by Marc Amblard / Founder & Managing Director, Orsay Consulting
Autonomous ride-hailing service is now deployed at small scale in three cities across the world. Scaling to more markets will take time. However, the industry is developing bespoke, fit-for-purpose vehicles, a.k.a. robotaxis, to serve this promising market. These vehicles will operate from any point A to any point B in a given Operating Design Domain (ODD, i.e., a set of conditions such as geography, weather or time of day) – unlike shuttles which already operate on fixed routes.
Autonomous ride-hailing service without a safety operator is currently offered by Waymo in Phoenix (and soon San Francisco), WeRide in Shenzhen and Cruise in San Francisco. Scaling to more cities will be slow initially until companies establish more efficient methodologies to learn new environments and driving conditions faster.
By signing up to the Autonomy newsletter you agree to receive electronic communications from us that may sometimes include advertisements or sponsored content.
The necessary Level 4 (L4) autonomous driving (AD) tech has benefited from tens of billions of dollars in capital injection in the last few years, including about $20B between mid-2020 and the end of 2021. There has also been signification concentration and vertical integration recently among the top players. All of this will result in fewer companies developing AD tech faster.
Product Approach, Tech, Features and Differentiation
Operators of autonomous ride-hailing service are currently using retrofitted, standard passenger vehicles which provide a temporary solution. Fit-for-purpose robotaxis won’t require steering wheels, pedals, and rearview mirrors. Instead, they must be developed from the ground up with things like redundant x-by-wire functions, integrated sensors (Lidar, radar, cameras), a specific electric / electronic architecture to connect sensors, actuators, and the AD computer, upgradable hardware and software, all of this with a longer life expectancy than current vehicles – perhaps a million km.
While designing robotaxis present technical challenges, it can also enable new experiences for users. For instance, the cabin layout can be arranged for more conviviality, i.e., face-to-face seating. It should be possible for passengers to import their digital profiles so as to create another living space. Monitoring the vehicle will be critical, e.g., confirming riders’ identification, assessing their behavior once onboard, identifying objects left behind or ensuring the cabin is clean before other riders are onboarded.
Differentiation among robotaxis will reside in their cabin and their ODD. I imagine various seating capacities, comfort levels, feature sets or branding which ride-hailing operators will market to different groups or use cases. I also anticipate that modular architectures will enable vehicles to morph from people movers to last-mile delivery vehicles depending on the time of day.
Products Already Announced or Even Presented
In January 2020, Cruise was the first to present a bespoke robotaxi. Origin (above) is a 6-passenger vehicle jointly developed by Cruise shareholders GM (now 80% of the equity) and Honda, with L4 AD provided by Cruise. The robotaxi will replace Chevrolet Bolts which Cruise currently use for its ride-hailing trial. Produced at a GM plant in Michigan, the vehicle is expected to be first deployed in both San Francisco and Dubai in 2023.
Amazon-owned Zoox unveiled its own robotaxi in December 2020. The 4-seat vehicle will be able drive in either direction for ease of operation (see at top of article and below). It will be capable of 120 km/h in order to drive on freeways for cross-city rides. Zoox will assemble vehicles in-house in the Bay Area from modules sources mainly from incumbent suppliers and proprietary L4 AD tech. We should see the first vehicles on the streets of Seattle later this year.
Waymo has been operating its ride-hailing service with Chrysler Pacificas in Phoenix and Jaguar iPaces in San Francisco. In late 2021, the Alphabet-owned company announced a partnership with China’s Zeekr (division of Geely), challenging Stellantis’ and JLR’s historic position as vehicle partners. The bespoke vehicle will be fitted with Waymo’s “Driver” (see below). A notable difference with the previous two robotaxis is that all seats are forward facing and packaged with a more conventional form factor.
Recently, Volkswagen announced its plan to develop bespoke autonomous vehicles to transport people or goods on the Buzz platform, leveraging the AD tech developed by Argo AI, a well-funded Silicon Valley startup in which Ford and VW each own 40%. These vehicles will likely be deployed on VW’s Moia mobility platform starting around 2025.
China’s Geely will not only partner with Waymo (above), but also with open-source AD tech developer Baidu. Their “Jidu” JV will jointly develop a robotaxi with the support of ZF, targeting to launch in 2023. The concept presented earlier this year offers an unconventional form factor (see below).
And we should expect the robotaxi offering to expand. For instance, Tesla’s Musk just announced a forthcoming, dedicated « robotaxi » model using the company’s AD software – which has yet to make significant progress to reach L4.
Most OEMs will progressively join the bandwagon either as vehicle providers or as full-service operators, offering autonomous ride-hailing service as well. However, I suspect most will remain on the sideline until they have clear visibility as to when the AD tech will be mature enough to scale the service.
When Should We Expect Robotaxis to Become More Widely Available?
Good question! As stated earlier, autonomous ride-hailing service is available at very limited scale in only three cities. In parallel, operators are learning the environment of several other cities, creating ad hoc 3D HD maps and training their AD models on all-important simulators. This remains a complex and expensive process though companies are learning how to be more efficient at scaling.
In addition, the regulatory framework is yet to be formalized in many jurisdictions. Last summer, Germany passed a bill authorizing L4 vehicles (without operators) on public roads and France passed one to define responsibilities in case of an accident. Last month, Japan reached a status similar to that of Germany and the USA’s NHTSA issued final rules for vehicles without manual controls. Yet, the industry is still missing clear qualification requirements to qualify for L4 robotaxi operations.
In the end, it will be easier to scale fleets in cities where robotaxi operations exist than to open new cities. Nevertheless, I expect robotaxis to roam in a couple dozen cities by the end of the decade.