Xcelerating Innovation, Honda-Style

Honda’s pragmatic approach to tech-startup partners is delivering promising results.

Xcelerator partner SoundHound helped develop Honda’s Personal Assistant voice-enabled AI that debuts in the 2020 Honda E electric vehicle. (Honda)

For the past five years, Honda Innovations, part of the automaker’s R&D enterprise, has been quietly running a collaborative, open-innovation program at six locations around the world. Known as the Honda Xcelerator, its overall mission is not unlike those of other mobility OEMs looking to engage with tech startups. This equates to funding for rapid prototyping, providing access to a collaborative workspace, and letting them work with mentors within Honda to develop products for at least one of eight focus areas. These areas include AI and robotics, connected vehicles, human-machine interface (HMI), industrial innovation, personal or shared mobility, vehicle data business, in-vehicle apps and energy innovation.

Augmented-ergonomics tech from Skelex (left) and Noonee are being tested in Honda production facilities. (Honda)

But Honda’s approach, at its Xcelerator offices in Israel, Japan, Detroit, China and Mountain View, California, is unique. The 72-year-old mobility company is careful about the startups with which it chooses to work. "Honda is very pragmatic," said Dennis Clark, managing director of strategic venture partnerships at Honda Innovations. "We don't have a speculative venture fund. All the work that we do with the startups, they're pretty much fully vetted. We have a lot of conversations and initial testing” in the early getting-to-know-you phase of the relationship.

Clark said start-ups seeking to engage with Honda first submit an application. Candidates who are selected are asked to follow-up with a demo. If it goes well, Honda and the startup will jointly identify a value creation opportunity and define a rapid prototype to work on. Then upon acceptance into Xcelerator, the startup and Honda collaborate on a proof-of-concept prototype that may then be demonstrated to Honda executives and development teams.

Bringing AI to the Honda E

Monolith AI's software analyzes old Honda vehicle data then applies deep learning to the design of parts, specifically CAD modeling. Three Honda plants are testing the package. (Honda)

The results have been positive, Clark told SAE’s Automotive Engineering. A long list of companies have worked with Xcelerator in some fashion since its founding in 2015. They include Heuro Labs, Ubitricity, WayRay, BRAIQ, Haas Alert, Tome and Nextremer. Some partnerships go further than others. Drivemode, a software developer that builds apps for drivers, entered the Xcelerator program in 2015 only to be acquired by Honda in 2019.

The technology resulted in an app that will debut later this year. It will allow riders of some Honda two-wheelers to get information from their bikes on their smartphone. SoundHound, another Xcelerator veteran, is the source of the technology Honda uses for its Personal Assistant, the voice-enabled AI that will debut in the all-new 2020 Honda E electric vehicle. Honda and SoundHound announced their partnership in 2018. This and other partnerships reflect where outside investment money has been flowing, Clark said.

"For the last five or six years, venture capitalists have funded R&D, to a certain extent, for automotive OEMs, to supplement the work that we're doing in-house," he said. "There are areas where we need support and help because our products are changing so much. They are getting electrified, they're bringing digital services into the vehicle. That's something we've never really done in our history."

This was evident at CES 2020, where Honda brought six of its Xcelerator-affiliated startups to join the automaker in its booth. SoundHound and Drivemode represented "successful collaborations." Two others, Noonee and Skelex, showed their work on augmented ergonomic technologies for Honda employees. And Monolith AI and UVeye represented how Honda is introducing AI technology to its production process in order to help Honda workers.

"It's not about replacing a human being,” Clark explained. “It's about supplementing and augmenting their activities. That's really where we see, as a company, the value of AI. It almost gives us superpowers." The artificial intelligence in Monolith AI's name refers to the way the company's software analyzes old Honda vehicle data then applies deep learning to the design of parts, specifically CAD modeling. Honda has designated three factories as test sites. Integration of the technology into real-world business applications would be in a year or two, Clark surmises.

Eliminating dead ends

"Leveraging AI with past data can help you better understand your current design," said Monolith AI's senior AI engineer, Joël Henry. "AI or machine leaning will tell you, based on past designs, that this specific part of your design will not work or will not be compliant or will fail this particular test and therefore you should change it."

Monolith AI's technology is used in the aerospace industry as well as automotive and works better the more historical data there is available to analyze. "Instead of spending weeks or even months to build and test something, you could have the first insight from the AI tell you that's not going to work, you shouldn't even bother," Henry said. "Sometimes, it will tell you something may work, and you can get rid of all of the dead-end paths."

UVEye’s vision machine learning is used to conduct inspections as a vehicle moves down the production line. Unlike Monolith AI, UVeye, does not need original data files from an automaker. This allows UVeye to work with the aftermarket as well as helping build new products. COO Ohad Hever said his company already has "millions of scans" that it can offer to used-car dealers, for example, who want to look over vehicles they are thinking of purchasing.

On a production line, the AI uses cameras to learn what a correct part or vehicle should look like and then compares that with each unit as it goes past. The Honda project is one of UVeye's biggest, but the company has talked publicly about working with a handful of other OEMs, including Volvo, Toyota, Daimler and Skoda.

Hever said letting an AI pour over moving parts is not only easier than having humans do it, but it's also better, since the software can find smaller issues before they become big ones. "We are detecting anomalies in the early stages, so they don't need to find it in the final stages and return the vehicle," he said. "This is very helpful to them."

The overall Xcelerator project has been helpful to Honda, too, said Clark, and proves that OEMs need to work with startups to succeed. "Three or four years ago, most automotive engineers were very reluctant to work with startups unless they were doing something very research-y," he said. "That is changing. There are thousands of startups out there that are potential partners."