"Unmanned vehicles 'go to the safety officer' will lead to qualitative changes in the format."
Author | Rob
Edited by | Wenliang
In the current Robotaxi field, there is a role that self-driving companies love and hate, that is, the driver of the vehicle driving position - the industry is more accustomed to calling it a safety officer.
In the early days of most Robotaxi projects, the safety officer was the most critical supervisory role in the car.
In addition to the difficult driving skills such as emergency parking and emergency avoidance, safety officers often undertake the work of assisting Robotaxi in research and development, related data collection and testing.
However, as self-driving technology matures, this role is becoming a key part of the further commercialization of the Robotaxi industry.
Today, the Beijing Intelligent Connected Vehicle Policy Pilot Zone issued the "Implementation Rules for the Management of Unmanned Road Test and Demonstration Application of Intelligent Connected Vehicles in the Beijing Intelligent Connected Vehicle Policy Pilot Zone" (hereinafter referred to as the "Implementation Rules"), and issued the first batch of unmanned manned demonstration application licenses to Baidu.
Baidu Radish Run also officially opened the Robotaxi service that was not driven by anyone in the morning of the same day.
This is undoubtedly an important breakthrough in the development of China's autonomous driving.
Because although there were also sporadic Robotaxi road tests in various places, this time it was officially from the policy level to allow the Robotaxi driver to remove the safety officer.
Security officer, it is time to slowly withdraw from the stage of history.
Removing Security Officers: One step closer to forming a commercial closed loop
According to the Detailed Implementing Rules, the demonstration application subjects that have obtained the notice can carry out unmanned manned demonstration applications of open roads within 60 square kilometers of the Beijing High-level Autonomous Driving Demonstration Zone.
It is understood that Baidu carried out unmanned manned demonstration applications in accordance with the requirements of the Detailed Implementing Rules, invested 10 unmanned vehicles in the first batch, and planned to add 30 unmanned vehicles in the future.
Baidu Apollo told New Smart Drive that since April 28, they will operate in a "mixed dispatch" model to provide Users with Robotaxi services.
The so-called "mixed dispatch", that is, after the user places an order, radish run will randomly dispatch vehicles with safety officers and no safety officers based on the location of the transportation vehicle and the ordering station.
In the early days of most Robotaxi projects, self-driving vehicles were like toddlers, and the driving experience was not stable. When the vehicle is slowing down the street, it will carefully turn when it sees the intersection, and sometimes in the face of random road conditions, it will also lose its temper and suddenly brake.
At this time, the safety officer is equivalent to a big parent, who needs to hold the steering wheel and pull the child at a critical moment.
But as research and development progresses, the driving experience gets better, and safety officers are gradually handing over control of the steering wheel to the vehicle itself. Most of the time today, safety officers no longer need to do anything, but instead stare at the road ahead and the related events of vehicles, as well as respond to and take over emergencies.
In addition to safety officers, various autonomous driving companies are also developing remote cloud control platforms for Robotaxi, supporting services such as cloud driving and remote security officers.
The issuance of unmanned manned license plates allows Robotaxi to conduct unmanned tests on the main drive from the policy level, indicating that such vehicles are actually in the functional verification stage of L4 automatic driving, and Robotaxi has a high enough technical reliability and safety.
But the removal of "security officers" is by no means more than that.
Going to the security officer has always been considered a key part of Robotaxi's transition from the testing phase to commercial operations.
Whether the machine can replace manpower is also one of the important factors considering the commercial feasibility of autonomous driving.
CICC had predicted that Robotaxi's global market space would exceed $2 trillion by 2030.
But the premise of this prediction is based on robotaxi's operational capabilities beyond ride-hailing.
Because the current cost of manpower investment of drivers is large, and the service time is limited, once replaced by unmanned vehicles, these problems can be solved, and traffic accidents caused by factors such as fatigue driving can be avoided, thereby further saving operating costs.
Going to the security officer is a key part of this.
Therefore, the key point of whether the Robotaxi business model can form a closed loop lies in whether laws and regulations allow the removal of security officers, and technical support does not support the solution to security personnel.
McKinsey has predicted that Robotaxi's cost expenditure will reach an inflection point in 2025-2027 compared with traditional taxis, and it is expected that around 2025, after Robotaxi cancels security personnel and deploys on a large scale, its cost advantage will be more prominent.
Image source: Robin Li, author of "Intelligent Transportation"
Throughout the development of global autonomous driving, almost all autonomous driving companies have followed the expansion route of "closed road testing - open road testing - regional road network testing - small-scale operation - commercial operation - multi-city commercial operation".
Nowadays, the development of global automatic driving can actually achieve "unmanned driving", that is, to liberate the driver from the car and uniformly dispatch and operate vehicles through the remote operation control platform, but at present, this ability is only realized in closed and semi-closed commercial vehicle scenarios such as ports, mining areas, and logistics parks.
