Recently, Lin Wenqin (nickname “Meng Jianke”), the founder of Shangshanruoshui Investment Management Company, Yitongtianxia Restaurant Management Company and Mei Yi Hao Brand Management Company, who was driving a NIO ES8 car in the autonomous driving mode (NOP Navigation Pilot) was unfortunately killed in a traffic accident on the Hanjiang section of the Shenhai Expressway, at the age of 31.
On the morning of August 16th, there was a new development in the case. The police have summoned a NIO technician for questioning regarding the unauthorized contact with the vehicle involved in the accident. If the investigation result is that the vehicle data has been tampered with or destroyed, NIO will be held fully responsible for the accident and may face criminal charges.
This incident has attracted a great deal of attention, and writing this article feels heavy-hearted. First of all, we express our condolences to the deceased. It is difficult for relatives and friends to accept such an event, and we who are strangers also feel regret. However, from a safety perspective, there are two major issues to be addressed in this accident.
NIO’s Smart Driving Capability
The two major issues in this accident that need attention are: first, the safety confirmation of the vehicle after it collided, especially the battery detection, to ensure that there is no risk of fire; second, whether NIO’s advanced driving assistance system has any problems.
The first issue has been discussed extensively before. If there is a collision, it is necessary to inspect the battery for damage as soon as possible and avoid the risk of fire. I will not go into details here.
Let’s focus on and understand NIO’s advanced driving assistance capability NOP. The full name of NOP is Navigate on Pilot, which is translated as Navigation Assistant in Chinese.
As shown in the following figure, there are mainly two generations: the first generation NT1.0 is deployed on ES8, ES6 and EC6, and has been iterated since 2017. On ET7, which has not yet been officially launched, NIO has upgraded its intelligent driving technology platform to NT2.0. Through its own research and development, NIO has gradually established a full-stack autonomous driving technology capability including perception algorithms, map positioning, control strategies and underlying systems.
In December 2017, NIO Pilot, a driving assistance system, was first introduced at the “NIO Day” event by NIO. The company adopted a method of pre-installing hardware and subsequent upgrade for a long period of time to assist driving at the perception level, which consists of three forward-facing cameras, four surround-view cameras, five millimeter-wave radars, twelve ultrasonic sensors, and one in-car driving status detection camera.
In 2018, the assistive driving was mainly expressed in the form of warning. At the end of 2018, automatic emergency braking was added, and in April 2019, adaptive cruise control was added. The major update milestone was in June 2019 when NIO Pilot completed the first major upgrade, adding seven features, including high-speed automatic assistive driving, traffic jam assistive driving, lane change controlled by turn signal, road traffic sign recognition, lane-keeping function, forward collision warning, and automatic parking assist system. A full-automatic parking system was added one month later.
In February 2020, NIO Pilot added automatic emergency braking (pedestrian and bicycle), overtaking assistance, and lane departure avoidance function. In October 2020, NIO Pilot system started using high-precision maps and added automatic assistive navigation driving function (NOA). On the basis of the alert function for vehicles crossing from the rear side, active braking function was added. The driver fatigue monitoring function was upgraded, and camera recognition information of the driver’s facial, eye, and head posture features was added for comprehensive judgement. In January this year, NIO Pilot added visual fusion full-automatic parking and vehicle close-range summoning, and optimized the NOA function to enhance its stability in active lane changing and merging/leaving main road scenes. By this stage, the layout of the first-generation software system of NIO has been basically completed in terms of functions.
Should NIO Smart Driving Take the Blame?
Figure 3 can be simply understood as the current classification of intelligent driving has six levels, with Level L1-L2 being the assistive driving level, and Level L3-L5 being the automatic driving level. (Editor’s note)
The NIO assistive driving system is set up according to Level L2, deploying hardware first and then gradually improving software. From this concept, several points need to be focused on by car owners and users regarding current Level L2 intelligent driving:
1) The system only works under the supervision of human assistance, and its scope is strictly limited. If working conditions exceed its scope, it will automatically become ineffective or exit.
