Baidu, a leading Chinese company, began its foray into autonomous driving as early as 2013. Now, the robotaxi trend has spread to the other side of the ocean, even though Elon Musk's Tesla self-driving robotaxi is scheduled to be launched in Texas, USA on June 22.
The WeChat account "Autonomous Cars Are Here" believes it's time for more people to understand the true nature of autonomous driving!
(Further reading:)
"A Century of Autonomous Driving History: 1925-2024 – A Grand and Wonderful Era, with Some People and Events That Will Be Remembered!"
I. What is autonomous driving?
An automatic driving system refers to a complete train operation system in which the work performed by the driver can be fully automated and is under highly centralized control.
In layman's terms, a mature autonomous driving system can replace a human driver and complete the entire process of using the vehicle.
Therefore, some people refer to fully autonomous driving as driverless driving.
II. Current Application Directions of Autonomous Driving
1. Provide intelligent driving convenience for private cars
In recent years, consumers who pay attention to family cars have seen terms such as L2/L2+ in the introduction of different vehicles, and L2 here is actually closely related to autonomous driving.
In the classification of autonomous driving levels, L2 belongs to partial autonomous driving, which generally has four functions: adaptive cruise control, lane keeping, automatic braking assist, and automatic parking. Under specific conditions, it can temporarily disengage from the driver's control, bringing convenience and safety to car travel.
(Further reading:)
Intelligent driving, assisted driving, autonomous driving, driverless driving: an evolutionary revolution from "human control" to "machine control"
Currently, the more accurate term for autonomous driving below Level 3 is intelligent driving or assisted driving.
2. Autonomous driving empowers mobility services
The well-known Carrot Express has already achieved fully driverless travel services.
Based on autonomous driving technology, the vehicles operated by Luobo Express no longer require driver participation. The latest generation of mass-produced unmanned vehicles (sixth generation) in Wuhan can achieve 24/7 fully unmanned travel.
China's Pony.ai and WeRide have already achieved Level 4 autonomous driving.
Wayomo, a technology company under Google in the United States, has also achieved fully driverless driving!
Tesla's CyberCab self-driving taxi, scheduled to debut on June 22, is touted as being fully driverless.
3. Autonomous driving empowers the commercial vehicle sector
(Further reading:)
"Autonomous Commercial Vehicles: Why Has Capital Suddenly Fallen in Love with 'Freight Vehicles'?"
Autonomous passenger vehicles, often seen as "internet celebrities," are poised to empower human mobility with a promising future. Commercial vehicles, on the other hand, are more pragmatic, focusing on closed/semi-closed scenarios such as ports, mines, sanitation, and long-haul logistics. These locations offer fixed environments and simple routes, making technology implementation easier. Currently, unmanned delivery vehicles, unmanned trucks, unmanned mining trucks, unmanned sanitation vehicles, and unmanned agricultural machinery are rapidly developing. This article primarily introduces the situation of autonomous passenger vehicles, which are becoming increasingly integrated into our daily lives.
III. How safe is autonomous driving?
The biggest concern for everyone is whether autonomous driving is safe.
Ultimately, there are countless paths to take, but safety is paramount. Let's take the "Radish Run" as an example to illustrate this point.
1. Vehicle-side hardware infrastructure
Luobo Express's autonomous vehicles employ a multi-sensor fusion solution, combining LiDAR, cameras, millimeter-wave radar, and ultrasonic radar to achieve Level 4 autonomous driving capabilities. Sensor configuration and function descriptions for typical models (such as the Apollo RT6):
Sensor Configuration Overview
1.1 Sensor Layout and Division of Labor
LiDAR: 1 main radar on the roof (forward long-range detection) + 1 blind spot radar on each side (lateral coverage) + 1 rear radar (blind spot elimination).
It achieves 5×360° perception without blind spots, and its point cloud density is said to be 1.5 times that of Waymo vehicles.
Cameras: 4 front-view, 4 side-view, 2 rear-view, and 2 surround-view cameras, covering dynamic targets within 200 meters.
Millimeter-wave radar: forward (3) + lateral (2) radars, complementing lidar data to solve the problem of detecting obstructions (such as pedestrians suddenly crossing the road).
2. Underlying logical security strategy
L2 and L4 autonomous driving are not linear evolutions, but two technologies based on completely contradictory safety responsibility models. Their differences are like the gap between feature phones and smartphones, which are different in underlying safety logic.
