According to a 2018 Harvard Law School report, more than 80 percent of Americans admit to throwing away food before it has expired – just in case it makes them ill. They aren’t alone: most people in wealthier nations tend to be extremely cautious about food safety and it’s a major contributing factor to a third of all food produced each year being wasted.
Many people revert to their senses, using a ‘sniff test’ but this is no more reliable than a ‘use by’ date as food that smells fine to us can still contain enough food poisoning bacteria to make you very sick. As engineers, we spend our lives searching for answers to human challenges that technology can solve and now it’s the turn of food safety to get an intelligence make-over.
That make-over comes in the form of a highly-sensitive computer ‘e-nose’ being developed by a group led by my team in Arm’s research division. The so-called PlasticArmPit project aims to develop tiny smell sensors on plastic chips that can be in embedded anywhere – from food packaging to the armpit of a T-shirt. The aim is to create a sensor so sensitive and accurate that it can signal when as odor is signaling a problem.
Putting a team together
The project is supported by Unilever, which is looking to develop follow-on consumer applications, PragmatIC, a world leader in ultra-low-cost flexible electronics, the University of Manchester with its expertise in printed Organic Field-effect Transistors (OFET) and UK funding body Innovate UK.
The team has now produced a PlasticArmpit prototype in the form of a system-on-chip (SoC) on a flexible clothing insert, capable of monitoring wearers’ odor levels. Should an unexpected run for the bus or particularly tense business meeting result in elevated body odor, this machine learning (ML) powered wingman could politely encourage the wearer to take a shower – or at least temporarily mask the smell with a deodorant veneer. This is the first step and once proven the application could be taken in any direction, food being another prime candidate.
To make this possible, the team had to design a super low-cost microcontroller, so cheap it could be embedded in billions of low-cost products. It needed to have a small enough form factor so it could be embedded into any garment, container or wrapping, and be resistant to environment challenges (such as washing). To achieve all of these things, we looked past expensive silicon-based chips and started experimenting with plastic chips and specialist flexible electronics fabrication processes.
Plastic chips offer flexibility
Plastic chips have their trade-offs; silicon chips fit billions of transistors, while the largest published designs for plastic run to just thousands. For silicon designers, this is like time-travelling back to the 1970s – so designs needed to make the most of every transistor. The only way to do this, currently, is to make chips non-programmable, with a common base design customized to suit each specific purpose.
However, our design approach now is to give these plastic chips the ability to learn from running neural networks (NNs). We think this will be the key to accelerate development of low-cost flexible, integrated smart systems that are customized for a specific application. Capable of operating in extremely parallel fashion to achieve high performance and consume low power, ProjectArmPit will be the first application of artificial intelligence (AI) algorithms in a flexible smart device.
Reducing waste, increasing recycling
Coming back to potential applications, it is easy to see how plastic chips may be embedded into packaging around meat or fish, which could become dangerous rather than just unpleasant to consume when past its best. A sensor that reliably detects the presence or concentration of a particular odor or chemical indicating toxicity in the protective packaging atmosphere, displayed in an easy-to-understand format, would go a fair way to giving the consumer confidence to consume products which may have been otherwise wasted had they relied upon a ‘use by’ date. Wide adoption of plastic chips in packaging could even pave the way for smart recycling centers, enabling automatic sorting of various plastics and reduced contamination.
We’re still in the early stages of the project, but we’ve already established the fundamental building blocks required to support our vision for PlasticArmPit’s applications. What began with body odor is now growing rapidly as we discover ways to combine the new fields of ML and flexible integrated circuits. While reducing food waste is of utmost priority, we also see this technology playing a role in healthcare; smart bandages containing plastic chips could provide crucial early indicators of infection, helping to give respite to the 46,000 people that die each year from sepsis after injury or minor infection. And as with everything at Arm, we’re putting the design flows in place to ensure that this technology is available and accessible to everyone.
Note: The PlasticArmPit project began in October 2017 and is part-funded by InnovateUK, a British government backed funding group. The project’s plastic elements grew out of the PlasticArm project, itself an exploration of what printing integrated circuits onto flexible plastic chips will enable us to achieve that silicon cannot.