Introduction
In recent years, there have been significant changes in the pattern of the domestic automobile industry in China. NIO, XPeng, and Ideal have successively become listed and have achieved quite good results. Meanwhile, more and more consumer electronics and Internet companies such as Huawei, ZTE, and Xiaomi have entered the automotive field and begun to embark on the path of car-making.
Huawei, Xiaomi and other “disruptors” do not focus on the powertrain and chassis domains deeply ploughed by traditional players in the industry but on the intelligent automotive field that requires continuous software iteration based on massive data. Smart driving has a considerable space for imagination and commercial prospects in the intelligent automotive field, which has sparked intense pursuit from major companies.
The Three Major Scenes and Technological Development of Intelligent Driving
The technology stack of intelligent driving is not simple, covering the fields of mapping, positioning, perception, planning, control, etc. However, these are what engineers need to care about, and users cannot feel them. What users can feel is the scene in which they can use smart driving. From the perspective of industry development, there are three scenarios for intelligent driving applications: intelligent driving on highways and ring roads, autonomous parking, and intelligent driving in urban areas.
Intelligent Driving on Highways and Ring Roads
Driving on highways and ring roads usually accounts for most of the driving time during long-distance travel or daily commuting in first-tier cities.
Due to the relatively single scenario on highways and ring roads, there are not many frequent intersections with traffic lights or intersections with many vehicles and people. The emergency situations that need to be faced are relatively concentrated and it is easier to achieve a balance between safety and comfort. Therefore, this has become one of the key areas where major automakers have made efforts. With the assistance of user-provided navigation information, intelligent cars can automatically overtake, change lanes, and enter and exit ramps, instead of simply cruising along the lane.
In May 2019, Tesla launched the NOA function by OTA to Chinese owners, and the intelligent driving function on highways and ring roads came into people’s sight. Soon after, Chinese new forces in car-making followed Tesla’s footsteps. NIO released NOP, XPeng released NGP, and new forces such as Ideal and IM also have the NOA function on the way.
The navigation-assisted driving that has been mass-produced still has some scenarios that cannot be well solved on highways, such as construction zones or faulty stationary vehicles that suddenly appear while cruising on highways. These may cause accidents due to the missed detection of the assisted driving system, so users must be clear about the boundaries of this system and remain cautious when using it.
Autonomous Parking
Many experienced drivers have the dreadful experience of “driving for five minutes, parking for two hours.” In many cases, facing a parking lot with only a few scattered parking spaces, we easily try our luck like headless flies. Sometimes we find an empty parking space, but the surrounding cars are parked slantingly, making the entrance too narrow, and it is difficult not to touch the surrounding cars when parking, so we can only give up and continue to look for spaces.
Facing the pain points of the difficulty in finding parking spaces, parking, and retrieving cars for consumers, a company developing intelligent driving targeted the market with huge potential- Automated Valet Parking (AVP). Compared to high-speed intelligent driving, intelligent parking is concentrated in relatively closed parking areas, and the scenarios are not as complex as open roads, with overall controllable risks.
Intelligent Driving in Urban Areas
Compared to the scenarios of high-speed, ring road, and parking areas, the urban area scenario is more complex in the field of intelligent driving and also one of the most difficult to handle.
When the intelligent car is driving on urban roads or rural roads, it will encounter pedestrians, cyclists, irregular vehicles (cargo tricycles, construction vehicles), static obstacles (water horses, cones, trash cans, temporary barriers, etc.), small objects on the road (triangular warning signs, rocks, etc.). In addition to various bizarre obstacles, it may also encounter weak light, backlight, heavy rain, and heavy fog.
Any one of the above targets missed by the pure visual intelligent driving solution may cause serious consequences in intelligent driving. Due to the limitation of angular resolution, the detection signal-to-noise ratio of the millimeter-wave radar is too low for the detection of the above scenarios. If used rashly, it will cause many false alarms and false positives. Therefore, the intelligent driving solution with the main sensor scheme of vision + millimeter-wave radar has a significant hidden danger in these complex scenarios.
