Author|Wu Jingjing
Editor|Zhao Jian
On April 13, Didi Autonomous Driving held an open day and announced the current technical service progress, new products and innovative businesses, focusing on four main aspects:
· In terms of products, the first futuristic service concept car, DiDi Neuron, was released;
· In terms of autonomous driving services, the first autonomous driving operation and maintenance center was released;
· In terms of mass production, it announced that the first autonomous driving production model is planned to be connected to Didi's shared travel network in 2025, and released two core hardware for autonomous driving, "Beiyao Beta" lidar and three-domain converged computing platform "Orca";
· In the business segment, the innovative business Didi autonomous driving cargo KargoBot was unveiled.
Zhang Bo, CTO of Didi Chuxing and CEO of Didi Autonomous Driving, said at the event that at present, the full-stack core technology of Didi Autonomous Driving L4 autonomous driving has achieved autonomy and controllability: "With the accumulation of core technology research and development in the past six or seven years, we have mastered the full-stack core technology of L4 autonomous driving, including perception, prediction, decision-making, control, large-scale simulation systems, and large-scale machine learning platforms. ”
At present, Didi's autonomous driving travel services have been launched in Guangzhou and Shanghai, and Didi users can experience related services in the local autonomous driving demonstration operation area.
"We hope to work with upstream and downstream partners in the industry chain to deeply integrate software, hardware, services, operations and other aspects to innovate the next generation of autonomous driving mobility services." Zhang Bo said.
1. The new autonomous driving concept car DiDi Neuron was unveiled
At the event, the most eye-catching was the newly launched autonomous driving concept car DiDi Neuron.
"This is a car from the future." Meng Xing, COO of Didi Autopilot, said.
At 4.5 meters long, the DiDi Neuron is about the length of a midsize SUV, but because the wheelbase reaches 3.2, it exceeds the standard for a midsize SUV, so it looks as big as a large SUV. The cabin has increased the space available to passengers by 50%.
In terms of interaction design, Neuron is equipped with a large-screen interactive system, and replaces the traditional lights with five intelligent interactive screens, aiming to transmit the automatic driving situation to the outside world through the window, so as to protect both passengers in the car and people outside the car. In addition, there are many flexible screens in the car, which can interact with people, and can also carry out entertainment activities such as audio and video playback, while also meeting the needs of office meetings anytime, anywhere.
In terms of door design, Neuron uses the design experience of D1 to install an electric sliding door, passengers can only get off on the side near the road, and a top wing door is also designed on the car, so that people with a height of 2 meters can get on and off without lowering their heads.
Another highlight is that the Neuron is equipped with a robotic arm that acts as a "butler" function, which can help carry luggage to the trunk and deliver water to the passengers in the seats.
In addition, in terms of hardware, Neuron is equipped with a total of 21 cameras, 8 lidars, and 6 millimeter-wave radars. All sensors adopt an integrated, modular design, "each set has two lidar and three cameras, so the modular design greatly reduces our production and maintenance costs." Meng Xing said.
In terms of power system, Neuron combines in-wheel motors and unmanned ride-hailing, and the four tires can be flexibly steered, which can realize functions such as smaller radius turns and diagonal driving, even in difficult scenarios such as turning around on narrow roads and wrong driving.
"The Neuron concept car answers what kind of robotaxi service we need." Meng Xing said.
2. New progress of autonomous driving services and mass production
In addition to the unveiling of the new car of the future concept, Didi Autonomous Driving also announced the latest progress of the current autonomous driving business and further plans for mass production.
Meng Xing called the new stage of Didi's autonomous driving the autonomous driving service stage: "The past three years are our first stage, that is, the safety phase of autonomous driving, and now on the basis of safety, we have entered a new stage of autonomous driving service, starting to think more about service, focusing on the experience of passengers and the experience of operation." ”
In order to serve more customers, in terms of coverage area, Didi Robotaxi's autonomous driving coverage area in Shanghai has expanded by 7 times in the past three years, and the latest version of Didi Chuxing App can now achieve hybrid dispatch in designated areas in Shanghai and Guangzhou. That is, orders that start and end in the region, users can directly see the "autonomous driving" option on the Didi App and request a ride.
