Superfast mobile networks, lower power requirements, smaller batteries and improvements in sensor capability mean that soon, almost any device you can think of, from your fridge to your coffee cup, has the potential to be smart.
The modern factory has long been the realm of the sensor, which has taken over the functions of human senses to enable the automation of a wide variety of production processes, but now that same automation is moving into our offices, homes, pockets, and even bodies. Sensors look set to fundamentally transform our relationship to a myriad of daily activities and open up business models that were previously unthought of.
Sensors in industry
Sensor technology has been critical to the automation of industrial processes for decades, taking over from human eyes, ears and touch on the production line and allowing robots and other automatic devices to carry out many repeatable functions. The ability to connect those sensors, and harness them to effectively unlimited cloud computing resources, paves the way for industrial facilities that are entirely devoid of humans. As David Hannaby, Product Manager at sensor manufacturer SICK UK says:
“Industry 4.0 brings with it the opportunity for distributed intelligent control, where a greater number of process decisions are made autonomously by devices at the field level. As a result, we not only have new levels of flexibility, remote monitoring and diagnostics, but we can also build in intelligent capability for sensors and systems to monitor, identify and respond automatically to situations on the shop floor in real time, for example, to change parameter settings on a product changeover.”
According to David, there are four key dimensions which need to be in place to make a sensor smart:
- Enhanced sensing – manufacturers are very close to developing sensors that can detect and measure any object, including difficult-to-see transparent, semi-transparent, uneven and highly-reflective objects like glass and plastic.
- Efficient communication – the latest sensors make use of IO-Link, a universal, open communication system that allows two-way communication between the sensor and its controller. This means businesses can now access sensor data in real time and use it in more meaningful ways.
- Diagnostics – connected sensors can provide real time data on their own status (battery life, calibration etc) and on the status of the devices in which they’re embedded. This allows the scheduling of preventative maintenance and for replacement components to be ordered in on a just-in-time basis.
- Smart tasks – smaller, cheaper, more powerful chips are allowing sensors to perform much more of the computing work required for complex tasks in-situ. This not only lowers the computational burden on remote servers but can also mean that the sensor is capable of performing even when the connection drops.
Sensors are breaking out of the factory
Heavy industry has pioneered the use of sensors because, even when they were still relatively expensive, sensors reduced the need for human labour and the attendant costs. But as sensors get cheaper, they’re now moving out of the factories and are set to automate a much wider variety of tasks.
Enabled by 5G networks, and small, inexpensive computers, sensors are going to be able to be built into almost anything.
Enabled by 5G networks, and small, inexpensive computers, sensors are going to be able to be built into almost anything. IBM recently announced a computer with the same power as a 90s desktop which is the size of a grain of salt and which costs under 10 cents to produce. That chip is currently armed with just a photo detector, but there’s no reason it couldn’t be used to power a much wider range of sensors.
Sensors in the smart home
Smart homes are already a reality for lots of us, who use apps to set our preferences for things like room temperature and lighting states, and then let sensors take care of the rest. Where the smart home is about to get really interesting though, is where it overlaps with, and starts to automate, our behaviour as consumers.
For example, a smart kitchen might know what we’re running low on, factor in our travel plans, the weather, who is visiting and their dietary preferences, and then use Ocado to automatically restock the house. This is a major reconfiguration of the shopper-retailer relationship. From a retailer’s perspective, how do you get your product into the selection if it’s not something the consumer has bought already? From a consumer’s perspective, what would attempts to influence my smart-kitchen-mediated purchasing behaviour look like?
Ocado’s strategy isn’t even to be a supermarket – they’re just running a delivery service as training data for the technology they’ve built. Their aim is to sell their logistics capability to other retailers and they’ve already cut thousands of humans out of the process.
Sensors in the smart supply chain
A 2017 survey by Deloitte and MHI found that 43% of companies already use sensor technology. The “always on” supply chain enables pharmaceutical and life sciences companies to monitor the cold chain in real-time and ensure that high value time-and-temperature-sensitive goods stay within safe ranges.
Now imagine a pack of chicken that senses when it has been stored outside of its required temperature range. Maybe it releases a harmless dye onto the chicken rendering it unable to be sold and unlikely to be used, and maybe it also tells the manufacturer the GPS coordinates of where it was when it fell out of the safe temperature range. You could probably deliver that for less than a couple of dollars today. In time, you can see the cost dropping under a dollar. There probably isn’t enough of a margin on a basic supermarket pack of chicken to make this viable, but for premium products, such as organic whole chickens or prime cuts of beef, it might work.