Due to the great difficulty of intelligent driving in urban areas, only a few car manufacturers such as Zhi Ji and Ji Hu have tested intelligent driving in urban areas on public roads.
Closed Loop of Intelligent Driving Scenarios
Both mainstream car companies and “disruptors” represented by consumer electronics and Internet companies recognize the importance of linking urban/highway/parking lot closed-loop of intelligent driving scenarios. The closed loop of the full scenario means that the car can automatically drive out of the parking lot, automatic driving in urban areas, automatic highway driving, and then automatically park into the parking lot without the driver’s intervention, which will greatly improve the driving experience.
When designing the route of intelligent driving, many car companies will first sort out a User Story to help them better plan the technology. The following picture is Zhi Ji’s understanding of the closed loop of intelligent driving scenarios. This picture has a vivid name D2D (Door to Door) Pilot.
By integrating the functions of Intelligent Summon, AVP (Automated Valet Parking), Highway NOA (Highway Navigation on Autopilot), and City NOA (City Navigation on Autopilot), the assisted driving system can complete the entire driving mission. Any interruption of AVP, Highway NOA, or City NOA will result in the interruption of the entire driving task.
Hardware Upgrade for Intelligent Driving
In the previous section on intelligent driving in urban areas, we mentioned that although the intelligent driving solution based on visual and millimeter-wave radar can realize simple intelligent driving functions in urban areas, there are more or less safety hazards that affect the safety of drivers or traffic participants. The addition of lidar is to help intelligent vehicles solve these potential safety issues.
If the camera is compared to human vision, then lidar can be compared to human touch. The laser beam emitted by the lidar is like a hand extended by a person, touching the road surface or the wall. The redundancy of multiple sensors, especially the addition of lidar, can better guarantee the safety of intelligent driving.
From the recently released vehicle models with intelligent driving capabilities, the models equipped with lidar are either top-of-the-line or second-tier, and the mid-to-low-end models do not have a hardware upgrade path after purchase. This makes the intelligent driving potential of these models limited, and they may not be able to achieve higher-level intelligent driving functions technically. After all, cars, unlike electronic products such as smartphones, cannot be replaced every one or two years. This makes car owners with a limited budget but hoping for future intelligent driving upgrades confused for a while.
This is a commonly overlooked issue in the field of intelligent driving – hardware upgrades. Consumers’ demand for hardware upgrades is there. Therefore, how to enable cars to have hardware upgradeable capabilities should be a question that major automakers need to consider.
Currently, as far as I know, only IM Auto has made technical plans for hardware upgrades. They have designed a compatible lidar software and hardware architecture redundancy solution in the area of sensors and computing platforms. Below is the sensor layout of IM Auto related to intelligent driving. Among these sensors, three lidars can be installed by hardware upgrade, providing the possibility of hardware upgrades for mid-to-low-end models.
Apart from upgrading the hardware of the lidar sensor, IM Auto can also upgrade the computing platform. The main control chip is upgraded from Nvidia Xavier (30 TOPS) to Nvidia Orin X (500+TOPS), achieving a leap in computing power from the level of 100 trillion operations per second to the level of millions of trillions of operations per second, in order to support higher-level intelligent driving functions.
Summary
What is the significance of intelligent driving?
For users, when intelligent driving in a full-scenario is realized, their comfort and efficiency of travel will greatly improve, and more and more users will pay for these services.
For car manufacturers or technology providers, the intelligent driving technology market is large, whoever can occupy the intelligent driving market earlier means that they will make more money in the future.
After experiencing the baptism of the mobile Internet wave, both car companies and consumer electronics and Internet companies have a deep understanding of the value of data, originally seemingly unimportant in cars. Of course, they don’t want to miss the “data terminal” of intelligent cars. Intelligent driving, as the most important data center on intelligent cars, has become a field that major companies are competing for.
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.