WeRide unveils WeRide Master Platform (WMP) as the key technology to mass production of Level 4 autonomous vehicles
The year of 2020 witnessed two most important milestones in the Autonomous Driving (AD) industry achieved by WeRide, China’s leading L4 autonomous driving mobility company. In April 2020, the company unveiled a fleet of 100 AD vehicles, retrofitted from mass produced vehicles. Later in November 2020, WeRide rolled out its Mini Robobuses produced directly on the OEM assembly lines embedded with WeRide’s L4 AD hardware and software solutions. These fully driverless minibuses have been deployed for testing on public roads in Guangzhou, Nanjing and Zhengzhou.
How can WeRide, a company founded less than 4 years ago, achieve such as a rapid progress in manufacturing Level 4 autonomous vehicles? The key is WMP (WeRide Master Platform), a core AD technology platform that the company has recently unveiled.
1. Sensor Integration Solution: modular, configurable, adaptive
Features: Not limited to certain vehicle models
Starting from Lincoln MKZ as a widely used preliminary model, AD vehicles have now been expanded to a wide range of OEM brands and models. Does any addition of a new model require a set of brand-new sensor suite design? Does the suite design for Robotaxi and Robobus have to be completely different? How to maintain the same Level 4 AD performance on different models in a cost-effective manner?
WeRide has pioneered a unique modular design based on self-developed Lidar Camera System (LCS). The customized unit modules are fully configurable. Simply by changing the number of sensors, different types of sensor suites could be rapidly manufactured and adapted to different vehicle models.
WeRide’s LCS sensor integration solution allows parts and components to be fully standardized, which makes supply chain management efficient and controllable. The cost of designing, assembling and maintaining these sensor suites become much lower while the efficiency is optimized.
Since the relative location of individual sensors remain unchanged within unit modules, they could share the same perception algorithm, saving a lot of time to configure the hardware and software involved.
The highly integrated LCS also consists of large number of sensors, ensuring both adaptation efficiency and detection accuracy.
Modular, configurable and adaptive LCS makes WeRide ready for future fine tunes of sensor suite design to handle complex urban open roads and possible mass-produced new autonomous driving vehicle platforms.
2. WeRide ONE: A universal self-driving algorithm for comprehensive open urban roads
Features: Frame design
The biggest challenge for AD software algorithms is to not only address the variance in hardware but also in traffic scenarios. WeRide’s solution to this is a set of universal algorithms that can work on different hardware configurations and with different urban traffic scenarios. It is independent of underlying hardware configuration or ODDs (Operational Design Domains). Such a general solution provides much higher level of scalability when deploying in large areas.
The universal self-driving algorithm WeRide has developed is called WeRide ONE, and it could support self-driving under different scenarios and weather conditions on open urban roads.
So far, WeRide ONE has been verified extensively in both China and the US. The company’s AD vehicles have been undergoing testing and operation in eight cities, accumulating over four million kilometers of actual autonomous driving mileage in the past 3 years. WeRide ONE has gone through critical validation on city expressways and highways, in tunnels, ‘villages in the city’ and CBDs (Central Business District), meeting the needs of different modes of transportation.
3. Automatic Big Data Platform: Supporting large-scale commercialization in multiple cities
Features: Automatic data collection, processing and deployment
What is the biggest challenge posed by a self-driving fleet with over 100 vehicles (of numerous models) deployed in eight cities around the world? The answer is data. A highly efficient big data platform is essential in meeting the urgent needs of collecting, processing and deploying large amounts of data in real time; as well as improving the AD system through its deep-learning model training.
With robust local and global data centers, WeRide has built an automatic big data platform that supports simultaneous testing and operation of large-scale fleets in multiple cities, realizing an efficient and closed-loop data processing.