At present, the Robotaxi test of most of the world's autonomous driving companies and the unmanned manned license issued by Yizhuang this time only remove the safety officer in the main driver's seat, and in order to cope with the emergency, the co-driver or the rear row still retains the safety officer.
Baidu Apollo VS Waymo: China's moon is rounder
In recent years, many of the industry's technology bulls have returned from overseas, and these people have either been dug back to important positions in domestic factories with professional technology and rich experience, or they have returned to China to start a business, or they have moved the headquarters or main business of the company originally founded overseas back to China.
For these self-driving people who have returned from overseas, the moon at home seems to be rounder than abroad.
New Intelligent Driving has previously interviewed a number of executives of self-driving companies, who generally believe that China has a more prominent advantage in terms of data quality.
On the whole, autonomous driving has a more fertile soil for technological iteration in China, and at the same time has more possibilities for commercialization. The soil of technological iteration refers to the rich and diversified road data born under China's complex road conditions.
Overall, the mainstream roads and urban planning in Europe and the United States, the urban core area is very small, the residential and living areas are relatively scattered and sparse, the connections between the main scenes are mostly trunk road expressways and branch roads with relatively low traffic flow, the overall road scene is more standardized, and the traffic participants are relatively small.
The typical urban scene in China presents the following characteristics:
Large area of high-density urban core area, living area, commercial area, residential area mixed distribution, congestion during the peak of travel is the norm;
In addition to motorized vehicles, traffic participants also include a large number of pedestrians and non-motorized vehicles, especially there will be various courier trucks takeaway trucks;
In addition, China's infrastructure construction frequency will be higher, and autonomous vehicles will face more construction and road environment changes, which further increases the complexity of the scene.
"Baidu's car is casually wandering around Beijing, and the data is enough for Waymo to make a year." ——In a previous interview with the media, some insiders pointed out from the perspective of data collection.
According to the 2021 California Road Test Report released in the first quarter of 2022, the apparent value of Waymo's takeover frequency in terms of perception problems, pedestrian problems, software problems, etc., has not changed significantly compared to 2016.
According to the analysis of Guotai Junan Securities, Waymo's apparent "regression" in software problems, pedestrian problems, etc., is its continuous expansion of the scene in order to cover the corner case.
However, from the data point of view, in the 2019 test, the number of manual takeovers that occurred in the "street" scene of Waymo accounted for 76%, while in the 2021 test, the proportion of manual takeovers that occurred in the "street" scene increased to 96%.
That is, in the "street" scenario, Waymo's Robotaxi team's dependence on safety personnel is not reduced but rather increased.
This has made both Waymo and Cruise extremely cautious and slow in expanding the test range of their Robotaxi teams, both of which have been limited to manned operations in California and Arizona.
In this regard, although Baidu Apollo started late, due to its ability to test in a large number of construction occupations, complex road conditions and other diversified scenarios, its test mileage and operation range have increased significantly and the expansion speed has increased significantly.
Up to now, the Apollo test fleet has reached the scale of 500 vehicles, which can carry out manned operations in 9 cities in China, including megacities such as Beijing, Shanghai, Guangzhou, and Shenzhen, and open road tests in nearly 30 cities around the world, and the actual road test mileage of urban roads exceeds the total test mileage of 27 million kilometers.
From the perspective of the number of Robotaxi orders, according to the data of Baidu Apollo Day Technology Open Day on March 8, the number of orders for radish running for half a year has exceeded 300,000 orders, and the average daily peak of bicycles is 28.72, while the order volume of waymo Robotaxi fleet, according to estimates, is only 26,000 to 52,000 orders per quarter.
More perfect, sufficient, and diversified scene verification means that Robotaxi can establish a more perfect automated problem collection and scene clustering system, which also means that there is a stronger security guarantee, in this regard, Baidu Apollo has gradually thrown off Waymo.
In addition, the high cost of bicycles is also an important factor hindering the further mass production of Robotaxi.
In fact, whether it is automatic driving or unmanned driving, its foothold is the carrier of "car", and it is inseparable from cooperation with car companies, in this regard, Waymo is not as good as GM's "pro-son" Cruise, nor as good as Baidu Apollo.
As early as 2019, Baidu and FAW Hongqi joined hands to create the first L4-level autonomous passenger car front-loading production line, developed China's first batch of front-loading mass-produced unmanned vehicles, and ran at the forefront of the industry.
In June last year, Baidu Apollo and BAIC Jihu jointly released the fifth generation of shared unmanned vehicle Apollo Moon, at the same time, Baidu also cooperated with WM and GAC Aeon for Apollo Moon, gradually building Apollo Moon into a product line.