Sensors in the smart workplace
Even with wave after wave of automation sweeping through almost every vertical, people will remain the largest expense and most important investment for businesses, so that even small improvements in people’s productivity and wellbeing can have dramatic impacts.
Beyond simply allowing us to monitor and analyse employee use of the workspace, and optimise climatic conditions and energy use, smart sensors could also be used to detect indicators of stress and conflict (though there are obvious ethical questions raised when we start talking about gathering and processing physiological data), or help keep workers in potentially dangerous workplaces safe.
Sensors in the smart city
Monitoring the occupancy of parking spaces is just the start. A proliferation of low-maintenance smart sensors with battery life measured in years could open up an incredible range of analytic possibilities for city planners, as well as the option of handing over the operation and maintenance of key pieces of infrastructure to intelligent, automated systems.
Sensors in roads could enable dynamic traffic management to ease congestion, smart lighting systems could turn themselves off when no one’s around, sensors in bins could help waste management services plan optimal collection routes.
Sensors on the roads and in the air
To be effective and safe, autonomous vehicles are going to need incredibly sensitive and reliable sensors. The inability of its sensors to detect all objects is at the heart of the controversy surrounding Tesla’s Autopilot system and the accidents it’s been involved in. But, leaving aside autonomous and semi-autonomous vehicles for the moment, new automobiles are packed with an ever-growing number of sensors. They feed data to driver displays or intelligent systems in the vehicle designed to alert the user to maintenance issues, prevent damage or increase safety (and, sometimes, for more nefarious reasons).
Amazon’s famous unmanned drones look set to be equipped with a whole host of sensors that not only help them avoid objects but also interpret human gestures, so they can get packages into the hands of waiting recipients safely, and also realise when they’re being told they’ve got the wrong house.
Sensors in the human
In the rapidly growing area of connected health, whereby patients are encouraged to become more involved in their healthcare through continuous monitoring, allowing for more proactive and preventative intervention, there is an increasing reliance on the use of connected sensors. Recent examples include a sock which measures body temperature to detect inflammation in diabetes patients, a wearable sensor to detect hydration and a smartwatch which monitors blood pressure.
Some pretty incredible sensors are being developed to monitor conditions inside the human body, including a miniature wireless pressure and temperature sensor for use in the brain which dissolves when it’s no longer needed.
The current trend towards the quantification of the self suggests that it’s only a matter of time before the Fitbits and assorted other sensors of physiological data also start to get under the skin, though recent concerns over entrusting corporations with our most personal data may dampen people’s enthusiasm for putting their health data in the cloud.
When it comes to imagining use cases for connected sensors, our creativity seems to be the only limit. There are, however, some technical hurdles to overcome before sensors can achieve their full potential.
Estimates of connected devices range from around 20 billion to 50 billion – that’s excluding smartphones and tablets.
For a start, no one can agree on how many devices there are likely to be. Estimates range from around 20 billion to 50 billion – that’s excluding smartphones and tablets. Supporting them will require widespread coverage of high-speed 5G networks, and as we’ve previously noted, there’s uncertainty over how long that will take to roll out.
Another issue is calibration. Anyone who’s had to embarrass themselves in a public place by waving their smartphone about to re-calibrate the compass will be familiar with the problem: the quality of the data coming from sensors degrades over time, or if the sensor is improperly placed, or if environmental conditions are sub-optimal. While it may not be such a big deal to get an extra bag of rice in your Ocado delivery because your smart kitchen didn’t spot that you had some, the consequences of bad data coming from sensors can be a lot more serious in healthcare, transport and industrial settings. And these kinds of sensors are often much harder to reach for recalibration.
There are ways to get around this: one involves adding lots more sensors (redundancy); another uses machine learning to compensate for calibration errors by comparing different sources of information. The first method relies on the falling cost of sensor production; the second on improved ML techniques and the falling cost of cloud computing.
Sensors: the unseen eyes of the fourth industrial revolution
As we’ve noted, the fourth industrial revolution is a revolution of convergence, bringing together many disparate, already-existing technologies to open up a vast array of new economic possibilities. Sensors are just one part of this, their ubiquity increasing in lock step with advancements in machine learning, wireless connectivity, robotics, distributed databases and human-machine interfaces. Most sensors will be designed to be unremarkable: hidden from view or lacking any clue as to their function. But while they will be easily ignored by their human users, the sensors ignore nothing, as they quietly construct a model of the world to be interpreted and acted upon by non-human minds. As those models become more comprehensive and finely-detailed, the range of tasks which can be safely handed over to machine intelligence will grow ever larger.