The essence of Level 2 assisted driving is "human-machine co-driving, with humans bearing the ultimate responsibility." As an enhancement tool for the driver, it requires humans to always monitor the system as the ultimate responsible party. Its core risk lies in the "ambiguity of human and machine responsibilities." The contradiction between the driver's excessive trust and the system's capability boundaries can easily lead to takeover failure at critical moments.
Level 4 autonomous driving is characterized by "system-led, closed-loop responsibility." It is designed to replace humans as the fully autonomous driver in specific scenarios, with the system bearing complete responsibility for safety. The vehicle must independently handle all emergencies and ensure safety through system redundancy, fundamentally avoiding the risks that require human intervention.
Therefore, the leap from L2 to L4 is essentially a shift from "optimizing human driving" to "replacing human driving," and its core is a fundamental transfer of the subject of safety responsibility.
2.1 Ten-fold security redundancy strategy + six-fold MRC strategy
True autonomous driving hinges on absolute control over "what ifs".
To address this, Carrot Express has built a full-stack, ten-fold safety redundancy system that spans perception, computation, positioning, execution, and even power supply and heat dissipation. The design philosophy behind this architecture is that, in the event of any single failure, the vehicle can not only ensure safety but also continue to reliably complete its driving tasks.
Based on 170 million kilometers of real-world operational data from Luobo Express, they identified millions of complex scenarios and for the first time defined an unmanned MRC (Minimal Risk Condition) safety strategy, designing a six-fold MRC.
The system provides the most user-friendly safety experience through tiered or skip-tiered approaches. For example, in the event of inclement weather such as rain or fog, the system automatically identifies the problem and applies speed limits, allowing customers to continue their journey safely.
Compared to existing advanced driver assistance systems that automatically disengage upon malfunction and require driver intervention, the autonomous driving system, through its MRC (Mechanical Control Controller) strategy, can continue to operate safely even in the event of a low-probability malfunction.
2.2 Large-scale models enable leaps in autonomous driving technology capabilities
With the development of artificial intelligence technology, the shift from traditional modular design to large "end-to-end" models has become an important trend in autonomous driving technology.
Early autonomous driving systems mostly relied on rule-based architectures, typically consisting of multiple independent modules such as perception, decision-making, planning, and control, which worked together like an assembly line.
This method has certain limitations when dealing with complex and diverse traffic scenarios, especially in handling long-tail scenarios (such as road construction, extreme weather, and sudden obstacles).
In contrast, large-scale models employ ultra-large-scale parameters and complex neural network structures, and are trained on massive amounts of data, giving them stronger generalization capabilities and enabling them to cope with various complex traffic environments.
It not only demonstrates a higher level of intelligence in interaction and reasoning, but also handles long-tail and extreme scenarios more flexibly, greatly reducing potential security risks.
Since 2017, Baidu has been pushing its autonomous driving system to gradually shift from a rule-driven to a data-driven, model-based architecture.
By 2023, Baidu had taken the lead in applying its large visual model technology in the field of autonomous driving.
In 2024, Baidu launched Apollo ADFM, the world's first large-scale model supporting Level 4 autonomous driving. This model not only balances safety and generalization capabilities but also achieves comprehensive coverage of complex urban scenarios, potentially offering more than ten times the safety of human drivers.
3. Emotional and privacy security in travel service safety
Autonomous driving is not only changing the way passengers travel, but also gradually reshaping a warm and caring travel experience.
For passengers, driverless cars mean safe, convenient, and stress-free travel. They eliminate emotional risks—no road rage, no verbal conflicts with passengers, maintaining focus and efficiency at all times, and ensuring safety throughout the journey.
Taking Wuhan as an example, more than 90% of residents believe that driverless cars are safer than human drivers, reflecting the gradual establishment of public trust in technology.
In terms of emergency response, the unmanned vehicles are provided with "customer service"-style support by remote safety officers.
If a passenger suddenly feels unwell or encounters other emergencies, the safety officer can not only assist in safely terminating the trip, but also directly call 120 (emergency services) to ensure the passenger receives timely medical assistance. This proactive protection model has won the trust and praise of many users.
Privacy protection is also a major advantage of self-driving cars.
Driverless cars offer more private spaces and, unlike some ride-hailing drivers who may engage in unauthorized filming or recording, thus better protecting passengers' personal privacy.
More importantly, driverless cars transform automobiles from cold, impersonal devices into warm, intelligent companions, providing a more personalized and humane travel experience, especially for special groups such as people with color vision impairments, the elderly, and women.