In this process, Didi faced a problem not only how to expand the map, but how to face a more complex real scene. For example, during peak hours, when the car is forced to squeeze into the left turn lane on the straight road, how to find a gap and avoid a large number of motorcycles and bicycles crossing back to the straight road; How can a place with a well-developed night market respond flexibly to situations where road conditions are very different during the day and at night.
To this end, Didi Autonomous Driving launched the "City Generalization Engine" to solve the challenges of traffic flow, interaction and long-tail scenarios. The "urban generalization engine" includes a series of Didi automatic driving technologies, such as in order to solve the traffic flow problem, Didi uses a perception large model that can output all road perception results at one time, greatly reducing delay, improving accuracy, and also using a pre-trained model that can predict traffic behavior more accurately.
These technologies are the unique scene capabilities and technical capabilities that Didi autonomous driving has accumulated when entering the deep water area of the city in the past two years.
In order to further improve the user experience, Didi Autonomous Driving has also introduced unlimited parking technology to turn all sections into stations, and has also launched a 24-hour uninterrupted autonomous driving service.
Supporting these new actions, Didi Autonomous Driving has taken another step in operation and created the first autonomous driving automatic operation and maintenance center - Huiju Port, to automate and meet the needs of vehicles in the operation process. For example, automatic departure, order acceptance, return to Hong Kong, the robot can also intelligently flush, charge, detect, repair, maintain, park, etc. the vehicle, the whole process automation rate has reached 90%. Among them, the automation equipment includes robotic arms, AGV mobile robots, etc.
The latest data is that Didi's autonomous driving delivery time has reached 1.19 times that of ride-hailing, significantly shortening the time from 1.35 times in 2022, and at the same time, it has reached 93% of the efficiency of online ride-hailing this year in terms of hourly units, a significant improvement from 81% last year.
At present, Didi autonomous driving has been in continuous operation for 1,020 days, and in Shanghai Jiading and Guangzhou Huadu, users can already experience Didi's autonomous driving travel services in Didi autonomous driving demonstration operation areas.
For the next step of autonomous driving commercialization services, Didi Autonomous Driving also announced the latest mass production actions: launched two core hardware for autonomous driving - "Beiyao Beta" lidar and three-domain fusion computing platform "Orca".
Among them, Beiyao Beta is the first 2K image-level high-precision lidar in China, which can provide a large field of view of 3 million high point frequencies per second, ultra-512 lines, 120° horizontally and 25.6° longitudinally, and can support resolution adjustment to adapt to various scenarios.
Orca Killer Whale is the industry's first mass-produced three-domain converged computing platform developed by Didi Autonomous Driving, equipped with an on-board super central computing unit (SCCU), which integrates the three domains of intelligent driving, intelligent cockpit and networked domain. "Compared with the previous generation of computing platforms, the overall cost of killer whale products has been saved by 88%, the space volume of the whole vehicle has been reduced by 74%, and the core components and wiring harnesses have been optimized." Meng Xing said.
Didi's autonomous driving is one step closer to mass production.
3. New business exploration: self-driving cargo KargoBot
Unexpectedly at the scene was the unveiling of Didi's self-driving cargo KargoBot. Wei Junqing, head of Didi's autonomous driving innovation business, announced at the event that it has expanded its Level 4 autonomous passenger car solution to trunk logistics and freight transportation.
This is the first innovative business incubated by Didi autonomous driving.
The launch of KargoBot represents Didi Auto's targeting of the self-driving truck market beyond Robotaxi.
This is a new type of scenario that Didi autonomous driving began to explore in 2021. The reasons for choosing this scenario are the same as other participants who are optimistic about this track: "Freight business accounts for 5% of the country's GDP, but there is already a shortage of more than 8 million truck drivers in the country. ”
Didi autonomous driving truck chose bulk commodity transportation as the landing scenario and proposed a product solution for unmanned bulk commodity operation - HDS (Hybrid Driverless Solution) hybrid unmanned solution, which provides customers with end-to-end logistics and warehouse-to-warehouse transportation in the form of a manned L2 with assisted driving function and a fleet of multiple L4 fully unmanned vehicles.
