New Business
Autonomous Driving Cloud Service: A Powerful Tool for Developers

During the Shanghai International Automobile Industry Exhibition from April 16 to 25, the HUAWEI CLOUD Autonomous Driving Cloud Service (Octopus) made its big debut. Running on HUAWEI CLOUD, Octopus provides driving data, training, and simulation services that help developers more rapidly develop autonomous driving products and work together to explore today's new age of intelligent connected vehicles.

Artificial intelligence is bringing new vitality to the automobile industry, and the in-depth application of automatic driving technologies will create new opportunities for industry growth. The three services that Octopus provides, the data, training, and simulation, cover the entire life cycle of autonomous driving, enabling developers to more efficiently develop autonomous driving products. 

Driving Data PlatformOctopus processes different types of data from vehicle-mounted sensors and replays data from vehicle-mounted radar cameras. The platform can be equipped with functions such as petabytes of storage, interactive big data query, and massive data governance.

Model Training Platform

The data collected from sensors is used to train driving models. As new data is collected and fed into it, the accuracy is improved and as the model gets better, the driving gets safer. The platform provides software and hardware acceleration, which greatly shortens the time needed to train the models. Model training is made more efficient.

Driving Simulation Platform

Driving models built on a foundation of petabytes of driving data provide the basis for driving simulations. Octopus then adds scenario library management and an evaluation system to ensure that the vehicle drives safely and follows the rules of the road. Octopus also ensures that the vehicle's behavior can tracked, evaluated, and quickly integrated into advanced driver assistance systems (ADAS).

Solving Six Pain Points

There are a number of technologies where further research is critical to the mass production of self-driving cars. Octopus can attack industry pain points from multiple angles. When it comes to breaking through current limitations, it provides six major advantages:

-  Full Hosting

Octopus works right, out of the box. Users do not need to set up a complex automatic driving system from scratch. They are free to focus on core competencies like developing automatic driving algorithms and data labeling. It frees up developers to more quickly develop automatic driving services to keep up with the market.

-  Massive Data Management

A test vehicle is expected to generate 64TB data every day. The platform can store petabytes of data and can search hundreds of millions of records in just seconds.

-  Hardware and Software Acceleration

Model training, and simulation require tremendous GPU resources. The Octopus platform provides the compute power needed for millions of kilometers of simulations and model training every day. Meanwhile, Huawei's self-developed Ascend AI chips and AI frameworks greatly improve training efficiency.

-  Automatic Data Labeling

A drive test vehicle needs to label nearly 1 million pictures every day. Automating this process reduces the manpower needed by more than 70%. 

-  Driving Simulations

With more than 20,000 scenarios, Octopus covers almost all of the possibilities across six major scenarios, scenarios like smart driving, active security, and various dangerous driving scenarios.

-  Seamless connect to the cloud

Octopus connects seamlessly with a mobile data center. The vehicle-mounted hardware and the ADAS system work hand-in-hand producing vehicle-cloud synergy.