Digital Transformation
cancel
Showing results for 
Search instead for 
Did you mean: 

Cool Intelligent Edge and IoT demos at Discover, London 2016

mikeshaw747

I've just been to Discover 2016 in London. I thought I would do a writeup of the demos I saw in the "Intelligent Edge" and "Data Analytics" areas, because they might give you ideas as to how you might use Edge computers, beacons and high-speed, high-density wifi.

Smart hotel

In a demo of a smart hotel, the hotel notices you are approaching the hotel doors. It automatically checks you in, which avoids the check-in line and the check-in time.

If the hotel has preferences for you (or, has noticed a behaviour pattern from which they can infer preferences), your hotel app can give you the information you need and ask you about your needs for this stay. For example, if you often use room service, it could alert you to the latest room service menu. It might ask you if you want a morning paper (not sure someone who has this level of automation on their phone would read a physical paper, but you never know).

Smart hospital

A hospital found that 33% of patients and visitors had, at some time, been lost in the hospital. Either I've got a poor sense of direction or my local hospital is just badly signposted, but I've never not got lost to some degree in my hospital visits.

Anyway, in the smart hospital, you would be "checked-in" like in the smart hotel example above. And then, beacons around the hospital would guide you to the right department. I guess that you may well have multiple "checking-ins" because a hospital visit often involves visiting more than one department.

During the one of the conference keynotes, the CEO of Plymouth (the UK version) hospital also talked about how they were using this technology to track equipment. He explained that the emergency department's trolleys were "going for walks" and disappearing into other departments - actually, into one specific department the identify of which he wasn't prepared to divulge.

"Blue dot" navigation around a large hospital"Blue dot" navigation around a large hospital

Smart retail

"Click and collect" is great and it's part of retail's much vaunted "omni-channel retailing" - you can order from anywhere and collect anywhere. Research by a large retailer in the UK, however, found that the average wait time for someone collecting was 18 minutes. That delay starts to negate the wonders of "click and collect".

And so they created a pilot where you were detected approaching the store. This would then alert to pickers in the "click and collect" department and your goods were prepared for pickup. They also gave you a QR code (one of those view black and white things that lasers can read). You could buy anything else in the store, have the till operater scan your QR code, and the items would be put with your "click and collect" basket.

Of course, once the retailer knows you're in the store, and using beacons, where you are in the store, they can start to offer you incentives to up-sell and cross-sell you. At Discover, we had a dinner with a number of customers and experts on the subject of Digital Transformation. At my table, we discussed the cost/benefit equation for a customer giving up their geo-position. It's strange - when we browse using Google, we give them tons of information. We seem OK with this, maybe because it's tough to do otherwise. But when it comes our geo-position, we understand that this has real value and we need to be assured that we will get something return. I suspect that retailers will find they need to be careful not to inundate their customers with too many "useful offers".

One useful application that a number of supermarket retailers are working on is the "smart shopping list". You create a shopping list and when you get to the store, a beacon-based system will guide you to the items on your list, in an optimized way. The retailers will offer suggestions of "pairings" - "we notice you are buying chicken. There is a nice Sancerre on offer that pairs will with chicken". I guess the issue is that if the retailer helps you too much, you'll zoom around the store without stopping for impulse buys. I read today the a store in the UK is trialing a pattern of lines on the floor of their isles. If you move too quickly down the isle, the lines go past too fast and your instinct is to slow down.

This discussion around exactly what customers will accept and what they won't really underlies one of the key attributes of application development in a digital world and that is the concept of continuous, experimental innovation. Rather than the 18 month, mega-project of the past, digitally-transforming applications take a lot of steps - a lot of experimental steps. I believe that they must do this for two reasons.

  • Firstly, we the technology is moving so fast, we can't know what technology will be available to us in two year's time - beacons will have improved, data analytics will be better and faster, Edge computers (see below for a discussion on Edge Compute) will have become more powerful but less power consuming, 5G will start to be rolled out, etc.
  • And secondly, we really don't know how customers will react to our new applications - what will it take to get them to give up their geo-position, how many "friendly offers" can we give them before they shut down the app, will the connected car of the future have a ton of apps for us to play with and will we really want to play with them or will they confuse us like a restaurant that effectively asks you build your own meal?

