A Formula for Future Embedded Systems
July 14, 2021
There is beauty in the simplicity of a formula that describes the processes of the physical world. Examples which come to mind are Maxwell’s equations, or the relationship between mass and energy made famous by Einstein.
What if we could design a formula to describe the evolution of cyber-physical systems (CPS)? As part of our work on the HiPEAC Vision, funded by the European Commission, we’ve come up with one: future CPS= 5S.(CPS)2. Let’s break it down.
Cyber-physical systems are becoming more intelligent, actively responsive, and complex. In other words, what we are seeing increasingly are: cognitive cyber predictive physical systems of systems – CCPPSS, or CPS2. (OK, it’s geeky, but it’s elegant, right?)
Going into a little more detail, we see this panning out as follows:
Artificial intelligence (AI) is everywhere, and will help CPS to better “understand” their environment, to have a better perception of their surroundings. To move AI forward, we recommend investing in ultra-low-power accelerators and investigating training approaches which rely less on labelled data.
Europe was a leader in the development of the World Wide Web and hence of the cyber revolution, and we think it can still be at the forefront of the next web, which will add innovations on top of existing technologies. We can also make a mental connection to cybersecurity, which is becoming more and more important,
CPS should model their environments, and make forecasts of various potential future evolutions. They would therefore be able to predict, to some extent, how the environment will react.
Up to now, computing systems haven’t taken into account the constraints of the real world: time is so abstracted that most programming languages don’t even have a way to express it, while energy is often an afterthought, etc. CPS systems should take into account properties of the real world (i.e. non-functional properties) in order to fulfil their mission. We’d like to see more research into modelling how systems interact with the physical world. Thereby, they have to obey constraints such as timing and energy in order to have safer, more secure systems.
Systems of systems
With computing increasingly being carried out over a continuum from the edge to the cloud, developing tools to orchestrate large, dynamic, and heterogeneous systems should be a clear priority.
So what about 5S? Well, these are the non-functional properties we’ve chosen to focus on – which could be summarized as follows:
- Sober: this one’s all about energy efficiency – with billions of connected devices burgeoning over the coming years, battery life needs to be extended and systems need to be as energy efficient as possible. Energy is also an important part in the cost of ownership of servers, and the energy cost of communication should not be neglected either. We’ll need ultra-low power computing platforms covering the complete digital continuum, from deep-edge to cloud and high-performance computing (HPC), as well as tools allowing the assessment and design of systems with explicit power constraints.
- Secure: recent cybersecurity attacks have put security into the spotlight again. We think research should focus on the automated detection of security risks in existing systems, and on ensuring that they are as correct as possible by construction. We’d also like to see research into the means to automatically mitigate or remove security risks, the development of secure hardware, and the development of tools that can produce secure-by-design software and hardware.
- Safe: because no one wants to see the blue screen of death on their robot surgeon, or while their self-driving car is driving at high speed on the highway, systems should be safe by construction and/or should be able to be proven to be safe, with all possible failbacks implemented when possible.
- Straightforward: it goes without saying that future cyber-physical systems of systems will be extremely complex. How can we tame this hydra? We propose maximizing productivity to design and manage complex systems, including the use of artificial intelligence techniques. Humans should still be in the loop, but with the right level of abstraction so that they can be as efficient as possible and trust the systems.
- Sustainable: State-of-the-art electronic devices require the use of around 65 of the 102 elements on periodic table, and if these starting running out, the current chip shortages will be dwarfed in comparison. Add this to the fact that only 15% of computing devices globally are recycled, and the case for reusing components becomes even more compelling. We’d like to see further research to lower the embodied energy of devices, and help extend product lifetimes through upgrading, reuse, and repair, with Europe becoming a leader in the design of sustainable electronics.
If you’re interested in finding out more, check out the HiPEAC Vision – a series of readable articles from some of the leading computing systems researchers in Europe – on our website: https://www.hipeac.net/vision/#/latest/
Or have a look at the HiPEAC Vision two-minute explainer video: https://bit.ly/HiPEACVision2021_video
Marc Duranton is Editor in Chief of HiPEAC Vision. Marc is a veteran computing systems researcher with experience in the industry (Philips, NXP). He brings a holistic, philosophical and often thought-provoking angle to discussions on future directions in embedded computing.