Crowd-sourced data from built-in sensors in cars are combined into a map that includes the location of potholes, cracks and bumps.
For the past 16 months, 20 of the city of Detroit’s vehicles have been using Israeli software to automatically map hazards on the city’s roads.
Tactile Processor, developed by Tactile Mobility of Haifa, Israel, gathers data from each car’s built-in sensors and sends it to the company’s Tactile Cloud. Then, the Cloud uses a mathematical model called SurfaceDNA to combine the crowd-sourced data into a map, including each road’s grade and normalized grip level, and the locations of potholes, cracks and bumps.
Funded by PlanetM, this proof of concept is a collaboration between the company, the city, and a major local automaker that had asked the other partners not to name it.
The company was established in early 2012, under the name MobiWize, by Boaz Mizrachi and Yossi Shiri, who were later joined by their friend Alex Ackerman. While the last two have left years ago, Mizrachi remains the company’s chief technology officer.
In 2010, as a lecturer at the Technion in Haifa and an experienced entrepreneur, Mizrachi was presented with the idea to develop an app that would alert drivers if they were going too fast while nearing a stop sign. However, his plans were bigger.
“Gas was becoming very expensive, and everyone was trying to save it,” Mizrachi said. “I was trying to solve the problem of how to get a vehicle from point A to point B with a minimum of gas.”
“I knew nothing about cars or mapping, but I knew that I had to know things about the road, like grades and curvatures; and about the vehicle, like its capabilities and its weight.”
At some point, Mizrachi had realized that most of that data had not existed.
“I was sure that Google had a map that showed all of the grades on Earth, but it didn’t,” he said.
“Also, if you wanted to show a 5% saving in gas, you’d need to measure it with the mean error being 1% at most. However, a vehicle’s sensor’s mean error is up to 30%, as we had shown.”
The company started as part of a business incubator and developed an aftermarket system to be installed inside of a vehicle and instruct the driver on how to save gas based on the characteristics of the road and the vehicle.
“It would sometimes take it a week or two to calibrate itself for the specific car and roadmap, but eventually it would be super-accurate, like nowhere else in the world,” Mizrachi said.
About one year later, it was a Detroit company’s scout who challenged Mizrachi to create a cloud-based app that would control the car’s speed and gear-shifting according to where the car was heading. Mizrachi worked on the software for a few months, and then went to Detroit and connected his system to a car.
“The car would be driven once or twice, self-learn the map of road grades and what the road looked like, and save it to the cloud,” Mizrachi said. “The next vehicle to drive there would know what’s ahead of it.”
Within two weeks, Mizrachi managed to show a 5% saving in gas relatively to “standard” cruise control. “We had become known within the company, and I returned home with some money for that project,” he said.
Switch in Focus
In 2014 Mizrachi’s company recruited three private investors from Israel, Spain and Mexico, and got out of the incubator. He hired between 15 and 20 of his best students and got to work. “We created a beautiful technology,” he said. “We demonstrated the ability to estimate a truck’s weight with a 2% mean error.
“We started working with automakers, understanding their needs, and showing them that the same building blocks can create interesting insights.”
The twist in the company’s story came shortly afterward, when gas prices tanked. “Nobody cared about saving gas anymore,” Mizrachi said. “On the other hand, it turned out that all of the infrastructures that we had created were worth a lot, for they allowed a car to do other things, especially if that car was autonomous.”
While other companies trying to develop autonomous driving rely mostly on visual sensors, Mizrachi had realized that pictures do not show the full picture.
“Not everything can be seen,” Mizrachi said. “Some things you need to feel. Imagine yourself remotely driving a car using cameras and a steering wheel in your office. That is not the same.”
The company has created a language in which every pothole or bump is a different word. Those words are created using two mathematical models: While SurfaceDNA is a “digital twin” to the road, VehicleDNA is the same to the vehicle, creating real-time “virtual sensors” of the vehicle’s weight, fuel consumption, tire health and more.
Mizrachi said that the data that his company generated was more valuable than alternative solutions:
“If a driver turned on the wipers, you could guess it rained. If I had connected millions of cars to a cloud, I’d have a map of where it rained that day around the globe. However, just because it rained, and the road is wet doesn’t mean it’s slippery.
“You could drive the Cross-Israel Highway at 95 mph in pouring rain and everything would be fine,” he continued, “but in some places in Israel, even at 40 or 50 mph, very little rain can send you flying. It depends on the road’s texture, how much water has accumulated on it, the kind of asphalt on the road and your own tires.”
Automakers Step In
The company had been working with U.S. automakers for a few years, and then the Europeans joined in.
“Four years ago,” Mizrachi said, “a BMW software developer called me from Munich and said, ‘Imagine that in 2021 a car will be able to self-drive at 160 mph with the driver sleeping.’ I said, ‘I wish the driver luck. Why do you think I can help?’ He said that an essential condition for that would be knowing the grip level between the car and the road 300 yards ahead.”
