Hot Startups for March 2018: Nauto, DeepMap, TuSimple

It should come as no surprise that driverless cars are gearing up to be the next big thing in today’s world, and will change how people travel from destination to destination. According to a recent report, the global autonomous vehicles market is expected to reach $126.8 billion by 2027 and major auto companies are investing in AI to stay relevant in this new era.

Startups are also seeing major success in this booming industry. The startups we are featuring today—Nauto, DeepMap, and TuSimple—are just a small handful of companies developing the technology behind fully autonomous vehicles. “What we are seeing is that self-driving, or autonomous driving, is only at the beginning. I’m surprised that we are moving much faster than I expected five years ago,” notes DeepMap founding member, Tom Wang. “There are so many challenging problems —the hardware, software, deep learning, infrastructure problems. All of these I think are at the beginning, and it’s starting to evolve.”

Be sure to keep reading for more on Nauto, DeepMap, and TuSimple.

Nauto (Palo Alto, CA)

Founded in 2015, Nautohttps://www.nauto.com/ is focused on building a data platform to make vehicles on our roads safer while also improving the technology for autonomous vehicles in the future. The company’s multi-sensor windshield-mounted device assesses the full context of the driving environment—including the driver, the vehicle, and the road ahead — in real-time using bi-directional cameras and other sensors. The device’s external camera processes what’s in front of the vehicle to assess what risk elements are present. The internal camera is able to assess and understand the driver behavior using embedded computer vision, particularly whether the driver is distracted. Today, 70 percent of collisions and casualties from vehicles are due to drivers being distracted, with 1 in 4 drivers being distracted by mobile devices. Nauto’s intelligent edge device identifies driver distraction in real-time.

The second part of Nauto’s technology is their cloud-based software system for fleet management, which allows the fleet to understand the level of risk for potentially distracted drivers. Sanket Akerkar, Senior VP for Fleet and Insurance, explains that with the use of artificial intelligence and machine learning algorithms Nauto can “provide that information to the fleet, as well as the driver themselves, on their scoring and the safety level of their driving.”

Nauto’s platform allows fleet managers to monitor their drivers’ safety levels with what Akerkar describes as Attentive Driving (level of distraction) and Smooth Driving (acceleration, braking, and cornering). These events are scored for level of risk and uploaded into the cloud along with a 30-second clip (20 seconds before and 10 seconds after the event) that fleet managers can then review and use to coach and improve driver safety. The scoring is then aggregated into Nauto’s proprietary Visually-Enhanced Risk Assessment (VERA) Score—a single metric to that enables fleet managers to evaluate the safety level of their fleet and individual drivers.  Review of the events that contribute to low VERA Scores allows the fleet manager to make informed decisions about their drivers. As Akerkar puts it, “We want to make sure that they can understand the real context of what was going on, and then be able to provide the right coaching advice at that moment for that driver.”

DeepMap (Palo Alto, CA)

DeepMap is a full-stack mapping service that creates HD maps for the autonomous driving industry. Without a human driver, self-driving cars rely on maps with real-time localization to constantly update changes in road conditions, accidents, construction, and more. For cars to be able to maneuver autonomously, they must be aware of all conditions around them – DeepMap is helping solve this issue by providing real-time information that is then pushed to the map, and quick decisions can be made for safe driving. Instead of collecting data the traditional way using expensive survey vehicles, DeepMap is able to leverage existing hardware in vehicles to collect information and build HD maps for automotive firms. As Tom Wang, one of DeepMap’s founding members, explains, “Our map and localization service can be integrated with vehicles’ planning, perception, or control system. They can rely on our map to make better decisions, using more up-to-date information.”

So how exactly does DeepMap create their HD maps? Wang explains that the company’s technology allows them to work directly with their customers and the standard self-driving cars they provide and utilize existing cameras, lights, and sensors. Doing this allows the company to work with OEMs directly – they provide the cars and DeepMap’s technology will map the environment and create extremely accurate maps (down to the nearest 5 centimeters) that are shared with other vehicles on the DeepMap platform. Computer vision engineers will process the data and convert it to digital map data, while infrastructure engineers ensure they are able to build highly scalable maps. Deep learning engineers are making sure they are extracting features from the sensor data efficiently, which is important for lowering costs.

Wang also notes that, “Our mission was really to build a self-driving HD map’s Internet scale…we are cloud-agnostic, which means we will support any public cloud. We can also support a private cloud. We believe we have a big advantage when it’s time to roll out the self-driving technology to more cars, more roads, and more cities. We’ll be able to scale out really quickly.” DeepMap has raised a total of $32 million in three funding rounds since its inception in 2016.

TuSimple (San Diego, CA)

With the rise of e-commerce in today’s world, one industry has felt a noticeable change – the trucking industry. Strict regulations mean that truck drivers can only drive for eight hours per day, and there is a general shortage of long haul and line haul drivers crossing state lines, in part because of the toll it takes on a driver’s body. TuSimple is a level four autonomous commercial trucking company that uses deep learning and artificial intelligence to drive where a human driver normally would. Using an array of cameras, TuSimple’s platform scans the surrounding environment and makes decisions to safely navigate heavy freight trucks. The company is currently testing its product in Arizona, partnering with the University of Arizona’s College of Engineering to formalize internships and opportunities with other companies.

TuSimple—which recently raised $55 million in Series C funding led by Composite Capital—is developing deep learning algorithms to guide trucks along safe and fuel-efficient routes. The company executes their own HD mapping and their mapping vehicle creates routes for the trucks to then follow. The routes are fairly consistent, but do account for changes in weather and terrain. Robert Brown, Director of Government Relations at TuSimple, explains how the trucks drives just like a human would, noting, “If we hit blizzard conditions and a truck driver needs to pull over because he can’t see, it’s the same for us – we’ll have to pull over. As long as we’ve mapped it, the software stack can handle traffic variations and traffic jams, in and out.”

Be sure to check out our full interview with Robert here.