A few days ago, I used a navigation software to take a taxi. After selecting the “Quick car” option, I waited for about 5 minutes without anyone accepting my request. So I added more car types, selected “Premium car” and other ride-hailing platforms. Just less than 5 seconds after confirming the addition of car types, a driver accepted my request.
Soon, a Lexus RX450h pulled up in front of me. I was stunned for a moment, looked at the license plate and then glanced at the information in the mobile app before realizing that it was indeed the car that I had called for.
After fastening my seat belt, I didn’t open Weibo on my phone as usual, but stared blankly at the bustling night view outside the window. The retail price of a single RX450h for customers is more than 650,000 yuan. If this is purchased by a ride-hailing company, the average price per vehicle may be slightly lower. When the driver’s labor cost is added, the daily operating cost of this premium car is not low.
The RX450h, cost, and shared taxi reminded me of Pony.ai. If the driver in front of me is replaced by a robot, and some perception sensors are installed around the car body, it can be transformed into an unmanned vehicle. How much would I pay for the ride today? Will the riding experience be better? At least the “driver” won’t keep cursing by himself…
As a young person who has to buy a monthly pass just to ride a shared bike, I am always sensitive to travel costs. Even if unmanned taxis achieve large-scale commercial operations someday, if the riding price is ridiculously high, I believe no one will accept this mode of travel. Otherwise, what’s the point? Does it mean that there won’t even be anyone to help me carry my luggage anymore? Is it really feasible to spend millions of dollars like Waymo to build an unmanned vehicle for operation?
My “level” of thinking about this issue is still too superficial, so I “invited” some industry experts and “organized” a debate to discuss this issue.
A discussion on this scale is bound to attract the attention of the entire industry. Before the start time, the venue was already crowded. The familiar guests were chatting and catching up, and various reporters were recording the scene, afraid of missing any important information.
As the time approached, the guests took their seats one after another.
At this moment, the host of this historic debate, rushed to the center of the venue and tidied up his clothes. After clearing his throat, he said, “Thank you very much for taking time out of your busy schedules to participate in our 42 Debate and discuss the beautiful blueprint for the commercialization of Robotaxis.”Although you all are familiar with each other, please allow me to introduce the two most important debaters of today. Sitting on my left are members of the AutoX team: CEO Professor Jianxiong Xiao and Mr. Hai Deng, the CEO of AutoX China. I want to introduce Professor Xiao in particular: he is a well-known scholar in the field of deep learning in computer science, with a PhD from MIT and a professorship at Princeton University. After declining an offer from the Vice President of Autonomous Driving at Ford, he resolutely returned to China to start his own business. Welcome, Professor Xiao.
The venue immediately erupted in thunderous applause. Amid the applause, Jianxiong Xiao stood up slowly from his seat, waving to the audience and speaking in a not-very-standard Cantonese accent, “Thank you very much, thank you very much.”
Next, the host continued, “On my right are representatives from Baidu Apollo, including Baidu founder and CEO Robin Li. Today, Robin leads by example, a testament to how seriously he takes this debate. Welcome, Robin. Sitting next to Robin is Wei Dong, the Vice President of Baidu Intelligent Driving Group. Welcome, Mr. Wei.”
After introducing the guests, the host naturally explained the background and agenda of this discussion. The Robotaxi competition has reached its second half, and whoever can find a commercial path first will likely occupy a larger market share. As veterans of the industry, everyone present did not need an elaborate introduction. Let us get straight to the point and discuss the contradictions between the scalability and cost of Robotaxi. The floor is yours.
Upon seeing this topic, Robin adjusted the microphone and prepared to speak, “I believe that in terms of cost control, Baidu still has a strong voice. Apollo Project has been deeply rooted in the industry for many years, from the iteration of car models to multiple facets such as data from test mileage, the number of testing licenses obtained, and many more. We have a leading edge. As for the contradiction between scalability and cost, Mr. Wei, can you answer that?”
Caught off guard by the sudden cue, Wei Dong quickly adjusted his posture and replied, “Yes, we do have a say in cost control. Furthermore, in our Baidu, there is no contradiction between scalability and cost. The fifth generation Robotaxi we jointly developed with JiHu has reduced the cost to 480,000, successfully entering the price range of civilian cars.”
I will do the math for you. If a car of 480,000 RMB is put into operation for 5 years in the market, it would cost only 8,000 RMB per month, plus the costs for the operating personnel, which would be around 10,000 RMB per month in total. In a first-tier city, this price is similar to the price of manually operated ride-hailing services. If the price cannot be reduced, it will be difficult to expand. The difficulty of getting a ride is already a pain point for urban transportation, and if there is not enough supply, it will exacerbate the pain point and cause a vicious circle. Regarding this point, I would like to hear Professor Xiao’s thoughts. Your company is still doing addition. Can you explain your company’s philosophy?
Xiao Jianxiong had already anticipated the question and was well-prepared to answer. After clearing his throat, he began to explain his thoughts: First, I want to express my respect to Baidu Apollo. As one of the first companies in China to layout the autonomous driving industry, you are undoubtedly the pioneers. However, with regard to fully autonomous L5 vehicles, it is necessary to ensure that nothing goes wrong. Safety accidents have a huge impact on any company. No company can withstand it, not even Waymo or Uber. Our mission is to create autonomous transportation for the people, but when I say “for the people”, I refer to the end-users. They don’t care about how much the car is worth, they only care about two things: safety and cost.