2) When the system performs automated operations, the driver needs to monitor the road condition and be prepared to take over at any time.### 3) The ultimate driving responsibility lies with the driver. The system is designed for human-machine co-driving, and the driver needs to supervise the driving process to ensure safety. If an accident occurs due to lack of supervision or failure to take control in a timely manner, the driver is responsible.
As shown in Figure 4, NIO does introduce the NIO Pilot Navigation Assistance feature and emphasizes that the driving assistance function needs to be the responsibility of the driver.
At the beginning of 2021, NIO released the second generation of its assisted driving system, NAD. There are changes in the description of the promotional approach for this system, with the goal of moving towards a higher level of self-driving. NIO’s system includes the NIO Aquila ultrasonic system and the NIO Adam supercomputing platform. It can be said that NIO has officially begun promoting autonomous driving, with the goal of achieving point-to-point autonomous driving in all scenarios, including highways, urban areas, and parking, supported by powerful hardware (11 8-megapixel high-definition cameras, one long-distance high-precision lidar, five millimeter-wave radars, twelve ultrasonic radars, two high-precision positioning units, one car-road collaborative perception sensor, and one enhanced driver perception sensor), and gradually iterated software on the supercomputing platform.
From a promotional point of view, the author believes that NIO may need to exercise more control over its promotional approach, such as whether its two generations of systems are easily misunderstood in terms of their promotional approach, and whether they strictly follow the above differentiation in communicating information to consumers, which needs to be checked and confirmed.
Assisted Driving and Autonomous Driving
In the development and promotion of assisted driving and autonomous driving, due to different understandings of business models and technological classification, we see that there are still significant differences in the competition in intelligent driving between traditional and emerging car companies.
New car-making companies, represented by Tesla, mainly want to prioritize the experience. Tesla’s Autopilot FSD, according to the level of classification shown in Figure 3, is at the L2 level, which is an extreme performance experience based on a cost-controllable, pure visual solution. In Tesla’s view, autonomous driving is limited in terms of experience, and software optimization is achieved by comparing the system strategy with human driving through a large amount of data, forming a “shadow mode”. Therefore, this approach is characterized by a variety of functions and a good user experience, and Tesla is not concerned about the many constraints of software and hardware principles in the classification.Traditional automakers strictly follow SAE’s grading design, constantly considering the perception and safety requirements of excluding the driver from the previous L0, L1, and L2 levels. This aspect is largely constrained by regulations, such as Honda’s launch of L3, which was achieved under the restriction of Japan’s partially commercial use regulations.
More and more traditional automakers are making L3 and above autonomous driving functions an important component of future intelligent vehicles. From the current situation, high-speed closed roads are a relatively mature application scenario for assisted driving. Both new energy vehicle start-ups and traditional automakers have developed many L2 assisted driving systems, which are similar in hardware and software. Traditional automakers tend to be more conservative in their promotional messages.
Suggestions for car owners
After all this, we have some suggestions for users of intelligent driving:
First, at present, consumers are easily confused between “assisted driving systems” and “autonomous driving systems”. Assisted driving systems help us improve our lives, enhance driving skills, and reduce driving difficulty. People play a supervisory and controlling role- under these conditions, everyone can use assisted driving with peace of mind without giving up on this tool.
Secondly, using assisted driving systems can effectively reduce driving load, and objectively, there will be a positive cycle of increasingly trusting and relying on assisted driving. However, one must remain rational. Assisted driving systems have usage restrictions and boundaries. Do not treat assisted driving as autonomous driving, let go of the steering wheel completely, and neglect responsibilities for one’s own or others’ safety.
I believe that this accident is a supervision and a wake-up call for all automobile companies. With the MIIT’s control of the OTA of assisted driving functions, safety will definitely be supervised in advance to avoid tragic recurrence.
This article is a translation by ChatGPT of a Chinese report from 42HOW. If you have any questions about it, please email bd@42how.com.