Other German automakers quickly realized the potential of the unusual partner and also how unusual it was.
“We do things that are completely different than what the industry does,” Mizrachi said. “We are not car engineers. The first time that I arrived in Germany to install the system on a Porsche vehicle, they opened the engine cover and I asked, ‘Where has the engine gone?’ They told me that in Porsche cars, the engine is at the rear.”
In 2018, the company hired a vice president of business development, Eitan Grosbard, who was previously responsible for mergers and acquisitions at Gett.
“I had known the automobile industry, but from a different angle,” Grosbard said. “I have entered a whole new world of software and data monetization.”
In 2019, the company has gained two major investors: Porsche and Union Motors, Toyota’s importer into Israel.
Detroit’s Role
In October of that year, Detroit had become one of several cities around the world to collaborate with Tactile Mobility on a proof of concept.
“The automaker was the one that has put this whole story in motion,” Grosbard said. “Its people said that they were already selling cars to municipalities, but they wanted to start selling service as well.
“It was the first time that I wasn’t working with a city alone,” said Grosbard, who added that a similar type of collaboration is already taking place in Germany, with unnamed partners.
In recent years, the city would rely on citizens using the SeeClickFix app, and a team of people actively searching, to know about road deficiencies. The project’s manager on the city’s side, Samuel Krassenstein, said that Tactile Mobility’s data had changed things.
“Traditionally, when we’d do a road analysis, it was based on a visual assessment of the road, but that rating system isn’t necessarily aligned with where drivers are facing the most pain,” he said. “This type of analysis is based on ride quality. The combination of them ends up working very well.”
Krassenstein said that the project was made unique by the passive form of data collection.
“We did not tell our staff in our vehicles where to drive. We just let them go on their daily business. We’re pretty happy with the coverage of the city we got just on 20 vehicles.”
While the city currently receives raw-data monthly reports, Krassenstein expressed hope that in the future, the data would be more “real-time” and “consumer-ready.”
Grosbard said that a relevant system would be ready by Q3 2021.
Mizrachi said that even for cities with limited budgets, the system would be a worthwhile investment:
“When we installed a similar system in our hometown Haifa, they said, ‘We know that our roads are completely ruined. Instead of paying you, we could fix a couple of potholes that we know about.’ We explained to them that some roads are bad enough to make a car crash, even because of a small pothole — if its edge is steep enough, or if it’s exactly on the tire track. Especially if you have tons of potholes, you should know which would cause the most damage.”
Last year, both Porsche and BMW announced that they would use Tactile Mobility’s technology in their cars.
Last December, two more investors have joined in: Nexteer, a company that provides solutions to automakers in the steering department, and The Group Ventures, a Canadian venture capital fund for special AI-based technologies.
“We are preparing for another investment round this year,” Grosbard said.
Today, the company has almost 30 workers. The interim CEO for the past few months has been Rani Plaut, a well-known entrepreneur, and a board member and an adviser to the company for several years.
Probable Competitors
Both Grosbard and Mizrachi acknowledged that their system’s success may provoke a stiff competition but showed optimism.
“Competition only strengthens the notion that the product is needed,” Grosbard said. “That way, potential clients will also know that they aren’t dependent on us alone.”
Mizrachi said that the quality of Tactile Mobility’s engineering could give it the edge:
“Many machine-learning companies try to get into vehicles, and usually they rely on super-heavy resources,” he said. “NVidia is trying to get its heavy processors into vehicles. If you count on a very heavy processor in the car, in terms of price, power draw, size, etc., the number of cars that you’ll get into is small.
“However, if you aim your technology toward very lean and cheap computers — if you have reliable machine learning in 200 kilobytes — then you win. That is the difference between a company that sells millions of units and a company whose product is sold alongside a couple of luxury vehicles in Abu Dhabi.”
Grosbard said that data such as tire health, for example, had many possibilities for future monetization:
“Tire manufacturers can use this information for R&D — how long it takes their tires to wear out, what different geographical areas can indicate, the kind of drivers that use their tires and so on — and for marketing and sales.
“Vehicle fleets can use the same data. Tires are the second-highest operational expense for motor vehicle fleets, after gas. They’ll be able to receive an indication about tire health in all of their vehicles.
“As an insurance company, if I knew your tire condition, I could tell you that if you don’t replace your tires, your risk profile will be higher.
“Lastly, if the automaker knows that your tires are worn out, it can draw you into its repair shop.”
Mizrachi was looking a few years into the future:
“Our vision is ‘Tactile Inside.’ We want our software to be embedded in as many cars of possible out of the manufacturing strip — at least 10 million cars. As soon as that happens, we’ll have a map of the world’s roads with very high quality and refresh rate.”
“We’ll be happy to become a supplier of innovative information to all of the municipalities in the U.S.”
If you know of any interesting people from Israel with connections to Detroit, send an email to jheadapohl@thejewishnews.com.