Are AVs for People and Freight Close to Scaling?
For decades, people have dreamt of letting go of the steering wheel when driving conditions are boring, e.g., on the freeway or in heavy traffic. Significant resources have been committed in the past few years to make this a reality. This is particularly true over the past two years as the industry experienced staffing increases, consolidation and vertical integration supported by massive fund raising (see my Nov. 2020 and Apr. 2021 articles).
Level 4 autonomous driving (L4 AD) tech developers continue to increase the number of pilots with or without safety operators. The most mature ones have launched or announced commercial activities in the new term, addressing mainly robotaxi, long haul trucking as well as last- and middle mile logistics. Other players focus on less sexy but highly relevant spaces such as mining, agriculture, or yard operations. A few hundred vehicles currently provide commercial service across these segments without an operator onboard while more do so with a safety operator.
Where are we today? How and when can we expect these activities to scale?
About five years ago, Waymo, Cruise or Zoox were planning to launch commercial robotaxi services before the end of the last decade. What happened? Only Waymo did so, starting outside in Phoenix in late 2018, then removing safety operators a year after while limiting its operations to an Operating Design Domain (ODD) of about 120 km2. In China, Pony.ai launched a chauffeured robotaxi service in Dec 2018 and AutoX deployed a driverless fleet in Shenzhen in Jan 2021.
In late 2021, autonomous middle-mile logistics startup Gatik started moving goods for retail giant Walmart without a safety operator behind the wheel – though one is installed in the passenger seat for emergency. Einride has been operating several local logistics pilots in Europe for a couple of years. And Nuro started providing last-mile delivery services for several retailers in 2020. Looking further back, we should not ignore that less glamorous – yet highly relevant – huge trucks that have been operating autonomously for over 10 years in open mines.
Most of the efforts to bring L4 AD to market focused on moving people until about 2020. However, key players reshuffled priorities, putting autonomous trucking ahead of robotaxis or relatively accelerating the former. This is particularly the case for Waymo and Aurora. The higher-than-expected complexity of urban AD combined with a truck driver deficit, more obvious ROI, and potential efficiency gains for trucking lead to this choice.
One also should not forget the few hundred autonomous shuttles – e.g., Navya and Easymile – which have been operating commercially around the world starting in 2016, by and large staffed with on-board operators. However, these vehicles navigate through programmed waypoints rather than from any point A to any point B.
Where are we today?
Beyond the above-mentioned deployments, commercial activities were launched this year by Cruise (in part of San Francisco, without safety operators) and Waymo (in part of San Francisco, with operator). Yet, operating conditions remain quite constrained. Cruise’s permit allows a fleet of 200 vehicles to provide free, public ride-hailing service between 10pm and 6am at 50 km/h max under clement weather. A secondary permit will allow charging for the service. Waymo has started to give rides without an operator to its own employees, pending a broader permit. These are only the most mature examples.
In China, Pony.ai and Baidu were recently granted permits to operate paying ride-hailing activities in Beijing (60 km2 ODD), without safety operators. Pony received a similar permit in Guangzhou on a larger ODD.
On the trucking side, several AD tech developers have been operating pilots. Some have also launched commercial activities with safety operators. In the USA, this is for instance the case for Waymo, Aurora or TuSimple which all operate in southern states (Texas, Arizona) where the weather is more favorable and regulations are lighter. In China, a few players are progressing on a similar path, e.g., Plus and Pony.ai, whereas Locomation continues with its platooning solution where the driver in the following truck can rest.
On the matter of regulation, regions are progressing at various speeds across the world. Europe seems to be most advanced at least at the national level. Germany, France, and the UK leads the pack on matters related to testing and validation as well as liabilities.
In the USA, operators are currently facing a patchwork of regulations – or lack thereof. The federal government is largely absent, making future deployment across state lines an issue. However, the NHTSA, the safety-focused regulatory body, recently issued the final rules related to the absence of manual controls on highly automated vehicles. This opens the door to purpose-build vehicles.
What to expect in the coming years?
There have been many broken promises in the past when it comes to the commercial launch of autonomous vehicles or associated services. Thus, we must take all announcements with a grain of salt.
On the trucking front, several companies have announced the launch of their commercial activities (without safety operators) in 2023 to 2025. This is the case for Aurora Innovation (the first one in late 2023), TuSimple, Plus or Embark. Most will operate with a driver-as-a-Service business model, charging for their hardware and software on a per-km basis to operate on pre-mapped highways (on-ramp to off-ramp). TuSimple is also considering operating Trucking-as-a-Service though competing with clients may prove challenging. Most players have secured OEM partners (e.g., Aurora with Volvo Group) as well as fleet management and service partners to enable an efficient deployment on high-traffic corridors.
As far as robotaxis are concerned, tech developers that have already launched pilots and commercial activities recently stayed away from bold promises, except when they relate to purpose-built vehicles. Several robotaxis have been either presented (Zoox, Cruise Origin) or announced (Zeekr for Waymo, Arrival with Uber etc.) – see my Apr. 2022 article. I expect the former to make their way into pilot fleets starting in 2023.
For these robotaxi and autonomous trucking activities, I expect scaling to first translate in the deployment of more vehicles in established ODDs (cities, highways) and the expansion of these ODDs (new neighborhoods, more segments on the same highways). Opening new ODDs (new cities, new regions) will take more time as they require more complex and costly exercise, e.g., 3D HD mapping.
Most OEMs will do their best to keep up. For instance, VW plans to deploy in 2025 an ID Buzz-based driverless service to transport both people and goods in Hamburg – Argo AI will provide the L4 AD tech. Most other OEMs are certainly considering this option, combining their vehicle expertise with that of AD solution providers. This is a solution to boost revenue from mobility services (people and goods) as private vehicles sales will likely recede.
In the near term, OEMs will increasingly deploy safety-focused Level 2/2+ systems as well as L3 and valet parking solutions. As we move towards the software-defined vehicle, OEMs will have the ability to enhance these functionalities overtime, activate new features and generate revenue from subscriptions. Even if clients decide not to pay for these features, OEMs will be able to shadow-test their products and collect data to improve performance. Level 4 for vehicles targeting the general public is still far out.
Beyond the above modes, I expect L4 autonomy to become increasing pervasive in industries such as agriculture, mining, industrial logistics, professional lawn mowing and more. Autonomous driving is far from mature!
Managing Director, Orsay Consulting
Feel free to comment or like this article on LinkedIn. Thanks!