In fact, at the beginning of the development and design, Apollo Moon is a front-loading mass production vehicle for automatic driving, so it can help the automatic driving software and hardware system to get on the car through standardized processes, greatly shortening the time of large-scale production of vehicles and reducing manufacturing costs.
It is understood that the cost of an Apollo Moon is only 480,000 yuan, which is one-third of the average cost of the industry's L4 level autonomous driving models, which greatly reduces the operating costs of the fleet and becomes a new milestone for the real large-scale commercialization of unmanned vehicles.
In addition, Baidu Apollo has also launched solutions for car companies such as ANP, which helps them further collect data to form a closed loop. Waymo, on the other hand, has only cooperated with Chrysler on front-loading production vehicles, and it also lacks car company solutions for more scenarios such as ANP.
Lower bicycle costs, closer cooperation with car companies, and the introduction of more diversified scenario car company solutions have enabled Baidu Apollo to slowly surpass Waymo in cost control, large-scale operation, and commercialization capabilities, and walk in the forefront of the industry.
Robotaxi's three major obstacles: Baidu has entered the third level
Dai Yifan of Tsinghua Automobile Suzhou Research Institute believes that there are three major "roadblocks" in front of Robotaxi today:
The first is technology, although the problems that need to be solved by autonomous driving technology have been reduced, from exposing a technical problem every few kilometers traveled in the past to one or several key problems that may need to travel tens of thousands of kilometers in the future, but these problems still require a lot of time and effort to solve, especially some edge scenarios;
Secondly, the construction of urban roads in mainland China is very fast, which brings no small challenge to the data collection and update of high-precision maps;
Third, achieving Robotaxi's profitability is undoubtedly an important issue, but in fact, even the operational efficiency and cost of unmanned distribution have not reached the level of manual distribution, let alone Robotaxi.
Therefore, the polishing and verification of technology, the accumulation of operational experience, and the gradual reduction of costs will be the weapons for many autonomous driving players to open the next stage of large-scale commercial operations, and Baidu is undoubtedly full of confidence.
First of all, from a technical point of view, thanks to large-scale, high-quality road tests, Baidu can quickly simplify the complexity and turn the "long-tail scene" that demo believes to be an "ordinary scene" that the automatic driving system can easily deal with.
On April 26, World Intellectual Property Day, Baidu released the "Baidu Artificial Intelligence Patent White Paper 2022" to take stock of various patent achievements. Among them, in 2021, Baidu ranked first in the world in the number of deep learning patent applications and the number of autonomous driving patent families.
At present, Baidu Apollo has formed a three-layer safety system of bicycle intelligence, monitoring redundancy, parallel driving and a complete travel service system for Robotaxi.
In terms of the unmanned test and verification system, Baidu Apollo can repair more than 90% of the new collision problems every six months based on the amplification and verification capabilities of its billion kilometer simulation mileage, form a capability boundary assessment for the scene, and focus on the repair level verification of safety problems.
At the operational level, like ride-hailing, Robotaxi's business model will also clearly show the characteristics of bilateral effects.
Taking online ride-hailing as an example, the more online ride-hailing drivers, the better the user's taxi experience, the more users enter the taxi, and the more they enter and join the driver, and finally achieve the growth of flywheel under the bilateral market effect.
On March 26, the radish run platform officially landed in Wuzhen, and as of now, the platform has covered 9 cities; among them, Beijing, Chongqing and Yangquan have all opened official fee operations.
In addition, the value of the bilateral platform lies in the interaction between the demand side and the supply side, which requires Robotaxi to continuously improve its user experience.
To this end, under the premise of ensuring driving safety, Radish Fast Run optimizes the user's riding experience through technologies such as no sharp braking and advance prediction; for example, through real-time data updates to improve fleet scheduling efficiency and reduce passenger waiting time.
Among them, in the first half of 2021, the average waiting time of Baidu's autonomous driving service platform for cars is 4.6 minutes, and the average 42 seconds will be ordered, and the average 3.9 minutes can be picked up, fully achieving the rapid response of users' car calls.
In the past few years, Baidu Apollo has walked through three mountains called technology upgrading, cost control, and expansion of operations, and now it has crossed the third hurdle, and the sword blade points directly to the word "profit".
According to Baidu CEO Robin Li's plan, in 2025, Baidu will cover Robotaxi services in 65 cities across the country and achieve profitability as soon as 2025.
He believes that when there are 50 million orders per day, the cost of Robotaxi will be 1/5 of what it is now, when the autonomous driving industry will enter the full commercial stage.
END
The financing boom is back! Pick up 15 smart driving companies that can be expected in the future
This year, the executive personnel of the Chinese and foreign intelligent driving circles was greatly earthquake