Latest News

Wayve, a new AI-driven driving system that can learn like humans, is attracting automakers.

Wayve, a startup in autonomous-driving technology, is riding the wave of investor interest. The London-based startup has?raised $2.8 billion in funding from a list of investors and strategic partnerships that includes major names from the automotive and technology sectors, including Nvidia, Mercedes-Benz and?Nissan. Wayve announced in June that it would deploy its robotaxis system from Jeep maker Stellantis to Uber's ride hailing network.

Wayve uses an ?artificial-intelligence ?technology called end-to-end machine learning to navigate roadways, which is supposed to instantly translate sensor-generated data into driving decisions, much like a human driver. This is different from the more traditional approach that combines AI with high-definition mapping and software coding to create presets rules for how a car should react in various scenarios, such as unforeseen events.

Wayve's method is similar to that of another major autonomous-driving company, Tesla. Tesla moved to a end-to-end approach a few year ago. Wayve's technology is not like Tesla's, however, as it uses a variety of sensors, including AI chips, to create its system.

It could then license the technology to any developer of driverless cars, according to Wayve CEO Alex Kendall. Kendall is a 33-year old New Zealander and co-founded Wayve in 2017, the same year that he finished his PhD in AI deep-learning at Cambridge University, England.

Kendall said earlier this year that he wanted to make self-driving possible for any car, any brand and anywhere in the world. He was sitting behind the wheel of a Ford Mustang Mach-E equipped with Wayve’s driverless technology, which autonomously navigated San Francisco Bay Area neighbourhoods where the company has a major tech center.

WAYMO EXPANSION FUELS INDUSTRY MOMENTUM

After years of missed targets and overstated promises, the competition in the autonomous driving industry has intensified. Alphabet’s Waymo has experienced rapid growth over the last two years. It now offers paid rides in a dozen cities after more than 10 years of development.

Kendall was one of the few researchers who pursued end-to-end AI a decade ago. Many autonomous-driving developers are now integrating at least some aspects end-to-end AI into their systems.

The AI-centric approach does raise a dilemma: because end-to-end navigation systems are ambiguous and "black boxes" in nature, it is difficult to understand the driving decisions of the vehicle. In earlier versions of driverless vehicles, the software code was used to guide the vehicle to a safe route. It was therefore easier to understand why a car took a particular path.

Wayve's AI driving engine creates a safety map for traffic situations that are unfolding and determines the safest paths for vehicles. Wayve engineers believe the conventional, programming-intensive safety approach hinders an AI driving system's ability to stay safe in unusual cases because it is hard to write rules to prepare for very unusual situations.

Vijay Badrinarayanan is Wayve's vice-president of AI. He said that when such difficult-to-predict situations occur, the safety logic of a preprogrammed system becomes "brittle". Human drivers are safe because they adjust conservatively to unknown situations.

Shooting for Safety at Scale

Waymo is using end-to-end AI, but it also uses a conventional, rules-based method achieved by software coding and mapping, which, according to the company, is still necessary for safety.

The company said that "end-to-end" models were not enough to ensure safety on a large scale.

Nissan, one of Wayve's clients, has yet to feel comfortable with the safety approach.

Eiichi Akashi is Nissan's Tech?chief. His team is closely evaluating Wayve's technology ahead of Nissan's plans to deploy it on a people mover van named Elgrand in Japan during the year that ends March 2028. He says the system is the "most advanced" but that it's "difficult?to?peer in and see how it makes decision?"

Kendall says that Wayve with its major operations in Tokyo Stuttgart and Vancouver should be able expand into new markets quickly because it doesn't need to do the tiresome step of mapping and writing code for local road quirks. Wayve claims to have successfully tested its AI-driven driving system in hundreds cities around the globe without doing any initial preparation work.

Siddartha Khastgir, professor of safe autonomy at University of Warwick in England, says end-to-end model deployment should be quicker than more traditional approaches. He said that he would not say one technology was safer than another.

Phil Koopman, a Carnegie Mellon University computer-engineering professor and autonomous-technology expert, said Wayve's method for handling unusual traffic situations is but one approach, and others may also prove successful. He still believes it will take at least 10 years to safely deploy driverless vehicles across the U.S.

It will probably require new innovations to get there.

(source: Reuters)