm IP, tools, and other resources – to enable you to do your best research work, on your own terms.
Explore Research Enablement