Smart office

Does your office have high-speed, high-density wifi? Mine didn't until recently, and it might seem like a simple thing, but once you have a wifi blanketed office, you can get rid of physical phones and you can allow people to collaborate from their PCs, Mac, tablets or phones - all over the wifi.

OK - let's assume, that our office has high speed, high density wifi.

You then put beacons into meeting rooms. When you walk into a room in which you have a meeting booked, you are "checked in" to the room. This could mean that your collaboration call is setup using, say, MS Skype for Business.

That annoying situation where offices are booked but not used can be eliminated - a booked but unused office can be automatically freed up for use.

If you have a lot of hot desking, the system can "know" which hot desks are in use, and should you need one, it can tell you where the nearest one to you or to where a point of interest, is.

Herman Miller were even showing "smart office furniture" that adjusted itself to fit your needs (e.g. changing desk height).

Smart parking

It starts with a flat round, black pancake of about 12cm diameter. This is a parking sensor - it senses if someone is or isn't parking above it. It's wirelessly connected and the battery in it lasts ten to fifteen years.

Using these sensors, you can know which parking spaces are in use and which are free. And then, as with the hot-desking above, you can create parking apps that suggest the nearest free spaces. With public, ungated, parking spaces this will probably result in a mad rush by everyone with the parking app to one space. However, if the parking is gated, you could book a space and ANPR would allow you into the parking area at the barrier.

It's certainly a problem worth solving. I read today that 20-30% of the traffic congestion in downtown San Francisco is caused by people looking for a parking space.

Smart car

With wind, rain, temperature and traffic sensors along the side of the road there are a whole range of "connected car" applications that could be created. We could sense high winds and warm high sided vehicles in a specific way. So, rather than a general weather forecast warming of high winds, vehicles could get warnings as they approached specific high wind areas.

You could be warned as you go from one country to another. In Europe, car regulations change from country to country. In some countries, for example, you have to have your lights on.

When the connected car connects to other smart systems, the value is more than the sum of the parts. The connected car + smart parking. The connected car + home automation (heating and security).

I suspect that, as with retail, experimentation is going to be required. A car manufacturing might put a lot of effort into a connected car feature only to find that customers find it of little value.

## Augmenting existing manufacturing plant I recall reading that only 7% of manufacturing machinery is connected and that the vast majority of those inter-connections where completely closed and thus, couldn't be controlled or integrated with anything else. These closed systems are known as OT or Operational Technology.

One of the main promises of IoT in manufacturing is to use sensors (and in this, I include cameras) to augment existing manufacturing plant. Manufacturing plant is very expensive and it's unlikely a company would replace it just to get some sensors connected over an open, IT (Information Technology) network.

And so, at Discover, there were a number of demos where IoT was used to bridge between OT and IT.

Example 1 : FlowServe's pumps

FlowServe is a company that provides "fluid motion control systems" which includes, of course, pumps. Large, expensive pumps, which if they fail, cause large, expensive downtimes.

FlowServe has augmented its pumps with sensors - turbulence sensors, vibration sensors and temperature sensors. These sensors are then used to monitor the health of their pumps. This in itself is a big step forward, but there is more.

The data from these sensors is fed into a prediction algorithm that has learnt how to predict when a part in the pump is going to fail. In the demo, air bubbles where artificially introduced into the fluid flow (this has a technical term, but I can't remember what it is). The turbulence sensors notice these bubbles and thence the prediction engine tells us that if these bubbles carry continue, the impeller will fail in ten hours.

IoT and predictive maintenance is a hugely applicable use case. Anywhere you have machinery that, when broken, can cause either an expensive repair and/or expensive downtime, predictive maintenance can be applied.

And there is more. Someone needs to repair the pump. Well firstly, the maintenance engineer has to find the pump. This might not be so easy. Imagine, for example, that the pump is in a huge oil refinery. Once again, beacons (lots of them) and geo-positioning come to the rescue.

So, our engineer has quickly found the pump. It's a model II, but it's a long time since she fixed one of those - she's been doing model III's and IIIb's for the last few years. Not an issue. She points her tablet at the pump and augmented reality super-imposes a 3D model of the pump, and then it shows her how to disassemble, fix and reassemble the pump.

