Alibaba Cloud's "big computing power" supports Inceptio's autonomous driving simulation to speed up 20 times

In an interview with the media on May 24, Inceptio said that in the field of data intelligence systems, the company has carried out in-depth cooperation with Alibaba Cloud. Based on cloud computing, Inceptio has built an industry-leading high-concurrency and high-elasticity data storage. , computing and scheduling infrastructure platform to accelerate autonomous driving simulation and reduce computing resource consumption.

According to statistics, the current market size of trunk logistics in my country is nearly 4 trillion yuan, with huge potential. However, the low operating efficiency and high cost of large truck fleets are the pain points of the industry. Founded in 2018, Inceptio Technology solves the above-mentioned problems in trunk logistics through autonomous driving technology and operations.

The development, testing, and operation of autonomous driving will generate massive amounts of data. Roughly calculated, if a road test vehicle is fully collected, it will generate more than 2TB of data a day. As mass-produced vehicles continue to be delivered on a large scale, the amount of data will grow exponentially. The application of these real scene data from trunk transportation can greatly reduce the R&D investment in autonomous driving and improve R&D efficiency. "Nugget" in massive data requires "big computing power".

Take Inceptio's self-developed deep neural network model for Xuanyuan autonomous driving system as an example. Model development relies on an advanced, elastic and stable cloud AI training environment. The perception models and decision-making algorithms necessary for autonomous driving also require a large amount of data for continuous iterative updates. Alibaba Cloud not only provided Inceptio with a cloud training environment, but also opened hundreds of cloud servers with the most advanced GPU accelerators to accelerate the AI computing process, thereby shortening the product iteration cycle.

According to reports, in order to accelerate the research and development of AI algorithms, Inceptio and the Alibaba Cloud Container ACK team jointly explored cloud-native AI technology and further optimized the development efficiency of AI algorithms through a flexible customized and on-demand containerized Jupyter notebook development environment.

Under the separation architecture of computing and storage, Inceptio uses Fluid cloud-native data orchestration and acceleration to greatly optimize data access performance and improve resource utilization. Through the container ACK Serverless ECI elasticity and cloud-native data lake solution, the high-concurrency and low-latency large-scale elastic resource supply of Inceptio simulation experiments is satisfied, which makes the automatic driving simulation speed up by more than 20 times, and the consumption of computing resources is reduced by about 30%.

In the future, the two parties will deepen business cooperation and help Inceptio build a unified capacity operation management service platform serving the trunk transportation logistics system. Through self-developed full-stack L3 and L4 autonomous driving technologies, Inceptio provides logistics customers with a new generation of TaaS (Transportation-as-a-Service) freight network that is safer, more efficient and cost-effective. By the end of April 2022, the self-driving commercial operation mileage of smart heavy trucks equipped with the Inceptio Xuanyuan system has accumulated more than 2 million kilometers.