Now that the smoke from the November fires that burned huge swaths of California has begun to clear, talk has turned to how to prevent future disasters. A Silicon Valley startup has an idea.
Team Lali, part of the Hacking House’s first class of would be entrepreneurs, had a solution—a low cost wireless sensor network—looking for a problem. Thanks to an opportune talk by a few firefighters, they found one—fast, cheap, and, they hope, effective wildfire detection.
I met with Sucre Cando, from Guayaquil, Ecuador, and Nassim Bettach, from Lyon, France, in Fremont, Calif., in November. The air was full of smoke from the raging Camp Fire some 320 kilometers north. It was hard to breathe, even a little difficult to see, so it was definitely a good day to talk about fire.
The two came to Silicon Valley as part of the first class of Hacking House. SigFox, a company that markets a low-power wide area network for connecting IoT devices, launched the program in August. For each class, SigFox brings in one or two dozen young engineers and business school students who form teams and select problems to address. Most of the problems, submitted by SigFox clients, are along the lines of, “Is there any way we could do this?” SigFox eventually plans to establish 50 Hacking Houses around the world; the second, in Taipei, will welcome its first class in the first quarter of 2019.
Cando and Bettach arrived in California near the end of August, quickly forming a team of three with Jay Nagdeo, who hails from Nagpur, India. (I wasn’t able to meet with Nagdeo; he’d left the United States for the Netherlands until visa issues could be resolved.) The choice of partners was somewhat random, although Nagdeo knew Cando and Bettach slightly from previous tech events. But the partnership seems to have been blessed by the fates: Cando and Bettach later found out that they were born on the same day two years apart.
Neither had a particular project in mind when they started. They turned to the ideas submitted by SigFox customers and potential customers, and fire detection was on the list. But, says Bettach, “I didn’t want to pick this one at the beginning; I had no idea how we would solve it.”
But later that month, a talk by Fremont, Calif., firefighters at an open hackathon that was part of IBM’s Call for Code program convinced him. “They made me understand the scale of the problem, their difficulties in the field, and how we could help them save communities,” says Bettach.
Cando was already on board. “I tried to solve this problem before,” he said, “because Ecuador is burning all the time.” Cando had previously had reservations about the viability of an early-warning system because it seemed that the price of such a device would have to be too high. “But talking to the firemen, I found that they wouldn’t consider a $20 device expensive here [in the United States]; it would be too expensive for Ecuador.”
Today, fire departments usually find out that a fire has broken out from a 911 call. They also monitor strategically placed remote cameras, and get alerts from automated systems that use satellite data. A new satellite, the GOES-17, just got into position over California. It comes too late to help in tracking the Camp Fire, but it should help responders short circuit future fires, thanks to higher resolution and more frequent scans than was provided by the satellites that preceded it.
Existing approaches, the firemen explained in their presentation, don’t spot all wildfires soon enough. Ideally, they’d like to catch conflagrations before they spread beyond an acre. Detection within the first acre is critical because it typically allows firefighters to get to the scene before a fire spreads to three acres; at that point, containment becomes much more difficult.
With those important bits of information in mind, the trio used the ten-hour hackathon to build a prototype sensor package that monitors temperature and sends data to a SigFox base station every ten minutes. The base station relays the data to cloud-based software for processing. (A few sensor-based systems, like the FireBug system, are already in development, but Bettach thought the SigFox approach would give he and his teammates an edge in both cost and scalability. FireBug uses Zigbee mesh network technology to communicate.)
They decided to use IBM’s Watson system to learn about what constitutes a large temperature jump and to compare data from multiple sensors in order to avoid false alarms. The trio plans to eventually have Watson not only detect the presence of fire, but by monitoring temperature changes over multiple sensors, map its shape and predict its path—functions they expect to be hugely helpful for firefighting.
“We needed to make these as cheap as possible,” Bettach said. “Temperature is a pretty dumb sensor, but it works. If a fire is close by, you’ll see a big sudden increase in temperature and you are sure it is a fire. If [the sensor] is being warmed by the sun, the temperature will increase more slowly.”
He thinks the team has made the dollars make sense. The costs of communication and cloud computing would be minimal. The base stations, which cost a few thousand dollars each, can cover around 150,000 acres, and each sensor package, when manufactured in quantity, could cost as little as $5 (or $10 including installation costs). Meanwhile, a fire, Bettach explained, can cause damage of $7000 or more per acre.
Team Lali Wildfire Detection, as they now call themselves, came in second in that initial Call for Code hackathon, and went on to a third place finish in the Global Call for Code competition in October. They picked up $25,000 in prize money that they will use to continue development. Since then, they used a dataset from France to test the software, and did tests of 42 sensors installed in a small grove near their Fremont, Calif., offices. Next up is a series of larger tests involving some 1000 devices, likely in California and Ecuador.
With their time at Hacking House now over, all three have left Silicon Valley, at least for now. When I talked to them, Cando and Bettach also had visas due to expire. But work on the fire sensor network will continue.
“The development costs abroad are cheaper than here, so it’s OK that we need to go,” says Bettach. “But the go-to market is here in California, so we will be back.”