Example 2 : Augmenting the production line

In another demo, two sensors had been added to an "old" production line. The production line took pins and put them into holes.

The first sensor was a current sensor on the fan current. When the fan was in trouble the sensor saw a change in current. The sensor could equally have been an airflow sensor.

The second sensor was a visual sensor that could "see" the color of the pins and the color of the holes. This was used to re-program the production line to put red pins into red holes, blue pins into blue holes and green pins into green holes.

In a third demo, a production line was augmented by a series of "smart cameras". These cameras can programmed to look for different objects. The data from the cameras was then sent back to a data analytics system which is turn figured out where the "pinch points" in the process were. If you fix pinch points, you can improve the speed of the process (until you hit the next pinch point, of course).

Once these smart cameras and the collection of their data is in place, we can start to use "deep learning" systems to understand situations where the product line gets clogged up. Such technology is already in use in IT management. Deep learning systems look at data feeds and learn rules for what causes performance issues - "when the database starts to run out of space, a few minutes later, the checkin process starts to run slowly".

Linked IoT data sources

When we have a number of different IoT data sources (or any data sources, for that matter) we can use data analytics to create insights that are greater than the sum of the parts.

In a robot demo, the "brains" of a robot that carries parts were taken out and instead were provided by an Edge Compute server positioned nearby. Why do this? Firstly, it makes the cost of each robot lower - if you have 20 robots you would normally have compute in the robots for 20 brains. But if you do take out the "brains", you simply have one Edge Compute server. Secondly, the data from the sensors on the robots can now be combined.

In the demo, this data was combined with fixed proximity sensors around the factory and then used to work out the optimal paths for all the robots in the factory.

Secure ATM

The next demo was a regular ATM (automated teller machine) augmented with a series of sensors ..

  • Proximity sensors to reduce power when no-one was near to them
  • Vibration sensors to sense people trying to take them out the wall
  • And a camera with facial recognition that was linked with data from facial recognition systems on other ATMs so that someone going from ATM trying to steal money could be identified

The Technology

There were a lot of beacons at Discover, London, 2016!! As well as a lot of other sensors, as I've described above. The beacons use Bluetooth comms to your phone. They then connect using wifi to a server. They are about 3cm by 3cm and inside is a battery that lasts about 3 years.

And there were a lot of "EdgeLine 1000" servers. These sensors, including cameras, generate a lot of data. If the data has to go "back to the data center" for analysis, we have two potential problems. Firstly, it uses up a lot of bandwidth. Imagine the quantity of data from, say, 20 4K cameras around the parameter of an airfield. And secondly, the round trip can introduce a delay, again because of the sheer quantity of data being transmitted.

And so, many of these IoT-type scenarios need what is known as Edge Compute - the ability to do computing, but especially data analytics, locally.

By way of illustration of Edge Compute another demo had an HD camera and three screens. The first screen showed the real-time data feed from the camera. The second showed the data from a facial recognition system running on an Edge Compute server near to the camera. And the third showed the data from a facial recognition system running in the US (i.e. all the data had to go from London to the US). When you stuck your face in front of the camera, you instantly saw it on the first screen, followed pretty quickly by your face with a red box around it from the local, Edge Compute server. But the third screen, the one doing recognition back in the US, took many seconds to show your face.

Edge Compute is going to be important in a world of IoT. IDC estimates that, in the five years (by 2022), about 40% of all IoT computation will be done "at the Edge".

Mike Shaw
Director Strategic Marketing

twitter.gif@mike_j_shaw
linkedin.gifMike Shaw

0 Kudos
About the Author

mikeshaw747

Mike has been with HPE for 30 years. Half of that time was in research and development, mainly as an architect. The other 15 years has been spent in product management, product marketing, and now, strategic marketing. .

Labels
Events
28-30 November
Madrid, Spain
Discover 2017 Madrid
Join us for Hewlett Packard Enterprise Discover 2017 Madrid, taking place 28-30 November at the Feria de Madrid Convention Center
Read more
See posts for dates
Online
HPE Webinars - 2017
Find out about this year's live broadcasts and on-demand webinars.
Read more
View all