The autonomous driving industry is moving forward in a twists and turns.
On the one hand, after a cycle of technology and market explosions, the commercialization problems facing autonomous driving have made capital lose confidence. When the "burning money" game faced the constraints of cost and funds, many entrants fell before dawn one after another.
For example, in October 2022, Amazon, one of the world's first companies to deploy autonomous delivery business, announced that it would stop developing the "autonomous delivery vehicle" project Scout because it was always difficult to meet user needs and pain points in some aspects; in the same month, autonomous driving startup Argo AI announced the end of 7 years of operations.
On the other hand, as the unmanned process of testing and demonstration applications of autonomous driving technology accelerates in my country, management regulations in Shenzhen, Shanghai, Wuhan and other places have been issued one after another, which has also released a positive signal of national support and guidance for the compliant application of autonomous vehicles.
In July this year, the Ministry of Industry and Information Technology clearly supported the commercial application of autonomous driving at L3 and above. Some people in the industry said that this means that high-level autonomous driving will be paved by the previous regulations, select pilots, and officially move towards actual implementation.
The development of foreign autonomous driving is more radical. On August 10 this year, the California Public Utilities Commission (CPUC) approved Cruise and Waymo to provide all-weather (7 days a week, 24 hours a day) self-driving taxi (RoboTaxi) toll services in San Francisco.
"Autonomous driving represented by Robotaxi is a long track, and it is normal to experience several rounds of peaks and troughs. Friends who have made good time before the real industry outbreak period comes, the key is not to fall before dawn." Saiko Intelligent Dr. Yu Qiankun, the technical leader of Robotaxi , said.
As an important mission and support for SAIC Group's layout of "new four modernizations" transformation and innovation, SAIC AI LAB is responsible for the research and development and commercialization of high-level autonomous driving technology, and will officially undertake the Robotaxi project in 2021.
In the same year, relying on multi-dimensional technology accumulation and project accumulation in smart cockpits, smart driving, etc., the "SAIC Robotaxi" developed by SAIC Intelligent was launched in Jiading, Shanghai, and Suzhou; in 2022, Robotaxi 2.0 will be launched in Lingang, Shanghai, and Qianhai, Shenzhen Start the operation, realize the mass production of L4 level automatic driving, and launch RoboMPV and "full driverless" vehicles at the same time.
On the first anniversary of its landing in Lingang, Saike Intelligent released a series of results. In 2023, SECO Intelligence was awarded the first batch of national driverless intelligent connected car road test licenses based on legislation and the first batch of Shanghai smart connected rental demonstration operation qualifications. In the first half of the year, the operation scope of Saiko Intelligent Robotaxi has achieved full coverage of 68 square kilometers in the main urban area.
Preparing for the outbreak of autonomous driving in 2025
It can be said that autonomous driving has passed the "demo" stage.
"2020-2025 is the sprint stage of autonomous driving. It is expected that the industry will explode by 2025 or 2026. Especially since 2023, California has allowed Robotaxi to run around the clock, and all players are investing more. The fleet sprint is truly unmanned shared travel." Yu Qiankun said.
And firmly believing in the correctness and imminent dawn of autonomous driving, Seco Intelligence is trying to break the commercialization problem.
At the press conference to celebrate the first anniversary of its landing in Lingang, Yu Qiankun introduced the latest generation of fully independently developed driverless technology architecture of SECO Intelligent. One of its highlights is the adoption of multiple heterogeneous redundant autonomous driving software and hardware architecture to support safe autonomous driving with ultra-large computing power.
First, multiple redundancies in sensing are achieved. The latest perception technology architecture includes multiple solid-state lidar, peripheral view cameras, surround view cameras, millimeter wave radar and ultrasonic radar.
By leveraging the strengths of all sensors to offset their weaknesses, such as the strong depth perception of lidar and the strong visual perception of texture information, it is possible to achieve perception covering short, long and medium distances.
The second is to set up four layers of intelligent driving protection in the design of the intelligent driving system. The first layer is the main system of L4 intelligent driving, running on a high-performance HPC with a computing power of more than 1000Tops; the second layer is the first-level auxiliary intelligent driving system, which is implemented through a set of NVIDIA dual Xavier controllers with 64TOPS computing power. Safe shutdown when the main system fails; a 6-core ASIL-D MCU is built on the third layer. As a safety island for the entire system, it can still perform minimum emergency shutdown and other functions when the first two layers fail.
At the same time, the lowest level is also equipped with a remote driving control system developed by SECO Intelligent , which serves as the fourth layer of protection for the entire intelligent driving system. In the scenario where the self-driving vehicle cannot pass safely or the self-driving vehicle actively sends out a takeover request, it can help the vehicle get out of trouble or drive on a specific route by remotely issuing instructions or taking over the vehicle, so as to prepare for the realization of truly unmanned automatic driving.
