Part of series: SCC Distinguished Seminar Series

Are Sensor Networks a first step towards the Diamond Age?

Lancaster University

~900 words


Last modified: December 18th, 12,016 HE

I don’t think anyone in this room knows what loving grace is. You’re all engineers.

Dr Mark Rouncefield

Prof. McCann opened her instalment of the SCC Distinguished Seminar Series with a brief excerpt from Stephenson’s The Diamond Age: Or, A Young Lady’s Illustrated Primer, which formed the central pillar of her speech. The excerpt talked about two pieces of fictional technology in the world of The Diamond Age; the matter compiler and smart dust. The latter are intelligent, self-organising atoms that float around in the book’s universe until they are sucked in my a matter compiler to create any substance or object the user desires. She asked if, though we were seeing context-aware smart computing, we were anywhere close to smart dust.

She then went through the challenges facing current smart devices. Firstly, they are largely low-resourced and flaky due to a need to be low-powered and untethered. If you’re digging into the ground to put something under it, you don’t want to be digging it up again later to replace the batteries. This tendency to degrade over time is also an issue as many smart device applications are intended to be used long-term.

The issue of how to get these low-powered devices to have a manageable-but-still-useful view of network was achieved by looking at nature, but not stealing from nature in the form of fish shoals and fireflies. The latter look at their next-closest neighbour to sync their flashing with, which before long leads to a whole swarm pulsing in unison. This distributed networking had the advantage of being hugely scalable.

Nodes don’t always need to be running 24/7, however, but how can they be allowed to sleep when they may be called upon to assist the network that they are a part of at any time? The solution was to sync the nodes to only sleep at times when the network expected it — slot synchronisation.

The second issue is that a node cannot eat food — or rather battery power — to regain energy lost through exertion. The answer to this is energy neutral operation, where energy (e.g. solar, wind, vibration, etc.) is harvested to nullify the need for batteries. This can only be used, however, when the system has bottom-line QoS, with the example given of an atmospheric sensing system — a sample can be missed due to upload/download throttling without having too much impact.

The third obstacle is that the systems are bounded by the physical world. For example, a city may have over a billion devices operating within its walls, which means the city will have to be viewed as a whole and Shannon capacity taken into account.

Prof. McCann then demonstrated an interesting opportunistic network she had been developing in order to get around this. In the system, users run an app in the background of their phone to which sensors (in the example given located in Central Park) chirp and send sensor data. When this user walks past another user, the data hops across to their phone until it can find a free Wi-Fi upload point or some other path to its destination. To incentivise users to sacrifice their battery life and run the app, a financial reward was given for each bit of data sucked up and ferried along.

Explaining the final issue — the nature of cyberphysical systems — Prof. McCann explained that noise is not always noise. One can trim sensors when the network itself can be used to sense — in the example given, a simple sensor to monitor whether a machine was running correctly was placed next to it, and the disruption caused by the machine operating used to determine if it was doing so, rather than a more complex sensor inside the machine.

Of course, in the wake of the Mirai botnet, security in the IoT remained a topical issue. Opening with another Diamond Age quote — In an era when everything can be surveiled, all we have left is politeness — Prof. McCann detailed some of the issue that these systems face. Malicious and non-malicious data injections need to be separated from valid, unexpected results. This can usually be done by testing the surrounding nodes to see if they’re witnessing the same thing, again a la the fireflies, but this reliance on neighbours opens up a new attack vector.

In concluding that we are not quite at the level of smart dust just yet, Prof. McCann showed off her most interesting example; a capsule that could be launched into space and then print out hundreds of 3 mm2 sensor satellites. As the requirements may well change between lift-off and distribution, the printing could be remotely programmed from Earth to amend the specification of the satellites.

Bizarrely, Prof. McCann ended her talk with a clip from Adam Curtis’ anti-computer All Watched Over by Machines of Loving Grace, which prompted the best question of the session when Dr Mark Rouncefield asked if Prof. McCann knew what loving grace meant, before declaring that he [doesn’t] think anyone in this room knows what loving grace means and that we were all engineers who were just interested in building things. One can always rely on the Rounce to bring some colour to a technical talk.