Regarding safety, our fifth-generation Robotaxi Gen5 has a total of 50 sensing sensors, including 6 Lidars and 28 8-million-pixel cameras, which can sense 360 degrees without blind spots. The computing power of our XCU computing platform is also very high, with a total computing power of 2,200 TOPS. I do not think that piling up hardware is the wrong approach. I would never get on a 480,000 RMB autonomous vehicle. Furthermore, considering Moore’s Law in the field of computer hardware, our future costs can also be controlled at a very reasonable level.
His remarks instantly livened up the atmosphere in the conference. The guests whispered and exchanged views. At this point, Wei Dong couldn’t sit still and said with a slightly hurried tone, “Waymo’s autonomous vehicle based on Chrysler costs about 1 million dollars. Moreover, they have set a target of putting 20,000 autonomous vehicles into operation, which is difficult to achieve. Against such a backdrop, Waymo burns billions of dollars a year. As a startup company, you should find a path that suits you.”
Besides, Waymo has also made a profit from selling Lidars to others. Our approach is to bring L4 technology down to civilian cars. Our Apollo “ANP-Robotaxi” architecture has completed data full-cycle, and both legs are walking, empowering each other.The combination of punches was unexpected for Professor Xiao, as demonstrated by his unprepared response. When it comes to experience and business strategy, Baidu is indeed more comprehensive. The host quickly took over: Professor Xiao, we all know that AutoX has grand plans, and it’s remarkable that AutoX is the second company after Waymo to receive a full autonomous driving test license in California. How do you view the future of commercialization?
At this point, Deng Hai, who had been watching the debate, finally couldn’t sit still. He gently tapped Xiao Jianxiong’s arm, indicating that he could take over the question. Deng Hai continued: The application scenarios for unmanned vehicles are numerous. Perhaps everyone only focuses on our Gen5 Robotaxi being on the road, but our earliest venture in the United States was in unmanned robotic express delivery service, and our cooperation with fresh e-commerce company GrubMarket provided many users with the delivery of essential life necessities. This business still has a market, especially during the epidemic period.
In 2018, we signed a strategic cooperation agreement with ZTO Express to jointly develop unmanned delivery trucks. We aim to provide the “last mile” transportation solution for logistics and express delivery services in the park. We have also reached a strategic cooperation agreement with Meituan to provide unmanned meal delivery service. The number of takeaway orders and express parcels is enormous in China, and this scenario is also the fastest way for unmanned driving to achieve commercialization on a large scale. This is why we are also developing autonomous driving light trucks.
We don’t manufacture cars or operate them. We focus on technology and sell our electronic driver technology to car manufacturers, logistics companies, and taxi companies. We are also working with Dongfeng Motor to jointly develop unmanned trucks, and finally providing products to underground subways to complete unmanned logistics distribution services. Whether it’s from the perspective of Robotaxi or Robotruck, I am confident about our future commercialization strategy.
Our technology is applied directly to unmanned travel services, and in the process, we are also promoting the technology for civilian vehicles to develop advanced urban autonomous driving systems. I can see the beautiful blueprint of future travel from everyone’s statements. The host quickly intervened when the on-site discussion was still in the debating stage.
Of course, the contents of the debate above are fictitious, but they represent the two main directions of current Robotaxi development – cost reduction and increased business revenue.
At the same time, there is another voice that firmly believes that using Robotaxi as a commercial purpose is a wrong direction. Su Jing, the former chief architect of Huawei’s autonomous driving team, holds this view. In his opinion, the ride-hailing experience in China is already good enough, with a large number of cars available and low costs. China has already achieved Robotaxi, but the robot is human.From Su Jing’s perspective, the value of Robotaxi is not significant. It merely replaces human drivers who are least necessary to be replaced in shared travel, and autonomous driving for single vehicles neither improves traffic nor brings other associated value. So, what did Tesla do?
In fact, Tesla has long published its idea about Robotaxi, although it has been rarely mentioned recently. But, the words of “delayed delivery king” Musk will eventually become a reality someday in the future.
At Tesla Autonomy Day in 2019, Musk claimed that after achieving full self-driving (FSD), Tesla owners can join their Teslas to TESLA Network to share their vehicles. This project can bring about $30,000 in annual revenue for owners, but they have to bear the costs of electricity consumption, parking fees, and vehicle depreciation.
At first glance, this plan seems feasible and has a considerable market. However, one premise is that Tesla must not only achieve full autonomous driving, but also be willing to take responsibility for possible accidents.
Imagine our daily vehicle needs for commuting to work, which means that a car is mostly idle during its lifespan. If we use this period to send it out to make money, even if the vehicle owner can only get a 25%-30% cut, it is still a profitable business. For Tesla, this is also a business and value derived from autonomous driving research and development towards the final stage.
Although Robotaxi’s competition has entered the second half, few people have a clear view of the future situation. What we can do now is to keep testing on roads, accumulating data and experience. Nowadays, the trend for all companies is to remove safety supervisors, as unmanned tests are more meaningful. In terms of scaling, Baidu still has a leading position. Baidu plans to land in 30 cities, deploy 3,000 vehicles, serve 3 million users in the next three years, and explore more operational cities and travel scenarios.Some believe that Robotaxi could achieve lower costs and safer travels in the future. Others think that equipping the technology in cars before pursuing autonomous driving will pave the way for these unmanned travel companies, while some predict that acquisition is their ultimate fate. Looking back and ahead from the present day, the de-staffing of Robotaxi is undoubtedly a leap from 0 to 1, but the difficulties ahead still include policies, testing licenses and standards, technical maturity, and user support, among others.
The road ahead is long and winding, and the entire industry should work together to strive for progress.
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.