Big Data and Engineering Platform
The development of self-driving system requires long-term iterative optimization and needs a lot of road test data as support. Therefore, to meet the requirements of continuous operation and iteration of self-driving system, Hirain has developed a cloud-based big data system. Relying on 5G network, this system can support multiple functions such as remote data collection, compression, transmission, analysis, playback, and seamless docking with algorithm development tools, which can serve customers at different stages from research and development to commercial operation.
Overview of the system
This system achieves a very fast data loop. The full data of the autonomous driving fleet, including driving condition data, command data, business data, original point cloud data of lidar and camera image data can be sent back to the engineering data platform in real time, and can complete signal analysis, event analysis and data archiving in a short time (10~15 minutes). Therefore, algorithm and software development engineers can immediately retrieve abnormal data for analysis and can also import the data into engineering development software for playback and simulation, to quickly locate problems, modify defects, generate new software versions, and use OTA technology to update the vehicle-side code.
Advantages of the system
- Capability of full data uploading (including CAN and Ethernet operating condition data, laser point cloud and the raw image)
- Efficient data compression algorithm to reduce the pressure of 4G/5G traffic
- Automated data analysis and event analysis to reduce the working pressure of engineers
- Support online data browsing and data exporting offline view
With the support of the big data and engineering platform, the self-driving development team, and the operation team of Hirain can greatly reduce onsite data collection workload and obtain the required data remotely and quickly for system optimization. This system helps Hirain and its port customers to achieve a closed data loop for the development and operation of self-driving systems.