Why Cloud-based Solutions are the Engine of Autonomous Driving
In the lead-up to the Shanghai Auto Show, we've interviewed Shuman Liu, Vice President of Tencent Intelligent Mobility, about the importance of cloud-based solutions in the development of autonomous driving (AD), some technical highlights and why Tencent is the right cloud technology partner for its rollout.
Shuman Liu, Vice President of Tencent Intelligent Mobility.
What conditions are required for the development of AD technology?
With its complicated road systems, mega cities, dense populations and unique traffic behaviors, China poses challenges for AD development. It also provides richer scenarios, plentiful data and greater numbers of rare and unexpected so-called “corner cases” essential for the safe and effective evolution of AD technology.
Add to that a unique Gen-z demographic that plans to purchase a car with Level2+ standards of driver assistance functions and you have ideal training conditions for AD. That’s why global original equipment manufacturers (OEMs) are establishing major research and development (R&D) teams in China and promoting those solutions worldwide.
How is Tencent meeting the AD and cloud needs of international companies?
Tencent has taken the lead and launched a one-stop Intelligent Vehicle Cloud platform that can meet technological requirements at every stage of the AD development process. The foundation of the platform is the company’s security and compliance capabilities, along with all the necessary toolchains.
Our advanced cloud data storage ecosystem addresses challenges around the huge data volume and high storage costs involved in the development of AD. Tencent’s flexible out-of-the-box, low-cost tools also ensure a smooth user experience. Meanwhile, end-to-end security and compliance in our cloud-to-vehicle integration architecture resolve complex compliance issues when entering a new market.
What are some key features and advantages of the platform?
As to the security and compliance service, the loop begins within the car, where all perception data will be encrypted and transferred onto the platform. Only desensitized and anonymous data will be available for use by the R&D team working in a fully secure and compliant data-closed loop.
Tencent has developed software-hardware integrated architecture to provide stable operation and low energy consumption for managing such large volumes of data required for simulations and modeling. The system is high-density, high-performance, reliable and can reduce storage costs by 50 percent compared to standard practice.
What’s more, Tencent provides high-performance solutions by leveraging its AI capabilities for large-scale AD training. This can reduce network congestion by 40 percent and total training costs by 25 percent.
In building an AI model, polishing individual functions is not enough, so we equipped the TI platform with services covering:
1. Data transfer, when storage and computation can be separated, resulting in a reduction of the transfer period by 80 percent.
2. Data processing, with an easy-to-use rich labeling database that can cut labeling costs by 70 percent.
3. Model training with a drag-and-drop function, model evaluation and post-processing, which can improve efficiency by 50 percent.
To meet the validation requirements of AD testing, we also provide TAD Sim, an advanced simulation program that integrates gaming technology and cloud capacity. Performance-wise, the program can realize 90 percent accuracy for perception models, vehicle dynamics and traffic.
TAD Sim provides more than three million scenarios in its data set and supports real-time simulation for vast city-scale areas. Under this complex simulation environment, TAD Sim can generate random corner cases to test the completeness of the AD algorithm and enrich the scenario database. Its capabilities mean that the safety and efficiency of testing will greatly improve.
Accurate maps are another important component of AD development. Tencent has launched a vehicle-map-cloud solution to provide regular on-demand map updates and location-based services that can be tailored to different OEMs at low cost and with high coverage.
Can you provide examples of how international companies are benefiting from Tencent’s cloud solutions?
We are partnering with more than 100 OEMs and AD development companies, who have tapped us for our open ecosystem, technology and solid infrastructure.
We have partnered with Bosch to provide them with worry-free R&D environment, allowing their team to focus on AD technology development with our full-stack services.
Our strategic partnership with Nio has expanded from AD and maps to areas such as building a hybrid cloud infrastructure to support their AD development and operation, and providing our TAI smart cockpit solution to explore new in-car possibilities.
As autonomous driving enters the commercialization stage, OEMs should prepare by setting up dedicated teams to handle issues and continuously enhance the customer experience. Customer satisfaction is the core competency for any vehicle and our Intelligent Vehicle Cloud platform provides the framework to deliver that for companies.