Why are L4 Autonomous Driving Firms Delivering L2 Solutions?
The Driving Automation industry in China is competitive. A recent trend is that a number of companies defined as L4 autonomous driving are offering L2 and L2+ solutions to OEMs and Tier 1s. Due to the fact that progress in L4 autonomous driving technology commercialization is falling behind, L4 companies are taking a ‘dimension reduction strategy’. This sees them moving down the standard hierarchy to develop L2 technologies and collaborate with OEMs and Tier 1s.
Ways of Collaboration
We classified driving automation solutions providers into four types. The first type is companies that aim to develop comprehensive autonomous driving system solutions integrating software development and hardware design, such as Huawei ADS and HAOMO.AI. The second type includes Tier 1 ADAS suppliers such as BOSCH, Continental AG and Huizhou DESAY SV. The third type is software providers, who cooperate with OEMs and offer software technologies to them. The last is L4 autonomous driving companies that take a reduction strategy to develop and partnership deal with OEMs and Tier 1s for L2 technologies. For example, WeRide received a strategic investment from BOSCH, and reached a strategic cooperation agreement to jointly commercialize and mass produce the L2-L3 driving automation of passenger vehicles.
Why do OEMs not produce l2-l4 autonomous driving solutions by themselves?
It is risky to build a new team for traditional OEMs because they must make some internal changes in the company’s culture and organization. These changes may not be rewarded/fruitful/yield good results. For example, if a company builds an autonomous driving team, workers need time to know each other before producing efficient teamwork. Thus, the cost of time should be considered. Moreover, one of the most important reasons is the R&D risk. For the OEMs, it takes time for their R&D investments to yield outcomes. Sometimes, their money is spent in vain. Nowadays the driving automation (L2-L4) industry is crowded, as many players already possess (or boast of being about to have) mature technologies. Therefore, most OEMs choose to cooperate with autonomous driving firms for lower risks and costs; there is no need to reinvent the wheel.
Senior Automation (Chinese: 斯年智驾) is a company specializing in autonomous logistics heavy trucks and port Intelligent Guided Vehicles. In addition to developing autonomous driving software and hardware system, it also provides logistics and transportation services with pure electric robotrucks or IGVs. Up to now, Senior Automation has been commercialized in several large ports and has completed some pilot projects of Hub-to-Hub self-driving in Tangshan, Zhuhai and East China.
Wang Chuhao, the CSO of the Senior Automation, told EqualOcean in an interview that OEMs have strength in the capacity of system integration, which allows OEMs to offer unique capabilities and features. With components and sub-systems communicating with each other, the whole is better than the sum of individual components.
Why are L4 Autonomous Driving companies moving down the standard hierarchy to develop L2 technologies?
First, the L4 companies have some software advantages which include algorithms, system smoothness and architecture.
DeepRoute.ai (Chinese: 元戎启行), committed to advancing urban logistics and popularizing Robotaxis, is comprised of two product lines, DeepRoute-LINK (Robotruck) and DeepRoute-INJOY (Robotaxi).
Maxwell Zhou, the CEO of DeepRoute.ai, told EqualOcean that “L4 companies are proficient in deep learning algorithms as L4 autonomous driving needs to tackle much more complicated scenarios than L2. Also, the system smoothness and architecture. Currently, ADAS development on the L2/L2+ vehicles is achieved through different modules like ACC, AEB, LKA etc. and it needs continuous switching between individual systems during the ride. It happens that a few suppliers are involved in the L2/L2+ system development, so the system needs constant switching and may not run as smoothly as the case in which the L4 company takes the responsibility for the overall system. At DeepRoute, we constructed a one-code framework to accommodate both L2 and L4. This is very challenging on the technical side because L4 autonomous driving requires high computing power and energy consumption. Our algorithms are optimized to run efficiently and stably on low-cost, low-power consumption platforms.”
Second, the lack of LiDAR and computing power. LiDAR was the barrier, but with the development of the sensors in recent years, high-quality with lower-priced LiDARs are installed on vehicles. For the computing power, J5 is the third-generation AI chip that is developed by Horizon Robotics and has an eight-core ARM Cortex A55 CPU and a dual-core Horizon Bayesian architecture Brain Processor Unit (BPU) for AI computing, with a single chip capable of 128 TOPS of computing power. This meets the considerable computing power needs.
Third, commercialization. The commercialization of L4 autonomous driving is slow, and L4 companies are in dire need of money. L2 or L2+ autonomous driving is an opportunity for companies that focus on L4 autonomous driving.
He Bei, the CEO of the Senior Automation said that “Developing L2 autonomous driving technology does not mean that the companies will abandon the L4 technology. L2 and L4 teams are separated in the L4 companies. The main reason for L4 companies is focusing on L2 because they want sufficient cash flow. Meanwhile, developing L2 and L4 autonomous driving technologies makes these L4 companies more valuable.”
Hongjing Drive (Chinese: 宏景智驾) is a provider of computing platform solutions focused on autonomous driving. The company aims to develop comprehensive autonomous driving system solutions integrating software development and hardware design, providing a safe and efficient transportation method.
Liu Feilong, the Founder and CEO of Hongjing Drive, accepted EqualOcean’s exclusive interview and mentioned that the market size of driving automation will reach CNY 220 billion (USD 32.85 billion) by 2025, and reach CNY 500 billion by 2030. Initailly, Hongjing was defined as L4 company, but based on the L4 autonomous driving advantages, Hongjing is cooperating with OEMs to develop L2, L2+ and L3. By now, Hongjing believes that without mass production, commercialization is unsustainable. Moreover, to realize L4 and L5 autonomous driving, a step-by-step approach is best.
The cost and complexity
The costs of L1 and L2 are relatively low. Regarding the sensors, Millimeter-wave radar (MMW radar) and smart camera would be qualified for L1 and L2. However, with the L3 and L4 levels, which allow driver’s hands to be free from the steering wheel and pay less attention to the surroundings of the highway, the cost and the complexity of technologies will be augmented. L4 cars need different types of sensory devices to collect information. This includes historical information, vehicle positioning, identifying visible and hidden objects. Therefore, more type of sensors such as cameras, radars, lidars, sonars, infrared, inertial measurement units and global positioning systems. According to TSMC, it is possible that an L4 vehicle can be equipped with 60 sensors. Meanwhile, several sensors overlap to offer redundant functions for system backup in case of sensor breakdown.
The Opportunity of the L2
The figure above shows that the L2 ADAS penetration rates of passenger cars was 7.5%, 13%, 16%,21.5% and 23.2% from Q1 2021 to Q1 2022, it grew at a steady pace. Notably, the NEV penetration rate of ADA is 35%, much higher than the 19.9% of fuel-powered vehicles. With the development of NEVs, the penetration rate of ADAS will increase as well.
In this chart, we see the market share of China’s L2 autonomous driving passenger car shipments by different sources of power in Q1 2022. In this context, fuel-powered vehicles’ shipment takes L2 autonomous driving market share of 67.2%, while China’s EV- Marking brands such as Li Auto, NIO and Xpeng – which have been nicknamed ‘New Forces’ –take a market share of 21.9%, and the traditional OEMs producing NEVs have 10.9% of the market share.
L4 driving automation need more time to mass produce effectively. EqualOcean believes that by the year 2025, the mass production of L4 autonomous driving will be realized in airports and harbors, but the commercialization of L4 passenger vehicles will take a longer time. Moreover, due to the technologies limitations which slow down the L4 companies’ commercialization , and L4 companies need to supplement its cash flow, we think that more L4 companies will collaborate with OEMs and Tier 1s to develop ADAS technologies.