IBM Watson’s next mission? Keeping antelopes toasty warm

As the owner of a mini “zoo” comprising of two cats, and whatever assorted fleas currently sublet their fur, I can tell you it’s expensive. Scale that up to a full-size zoo, comprising of big cats, monkeys, rhinos and so on, and the problem increases exponentially.

Heating alone is a huge issue: if you lift a bunch of animals from sunny climes, where they’re used to constant sunshine, and drop them in England, they’re going to need a substantial heating solution or a selection of custom-made woolly jumpers. Either is pricey.

I was invited down to Marwell Zoo near Winchester recently to check out a pilot scheme looking to cut such costs, without letting the animals freeze. It’s a project made possible by fellow-Hampshire dwellers IBM which is bringing Internet of Things thinking to Marwell Zoo, one enclosure at a time.

Marwell’s seven nyala – a type of antelope – are the early beneficiaries. The reason for this is twofold: firstly, their enclosure is easily accessible. Secondly, they’re hardy enough not to really care if something goes wrong. “They’d probably be fine – they’re not like our okapi, who if it dropped below 15 [degrees] would panic,” hoofstock animal keeper Claire Curtis tells me. “[Nyalas are] happy if it’s a little bit chilly and happy if it’s hot. They’re quite laid back animals.”

Actually, even the okapi wouldn’t notice, yet. The system isn’t quite active, still in a testing phase where IBM’s cloud-based Watson AI is proving itself responsible enough to take care of several exotic pets. However, this is how the system will work: an infrared sensor sits in the rafters of the nyala house. It takes 16 temperature readings per second and sends the data to a Raspberry Pi, which passes it up to IBM Watson. Watson decides whether or not the readings suggest an animal is present. If they are, the heating comes on. If they’re not, it stays off.

“In the algorithm, there are various thresholds we had to calibrate,” IBM’s software developer Thomas Petty tells me. “For starters, when we pick up which data points are significantly hot, we had to make a decision on how many hot data points means we should turn the heater on, and how far above the average is significantly hot.

“To begin we just had a threshold that said if it’s two degrees above, that’s hot. Now we have a different thing where we measure degrees of heat above the average – if it’s one point, but it’s five degrees above average, there’s definitely some kind of animal there.”

The cash cost of conservation

So, it’s like Nest for antelope. So what? Well, that kind of indifference suggests you’ve never had to run a zoo, and almost certainly haven’t seen the kind of financial spreadsheets they have to deal with. “We think heating is responsible for about a third of our total electricity consumption,” Duncan East, Marwell Zoo’s head of sustainability tells me, adding that the electricity consumption triples from 100kw in summer to 300kw in winter.

“From the test data, the savings should be 30-40%,” he explains. “It won’t be quite that much because there’s a precautionary element to prevent damage to the bulb, but we should be saving around 30%. That’s something like £100 per heater, which across 50 or 60 heaters adds up to a reasonable amount of money and electricity.” Current estimates suggest Marwell could be looking at a saving of £5,000 to £10,000 per year.

We’ll have a better idea in February when Watson gets direct control of the heating. Currently, it’s just taking a photo so IBM can check the computer’s working. “I’ve got a big collection of photos, one pile with pictures of animals when we’d turned it on, and one with no animals when we turned it off,” explains Andy Stanford-Clark, chief technology officer for IBM UK and Ireland. In the three months Watson has been moonlighting as a wildlife photographer its accuracy has increased from 85% to 95%.

That’s certainly good enough for the nyala, but what about more sensitive species? What happens if the Wi-Fi drops? “If the Wi-Fi fails, or any part of it doesn’t work, the heater will fail on,” Petty tells me. “The Nyalas wouldn’t be in danger of freezing. And the zoo will be able to override it too.”

From coffee to antelope

Despite the obvious fit, the project didn’t begin with the aim of monitoring antelope. It actually started life in Germany – an IBM pet project (excuse the pun) to streamline office life. “I was doing a project in one of our IBM offices in Munich to monitor the coffee queue to see if it was busy or not,” Stanford-Clark recalls. A sensor, like the one now in Marwell’s nyala house, was mounted above the machine. “This was reflected on an LED display so you could glance at it and see how busy the coffee bar was, and maybe defer and go down later when it’s less busy.”

How did word spread to Marwell? Well, it turns out that East was mulling on a zoo-shaped problem when he saw Stanford-Clark talk at the Hampshire Chamber of Commerce back in 2017 – only, it wasn’t about heating then. It was about seeing whether people actually read those information signs at the zoo, and which were the most effective.

“I was looking at thermal imaging as a way of counting how many people are going past a certain point that wasn’t intrusive, like using an actual camera which is a bit Big Brothery,” East explains. “Obviously with thermal imaging you don’t recognise any individual, you just see bodies.”

But hearing Stanford-Clark talk, his thoughts quickly turned to the zoo’s monster heating bills. The foot-flow project is currently on hold – “this one took precedent,” East explains, for obvious reasons. It could well be returned to, of course. In fact, it could manifest itself in a number of ways.

“Armed with the knowledge of whether animals are in their house or not, you might have an app that will tell you ‘the tapirs are in bed’ or ‘the lions are out at the moment,’” Stanford-Clark muses. “You might plan your route around the zoo based on where the animals are at that time.”

Petty even thinks the current model could extend into basic animal healthcare. “If you find a really, really hot point – like ten or 15 degrees higher than you expect their body temperature should be – maybe we could send for a medic to check if they have a fever or something,” he suggests. “But that really is just a concept at the moment.”

Will it work?

We don’t even know for sure that the trial will successfully clear the next hurdle yet, though the signs are good. “The interest from the vets is whether [the scheme] changes the behaviour of the animals at all,” Stanford-Clark explains. “They might say ‘oh we’re not going there because there’s no heat’. The theory at the moment is that they will go there because they’re creatures of habit.”

If everything goes to plan, Marwell is hoping to expand it to other animal houses, all accessible remotely, beyond the initial setup. Petty says he’s only been down to the zoo once in three months and reckons it should be reasonably straightforward to adapt for other animal houses. “It shouldn’t take much too tweaking to move it over,” he says. Fellow developer Alexander Hurrell elaborates: “Of course there will need to be calibrations for different enclosures, different sizes, larger rooms, some may be warmer than others. But the tech and code wouldn’t have to change much.”

Other zoos are watching with interest. “The zoo community is quite a close one,” says Stanford-Clark. “Clearly this is a huge cost and carbon issue for all zoos, and we’d be very keen to roll it out to other zoos assuming it’s successful.”

It’s potentially one of those rarest situations where everybody wins: the zoo reduces its costs, the planet gets a little less carbon in the atmosphere, Watson learns some more life lessons and IBM makes a bit more money.

And the nyala? They don’t care, as long as they stay toasty.

Images: Marwell Zoo and Alan Martin