Add one more role to Troy Clarke’s already-extensive title at Navistar: Chairman, President, CEO—and Chief Business Anthropologist. “I see that as one of my most important job functions,” he told attendees of his opening keynote at the recent SAE COMVEC Technology Connection conference.
Business anthropology is the study of human behavior and culture, Clarke explained, but perhaps more significant to an industry delving deeper into big data, artificial intelligence (AI) and vehicles on the verge of becoming highly automated, it also is the study of change. “This includes understanding when the time is right for change, knowing when we’re at the cusp of a paradigm shift, and uncovering disruptions that can lead to unexpected leaps forward.”
Clarke expects more disruption to occur in the commercial-vehicle industry in the next 10 years than over the past 100, owing to three technology areas: automated driving, connected vehicles and electric propulsion. Each of these is nearing the so-called Golden Triangle, a “sweet spot” where three factors—advanced technology, economic value and societal benefit—align and the adoption of a technology accelerates, becoming a disruptive force.
With automated driving, the core technology is AI paired with advanced safety systems like radar, lidar and digital cameras. On the economic front, automated technologies promise to improve fuel economy, driver comfort and productivity, thus attracting more drivers to the profession, Clarke explained. In terms of societal benefit, increased automation can deliver reduced emissions and enhanced safety by reducing the number of accidents stemming from driver errors.
“AI augments the driver’s responses to traffic conditions, which is already in place with systems like lane departure warning, collision mitigation and adaptive cruise control,” he said. “Automation can save a great deal of money. Right now, 43% of fleets’ operating costs per mile can be attributed to the driver; that offers a lot of room for efficiency improvement, even when the driver is not replaced. And utilizing drivers to manage multiple vehicles [as in a platooning scenario] can help offset the current driver shortage and ultimately can save hundreds of billions—with a ‘b’—of dollars each year.”
Clarke estimated that 10-20% of the current truck fleet could be fully automated and not displace one driver, the driver-shortage issue being so acute.
As part of Navistar’s foray into business anthropology, executives talked with several disruptor companies such as Waymo and Uber to learn best practices.
“One big learning is it’s best to bring disruption in-house—in other words, do unto yourself before others do unto you,” Clarke said. “What was in common [among disruptors] is both digital innovation and extensive research and analysis. In some cases, they know as much or more about specific pieces of the business value chain in commercial trucking than we do—a company that has occupied this space for well over a hundred years.”
Creating a culture that enables disruption is paramount. “This type of culture is probably even harder to achieve than the innovation itself,” Clarke said. “The disruptive companies bring these young people in [from college] and ask them to do a ton of research—I need more researchers. I have no shortage of people who know what a good Class 6 truck looks like…But that’s not the skillset we need going forward.”
Amazon helps solve “gnarly” problems
Providing a disruptor’s viewpoint at COMVEC, keynoter Ranju Das, who’s part of the technical leadership team of Amazon Web Services (AWS), said that companies’ scientists can spend up to 80% of their time in data banking and managing the infrastructure, when they should be focusing on what they do best—running rapid experiments.
Among his recommendations for how companies can get started with machine learning (ML), Das said to hire ML experts—if you can. Even AWS has difficulty finding enough such experts.
“One thing we do is take engineers out of college and train them on machine learning—create a scalable workforce. There are more engineers than machine learning scientists out there,” he said.
Engineers can broaden their ML skills with, for example, DeepLens from AWS, a fully programmable video camera for developers. It includes tutorials, code and pre-trained models designed to expand deep learning skills.
The best ML solutions come from working backward from real customer problems, Das said, adding that the transportation sector’s move to automated vehicles presents some “really gnarly problems” for its scientists to help solve.
One example: autonomous technology company TuSimple, which was founded in 2015, relies on AWS’s ML expertise and compute and storage capabilities to achieve 1,000-meter perception range for its SAE Level 4 automated trucks. TuSimple spent two years developing the deep learning algorithms that “instruct” its camera-based system (radar and lidar are employed, too) and make sense of the several terabytes of data created per truck each day. The data is stored using AWS Snowball Edge devices, each capable of storing 100 TB of data, with onboard computational capacity that allows for local data analysis and data compression, prior to being uploaded to the AWS Cloud.
Another critical piece of advice Das shared at COMVEC: Never delete one iota of data, because you never know which input might be useful for an algorithm in the future.
“I think it’s criminal to not store data,” he said. “It’s not algorithms [that are important]—algorithms are a dime a dozen—it’s the data. Data is your IP.”Continue reading »