It is worth mentioning that the architecture fully adopts the BEV Transformer-based fusion perception solution to predict space occupancy, grid semantics, and motion status, and uniformly express various irregular obstacles and three-dimensional drivable spaces from the perspective of BEV, greatly It greatly reduces the missed detection rate and calmly copes with the changing driving environment.
"The basic application idea of BEV has been unified in the industry, but the key to doing it well is to deeply integrate the existing technical system with business scenarios; On the far side, focus on whether the obstacles on the driving trajectory are occupied." Yu Qiankun said.
According to reports, Seco Intelligence has implemented a perception framework of front fusion + rear fusion and full recall based on BEV. It specializes in special processing and identification of areas that may affect the driving path, cooperates with fusion positioning, and uses 5G edge cloud technology for remote driving control. The system realizes the deep adaptation of BEV and Robotaxi.
In addition, Saike Intelligent has also built a data factory to realize the unified management of vehicle data through the vehicle cloud channel, provide data development platform, simulation test platform, vehicle problem management platform and business data tool chain and other related services, and a full set of closed-loop automation tool chain to help Rapid algorithm iteration drives continuous innovation in mass-production car-level autonomous driving solutions.
No matter in terms of software or hardware, Saike Intelligent has already made great efforts to attack Robotaxi, and no one has commercialized it.
Opening a new paradigm for commercialization
In the second half of the competition, the commercialization of L4 Robotaxi is one of the most concerning issues in the industry, but there are still problems to be solved in terms of technology, cost, market and supervision.
For example, the commercial model of Robotaxi must rely on lower vehicle costs and operating costs. However, the cost of refitting key hardware such as software, computing chips, and lidar for autonomous driving is still high, making it difficult to support the large-scale implementation of Robotaxi.
As for the existing pain points of Robotaxi, the competitive advantage of Saike Intelligent, which is backed by SAIC, is particularly obvious, and it has opened a new paradigm of commercialization.
On the one hand, as a project led by SAIC, SECO Intelligence is more closely integrated with the manufacturing end of traditional car manufacturers. By implementing Robotaxi on SAIC's mass-produced vehicles, it has also won the opportunity to scale up to a certain extent.
It is true that unlike L4 autonomous driving companies, which generally use rear-end modified vehicles for Robotaxi, SAIC Intelligent directly imports demand into the production line of SAIC Passenger Vehicle's production base in Lingang, and has established a fleet on the production line. This results in a higher level of vehicle consistency and process integrity while achieving lower retrofit costs.
On the other hand, SECO Intelligence is working with partners to explore the "new city model" for Robotaxi operations in Lingang, Shanghai, strengthening infrastructure construction such as communication terminals, road terminals, and cloud terminals to form vehicle-road collaboration to achieve a closed-loop commercialization of Robotaxi.
It is reported that since the fourth anniversary of the establishment of the Lingang New Area, it has fully implemented the national transportation power and the development strategy of smart cars, giving full play to its institutional advantages, technological advantages, scene advantages and industrial advantages, and building policy systems, basic elements and application scenarios. In terms of innovation, a good foundation has been formed to actively promote the construction of the intelligent connected vehicle innovation leading area in Lingang New Area.
For Robotaxi, the advancement of intelligent connected vehicle-related technologies such as vehicle-road collaboration V2X means that the safety pressure that bicycle intelligence needs to bear can be allocated to the roadside. While reducing the cost of vehicle-end modifications, it can also rely on roadside intelligent equipment. Assistance to achieve vehicle-side intelligence.
In addition, Saike Intelligent also continues to explore more commercial potential of Robotaxi by expanding the cooperation in the ecological circle in Lingang.
For example, at the results conference site, the public beta test of Seco Intelligent Driverless Vehicle was launched, and a strategic cooperation framework agreement was signed with Lingang Jinjiang Site to explore sustainability, replicability, and promotion in scenarios such as travel, business travel, and life . autonomous driving business model.
Xike's smart car-hailing system "Xiao Ke Lai Lai" was also launched on August 26. By booking a car through the mobile app, you can summon a self-driving Robotaxi to pick you up for free. In the Lingang New Area, 29 autonomous driving points include campuses, commercial areas, residential areas, scenic spots, transportation hubs and other core areas.
"This year, Seco Intelligence will increase its investment in the fleet in Lingang. Currently, we have nearly 30 vehicles in operation. It is expected to reach more than 50 vehicles by the third quarter, and we will target a fleet of 100 vehicles by the end of the year." Yu Qiankun said .
It can be said that as we enter the second half of the autonomous driving competition, the development blueprint of SECCO Intelligence has emerged, which is to explore a new and complete commercial implementation model and truly realize the large-scale operation of Robotaxi.