China Autonomous Driving Companies’ Evolving Approaches
Autonomous driving companies have been making the headlines lately with moves like manufacturing cars. Deep down, it represents a rush to commercialization by evolving product offerings and go-to-market strategies. I aggregate these moves into three types of approaches commonly adopted by companies in the landscape.
Solution provider to make cars or to engage in services
In the autonomous driving ecosystem, there are three roles:
- Technology developer, like startups in the landscape.
- Vehicle manufacturer with assembly plants, hardware technologies and a supply network, a.k.a. OEM.
- Service provider like taxi companies, ride-sharing platforms, logistics operators, contractors in port/mining areas, etc.
Ideally, a developer could sell its solution to the OEM, and then the OEM could sell the end product, self-driving car, to the service provider. The AV startups may have initially identified a business model under this assumption, and some of the strategic partnerships and investments were made with the flow in mind.
However, synergies are easier to say than to create. With problems arising in the execution process, technology developers are eying the dominant power in making cars.
Baidu reacted quickly by creating Jidu Auto with Geely to produce smart electric vehicles for the consumer market. The first model is expected to debut at the 2022 Beijing Auto Show.
AutoX invested in ICONIQ, a lesser-known new generation OEM in China, aiming to produce premium vehicles with L4 features for the rental market. AutoX CEO mentioned in an interview that the move was more about having a say in the design and production of L4 vehicles by taking an additional role as the chief product officer at ICONIQ. The production experience in return could enhance its software capabilities.
Recently, TuSimple was reported to have launched a new project named Turing Auto to manufacture cargo vans and later heavy-duty trucks for autonomous fleet operations at lower cost. It will be conducted by a separate entity with funding from TuSimple and external investors.
Pony.ai and Didi have also been associated with plans to make cars but no official announcement yet.
Besides making cars, it is common for autonomous driving startups to be engaged in service operations, which sits downstream of the value chain. Operational experience could help design and improve products. Meanwhile, products could be deployed in real world scenarios once ready.
Inceptio has defined its business model as logistics-as-a-service (LAAS) since the very beginning. While developing autonomous driving technologies for trucks, it has been growing a logistics service network that currently operates over 100 routes with a thousand-sized fleets across China. According to Inceptio CEO, as the technologies mature, it can be incorporated into operations smoothly.
Senior.Auto is positioned as a solution and service provider. It has reached contracts with Ningbo port and Tangshan port, aiming to build a fleet of 60 autonomous trucks within the year. Although trucks are quite heavy assets for a startup, the founder firmly believes the team is gaining in-depth understanding of the scenarios through the process, which would take its products to the next level.
EQ (Yikong) from the mining sector has adopted a similar strategy by building a fleet of wide-body dump trucks to offer delivery operations and services to open-pit mining site.
Expanding to additional application scenarios
This strategy is straightforward, like robotaxi player can do robobus or others and vice versa.
Pony.ai announced the new business unit “PonyTron” in March dedicated to truck business alongside robotaxi operation. It will work closely with truck manufacturers and suppliers to develop autonomous driving trucks for long haul logistics. It claims 80% of the technology is transferrable from robotaxis to robotrucks.
WeRide launched a robobus trial service in Guangzhou in January. The mini robobus, co-developed with Yutong, has no steering wheel or pedals. It circulates in the last few miles within designated areas of public roads. Later in July, WeRide was reported to have acquired MoonX, an autonomous truck startup. However, it is not clear yet if the company is to expand the scope of AV technology to the trucking sector.
DeepRoute is engaged in robotaxi and smart port. It partners with Dongfeng, Caocao (ride-hailing platform) and Futian District respectively on robotaxi services. It has also teamed up with Dongfeng and Xiamen Ocean Gate Container Terminal to run autonomous container trucks. DeepRoute sticks to the solution provider role with “light assets”, which is different from Pony.ai or WeRide that owns the taxi fleets.
Trunk works closely with ForU Trucking on long haul logistics through their joint venture established in August 2020. Meanwhile, Trunk has applied the AV technology to port transportation with commercial operations in Tianjin, Ningbo and Shenzhen Mawan ports.
Fabu’s autonomous driving initially targeted goods delivery in urban areas and extended to smart port since 2019. Currently, it runs L4 container trucks operation in the Zhoushan port.
UISEE is well known by its autonomous tractors for airport while it also works on robobus and robotaxi projects.
“Dimensional reduction” to the ADAS market
“Dimensional Reduction” is a theory out of the science fiction novel The Three-Body Problem and often quoted as a business strategy. Here it means that the companies devoted to L4 autonomous driving offer L2 and plus solutions to OEMs and Tier 1s, competing with the ADAS suppliers. An ADAS solution deals with limited driving scenarios and the technologies are supposed to be transferrable from high to low although the saying is highly debatable in the industry.
Baidu’s Apollo Navigation Pilot (ANP) and Auto Valet Parking (AVP), presented during the auto show in April, are examples of twisting robotaxi capabilities into intelligent driving features for production vehicles. The ANP was developed on the same Apollo L4 technology structure but opted for a pure vision solution for lower cost. While the ANP is being tested and integrated by multiple partners, Baidu’s AVP has been carried on W6 from WM Motor and Moca from the Great Wall Motor’s WEY brand.
Momenta’s “two legs” strategy is similar. The company focuses on Mpilot and MSD in parallel. The two share the same sensor combinations except an extra LiDAR for MSD on robotaxi, through which way the real world data from Mpilot on production models could be leveraged to train MSD.
Haomo.ai, a spin off from the Great Wall Motor, primarily develops low-speed autonomous vehicles for deliveries with partners like SF Express and Dmall while it provides ADAS solutions to the GWM models.
UISEE has also supplied AVP to SAIC-GM-Wuling models.