Currently, autonomous driving has gradually moved from scientific research and experimentation to the stage of urban demonstration and application.
For example, the "Vehicle-Road-Cloud Integration" pilot program, launched in July 2024, has laid infrastructure in many cities across my country, promoted the intelligent upgrading of roads, and "paved the way" for unmanned vehicles.
Currently, the world is at a crossroads in the development of autonomous driving. Whoever can take the lead in transforming from technological breakthroughs to large-scale applications will gain a first-mover advantage and dominate the rules and standards of future transportation.
At the national level, positive signals are being continuously released, encouraging local areas to take the lead in piloting and allowing technologies to be validated and matured in a wider range of scenarios.
This will not only help Chinese companies take a leading position in the global autonomous driving field, but also give domestically developed technologies a favorable position in international competition.
As industry insiders have summarized: "We need to constantly experiment and optimize on the most realistic path in order to seize the commanding heights of the future."
IV. Will autonomous driving cause traffic jams?
If pedestrians jaywalk and cross the road illegally, and an autonomous vehicle chooses to yield for safety, is the traffic jam caused by the autonomous vehicle or the pedestrians jaywalking?
It should be said that autonomous driving itself does not reduce traffic efficiency; chaos is the source of traffic jams.
The current traffic congestion and complexity in China stem from the blurred lines between motor vehicles, non-motor vehicles, and pedestrians. Electric scooters often travel against traffic or cross the road without using crosswalks. Of course, there are also private cars that drive slowly, occupy lanes illegally, and disregard traffic signals.
If there were only self-driving vehicles on the road, it would probably be much more efficient!
Autonomous driving is the most steadfast enforcer of traffic rules!
Autonomous driving will operate at the maximum speed permitted by road speed limits;
Human drivers often determine their speed based on their mood, road conditions, weather, and driving skills, which can easily lead to ambiguity in the enforcement of traffic rules.
When faced with an intersection, autonomous driving will only follow the rules of stopping at red lights and going at green lights; and while waiting, large-scale autonomous driving fleets will be precisely linked together, squeezing out gaps in all directions to save road space and meet the needs of as many vehicles as possible to travel at the same time.
Therefore, autonomous driving is generally not the main solution in traffic jams!
The simplest way to avoid traffic jams is for everyone to follow the rules: don't park illegally, don't block lanes, and don't run red lights. This will ensure smooth traffic flow and is the real solution.
In the future, with the increase in autonomous vehicles, traffic order will improve, and traffic jams and accidents will drop dramatically!
V. Will autonomous driving take away the joy of driving?
Autonomous driving is a tool that makes travel more convenient and safer, but the tool itself is cold and impersonal. It may not focus on the feeling of being pushed back in your seat during acceleration or the thrill of drifting through corners.
So, once autonomous driving becomes mainstream, will we lose all the joy of driving?
Such concerns are not without merit, but they are not realistic.
Since the vast majority of car buyers use cars for transportation, and cars are essentially a means of transportation to get from point A to point B, autonomous driving is designed to better accomplish this task.
If you love driving and enjoy the fun of driving, you can still choose to drive a car yourself, or drive a go-kart in your spare time, or go to the track to participate in more competitive driving activities.
Enjoy autonomous driving, but don't rely on it entirely. No matter how high the level of automation, humans will always be the final decision-makers.
6. Does autonomous driving require absolutely no human intervention?
Early autonomous vehicles were equipped with onboard safety drivers, but with the relaxation of regulations and the advancement of technology, autonomous vehicles no longer require safety drivers or drivers to accompany them.
However, this does not mean that autonomous driving does not require human intervention at all. Autonomous driving may be too cautious in scenarios such as passing other vehicles on narrow roads, maneuvering at intersections, or being cut off. When it hesitates, remote support is still needed.
It can be said that autonomous driving does not mean that humans are not needed, but rather that humans are not needed to perform simple, repetitive, and fixed operations. What it needs more is human intelligence!
Therefore, the system will continue to closely monitor autonomous vehicles and communicate with passengers as necessary, since machines are more reliable than humans when it comes to driving.
7. Will autonomous driving affect employment?
Since we've discussed the relationship between autonomous driving and humans, will the widespread adoption of autonomous driving affect employment?
It depends on which angle you analyze this problem from.
Although autonomous driving does not require drivers, which will reduce the number of ride-hailing drivers, at the same time, intelligent connected vehicle testers, ground staff, and back-end support staff have become indispensable new positions in autonomous driving.