On this platform, hundreds of TB data are collected and uploaded to local data centers by WeRide self-driving fleets operating in various cities every day. These massive data are processed in global data center efficiently for key scenario collection, HD map generation, deep-learning model training, as well as scenario simulation and validation. WeRide’s cloud-based simulation cluster generates over 500,000 kilometers of simulated mileage on a daily basis. Tens of thousands of complex urban road scenarios and their variations are available, each with thousands of obstacles generated instantaneously from algorithms or from real-world data.
With this automatic big data platform, a local team made up of project managers, product engineers and safety drivers will be sufficient to conduct regular testing and operation of large-scale fleets on the ground, and meanwhile algorithm engineers can deal with most of the problems remotely. This platform can automatically complete data triage and diagnosis of problems as well as system update within 24 hours without the constraints of geographic locations, achieving much lower cost and higher efficiency.
4. Fully Safety Redundancy Design: Delivering the best AD safety solution
When it comes to the design of safety redundancy systems, WeRide adopts a four-pronged approach, which covers 1) sensors 2) compute units 3) DBW (drive-by-wire) and 4) network connection. In addition to safety redundancy systems, the company has also worked with OEMs and designed a systematic process for vehicle fault diagnosis and safety verification.
Over the past four years, WeRide has invested significant amount of resources into the R&D of the WeRide Master Platform, which is the backbone for all of its hardware & software solutions as well as data infrastructure. The company has been leveraging its large-scale publicly accessible Robotaxi services for verifying and improving the stability and safety of its Level 4 AD technologies. Now, the WMP is in turns enabling WeRide, a company originally focused on developing in Guangzhou and solely on Robotaxis, to rapidly expand its product capabilities to Robobuses and scale to new cities at an unprecedented speed.
“Autonomous driving is the perfect showcase for AI. It is the ultimate integration of multiple technologies and innovations. I believe that for an AD company, what matters the most is developing the core technologies. Supported by WMP, WeRide was able to deliver mass-production Level 4 AD vehicles through retrofitting in 2020. Into 2021, we have now made the significant leap into mass production directly on the OEM assembly lines. The new Mini Robobus that we have jointly designed, developed and manufactured with our strategic partner YuTong Group marks a truly important step forward for the entire industry.” Said Tony Han, Founder and CEO of WeRide. “Our mission at WeRide is to realize fully autonomous driving service while maintaining the industry’s highest standards on safety, stability and reliability.”
Cases:
1Teaming with Yutong Group and producing front-loaded self-driving minibuses
In December 2020, the self-driving Mini Robobus jointly created by WeRide and Yutong Group made its debut in Guangzhou. This is the first AD minibus in China designed specifically for open urban roads. It’s an innovative model without any steering wheel, accelerator or brake, but equipped with WeRide’s self-developed full stack software and hardware solutions, and capable of cruising under complex urban traffic conditions.
In just two months, the two companies succeeded in designing Mini Robobus 2.0 and its sensor suite. Since last November, regular testing of the Mini Robobus has been conducting in Zhengzhou, Guangzhou and Nanjing successively. The Mini Robobus has also been launched into the market with demand-based production and has already won hundreds of orders.
2Dongfeng Motor-led autonomous driving project
In the autonomous driving project organized by Dongfeng Motor, WeRide completed the retrofit of AD hardware on Dongfeng Fengshen E70, a brand-new vehicle mode for self-driving, within just two months. The configuration of hardware and software was finished within just one week. WeRide’s trial operation was conducted in the Development Zone with 22 major pick-up and drop-off points. WeRide’s HD maps cove 198-kilometer roads, making WeRide stand out from the many AD companies in the project. Currently, AD vehicles in this project are available for public reservation. WeRide, on its part, is working with Dongfeng Motor and conducting Robotaxi operation in core urban areas in Wuhan, delivering the benefits of autonomous driving to the local residents.
WMP is essential to mass production of Level 4 autonomous vehicles on OEM assembly lines. It has been constantly tested by WeRide to achieve a higher level of flexibility, efficiency, safety and stability. As the commercial use of fully driverless vehicles accelerates, WeRide will continue to enhance WMP through the testing and operation of its Robotaxis and Mini Robobuses in multiple cities.