This is a market that many people are optimistic about, but it is also facing practical problems. In the article "Self-driving Trucks: A Shortcut Not So Easy to Go", "Koshi Lightyear" talks about the opportunities and challenges of this track. In the traditional truck fleet, the largest consumption of a heavy truck comes from fuel consumption, tolls and labor costs, accounting for 30%, 30% and 20% respectively, and the cost of buying a car is spread evenly to only 10% per year. If a self-driving truck can bear labor costs and reduce fuel consumption, it can save 50% of the cost during vehicle operation and improve the efficiency and operation time of the vehicle. Under ideal conditions, such technology would have been paid for by the logistics fleet, but it still faces the problems of low revenue, high losses and long cycles.
Didi's autonomous driving innovation business has made breakthroughs in technology and scenarios, and has invested in large-scale autonomous driving commercial operations. Wei Junqing mentioned that there were already more than 50 self-driving trucks at the end of 2022, and today the number has exceeded 100, and the cumulative autonomous driving revenue will reach 50 million yuan in 2022, and the number will exceed 100 million at the end of the first quarter of 2023. In addition, KargoBot has hybrid intelligence in terms of unmanned, so it has been verified last year for full-stack unmanned technology. "This year, we will officially put autonomous unmanned transportation into small-scale commercial operations." Wei Junqing said.
KargoBot's latest progress is that at present, it has landed in freight scenarios and cargo truck automation, and has obtained the test license of Beijing self-driving truck in three months and started the test of self-driving truck. In addition, its full-stack software, systems and AI have made more than 50 innovations, allowing Didi Robotaxi technology to be used by trucks.
4. From 200 to 1,000 people, Didi's new journey of autonomous driving
This is an appearance after two years, and it is also a concentrated display of the achievements of Didi autonomous driving since the formation of the team in 2016.
Since 2016, Didi has been ready for the long term. Zhang Bo, CEO of Didi Autonomous Driving, said at the event that at that time, after visiting the world's top scientists, entrepreneurs and engineers in the field of autonomous driving, he established a new understanding: "L4 autonomous driving technology will take about ten years (2016 to 2026) to gradually enter people's lives, not many people say two or three years of energy production, nor is it as far away as some people say." ”
During this period, Didi autopilot walked its own way. In 2019, after Didi's autonomous driving division was upgraded to an independent company, Zhang Bo, CTO of Didi Chuxing, concurrently served as the CEO of Didi Autonomous Driving, with a team of less than 200 people at that time, and now the team size has reached about 1,000 people.
Today, three years before the end of the first decade.
Didi also has a clearer path.
It was mentioned on the spot that there are four key dimensions of Didi Robotaxi development: localization, pre-installation mass production, cost reduction, and safety guarantee:
The first is localization, Didi automatic driving hopes to achieve automatic control of the supply chain, 90% of key components to achieve localization, which is also one of the key points that Didi automatic driving has been emphasizing before, in 2021 Didi automatic driving released the Gemini platform mentioned that 80% of the time has chosen the localization solution.
Front-loading mass production has always been one of the core goals of Didi Autopilot, and the company needs to ensure reliability and consistency, while Didi Autonomous Driving also hopes to reduce the cost of the next generation by 4-5 times and lay the foundation for scale.
In addition, safety guarantee is the foundation of all services, and Didi's autonomous driving planning is to form the core standard of next-generation vehicles through various redundancy, such as chassis redundancy, communication redundancy, power supply redundancy and other guarantees.
Along this line, the development focus of Didi autonomous driving is to do a good job in the operation and service of autonomous driving; At the same time, it can continue to iterate the software; And technology to empower ride-hailing platforms.
In the future, Didi Autonomous Driving will work with new energy vehicle companies to jointly define mass-produced driverless new energy ride-hailing vehicles, and cooperate in product definition, model platform, mass production, production and other links.
"It is hoped that the first car can enter Didi's network for service in 2025, and realize all-weather, large-scale hybrid dispatch." Meng Xing said.