Furthermore, the development of autonomous driving requires vehicles and sensors, both of which are inseparable from the manufacturing industry.
The algorithms and models required for the system to run are inseparable from programmers and software development engineers.
The development of autonomous driving itself will not limit or reduce employment; it will only place new demands on workers. The more widespread autonomous driving becomes, the more new jobs will be created.
Just as automobiles replaced horse-drawn carriages, seemingly reducing the number of horse-drawn carriage drivers, but creating dozens of new professions such as automobile engineers, mechanics, designers, and insurance specialists.
The rise and fall of professions often vividly depict the evolution of society!
8. Can self-driving cars help us earn money by picking up fares?
Netizens have wilder imaginations than black holes, and one suggested that we should just buy a "Radish Express" car, which would pick up passengers during the day to make money, and then I could have someone pick me up after get off work. That way, I would not only have someone to drive me around, but also someone to help me make money.
So, will this model actually work?
This idea is certainly admirable, but it may face some difficulties in practice.
For example, regarding the division of responsibilities and rights, the autonomous ride-hailing services provided by Luobo Express involve vehicles operated uniformly by the company, which naturally has dedicated personnel for vehicle hardware maintenance and software debugging;
Furthermore, there is safety supervision during the service process.
If an uncontrollable accident occurs, Luobo Express will naturally bear the corresponding responsibility; however, if it is a privately provided autonomous ride-hailing service, who will bear the responsibility after an accident, and does the individual have the capacity to assume responsibility?
These issues may require further improvement of relevant regulations.
In addition, mass-operated autonomous vehicles can be centrally dispatched, which can increase energy utilization and smooth out peak and valley times through charging and discharging strategy management.
Moreover, they can park freely, and by analyzing travel data, they can deploy vehicles to areas with higher travel demand in advance at night.
If an individual owns such a car, should priority be given to accepting orders or to the owner's own travel needs?
In addition, the unseen idleness and wear and tear cannot be ignored. After careful consideration, I think taking a taxi is still the best option!
9. What is the future of autonomous driving?
Mature autonomous vehicles have now moved away from human drivers, which means that the traditional driver's seat design may be eliminated (due to policy reasons, the sixth generation of Carrot Express still retains a steering wheel, but it is said to be detachable). This frees up interior space, so autonomous driving will change the spatial layout of cars, giving them more freedom and more spacious seating.
When the interior space of a car is large enough, it may become like a house, with more functional areas. The car may then become a "third space" or a "mobile home" outside of people's home and office.
We can easily transform vehicles into cinemas, cafes, classrooms, hotels, restaurants, barbershops, and so on, and users can enjoy services while on the move. For example, I once saw a video of elementary school students doing their homework or playing the violin while sitting on a "carrot running" vehicle. These are all services that unmanned mobile vehicles can provide.
Furthermore, autonomous driving will not be limited to ride-hailing services or private cars; it can fully cover public transportation and subways, creating a completely new autonomous driving transportation system.
In this system, there are fixed areas to provide energy for autonomous vehicles, fixed roads to provide routes for autonomous vehicles, and fixed locations to facilitate passenger gathering and dispersal.
We can allow autonomous vehicles to travel at higher speeds, we can reduce the space they occupy, we can save on the cost of autonomous travel, and more powerful and accurate autonomous driving technology will make our travel safer.
10. Besides autonomous driving, what kind of automation do you need most?
In the wave of intelligence and automation, autonomous driving has certainly attracted much attention, but in addition to the innovation in the transportation sector, automation in other areas is also quietly changing our lives.
For example, a senior living community in Shanghai has introduced a variety of automated robots to improve the quality of life for its residents.
These robots not only have medical-grade disinfection capabilities, effectively preventing the risk of infection, but also enable medicine delivery between floors, reducing the tedious operations of traditional manual labor.
Some robots can even take on the responsibility of night patrols, providing safety for the elderly.
Some people expect that automated robots will help solve the traditional problem of "raising children to provide for old age," allowing the elderly to have a more secure and dignified later life.
So, besides transportation and elderly care, what are the automation scenarios you most urgently need in your daily life?
Do you hope that robots can solve some practical problems for you, such as home cleaning, health monitoring, barrier-free travel, or emotional companionship?
With the ever-evolving technology, each of us can benefit, as long as we find the automation solutions that best meet our needs in life, allowing technology to truly become a bridge to protect and improve our humanistic care.
