TUM’s autonomous shuttle, EDGAR, operates seamlessly with InoNet’s datalogger working efficiently in the background
As part of an autonomous driving project in Germany, the Technical University of Munich (TUM) is using the autonomous vehicle EDGAR (Excellence Driving GARching) as an autonomous Wiesn shuttle, which went into service late last year. This application places extreme demands on the technology and requires a solution that overcomes several complex technical challenges.
A key challenge is using artificial intelligence to process the large amounts of data generated during journeys, which requires a large number of suitable interfaces for seamless integration of sensors, including cameras, radar and lidar. Different interfaces, including CAN-FD, ethernet and USB, need to communicate with the GPU and the Raid controller via the CPU.
The data must also be stored and transferred to the hardware-in-the-loop system for later analysis. The data processed during the trip is used to visualize the detected objects – such as other vehicles, pedestrians, cyclists or traffic signs – on screens inside the vehicle. At the same time, the processed information must be transmitted to the engine control unit (ECU) to enable the autonomous vehicle to be controlled.
These requirements demand a hardware solution that is not only powerful and flexible, but also works reliably under the demanding operating conditions in the vehicle.
The solution
To meet the high project requirements, TUM relies on the powerful InoNet Mayflower-B17-LiQuid-vX in combination with the QuickTray, which is integrated in the trunk of the vehicle. This liquid-cooled hardware platform is specifically designed for demanding applications such as data acquisition and datalogging in ADAS and AD test applications, and is specifically tailored to the use case and customer requirements.
The Mayflower-B17-LiQuid-vX offers a combination of a high-performance CPU and the Nvidia RTX A6000 GPU, which enables the processing of large amounts of data thanks to its computing power. The efficient water cooling ensures reliable heat dissipation, even under demanding conditions. By integrating network cards, frame grabbers and bus measurement cards, the platform provides extensive connectivity and enables precise processing and analysis of sensor data. It also enables this data to be visualized directly on screens via its many interfaces.
The QuickTray complements the system as a high-performance storage module with a capacity of up to 120TB and write rates of up to 26GB/s (depending on the SSD type and manufacturer used, real measured with Iometer in continuous write mode), enabling continuous recording of vehicle data. The QuickTray can be quickly and easily replaced when needed, providing flexibility and ease of use.
Customer benefits
InoNet’s solution offers significant benefits to the customer by covering all requirements with a single PC, saving the cost of additional systems. Fast data processing during the trip means that only relevant information is stored, reducing the amount of data to be stored and enabling longer recording times.
The QuickTray is seamlessly interchangeable so that data can be easily transferred to an external evaluation station for HIL simulations. This quick transfer after the test drives reduces development time by one hour per test drive.
The ability to visualize the data during the journey offers the customer a further advantage: errors can be detected immediately and any necessary intervention can be made directly if the system has not recorded important objects correctly. This supports the driver and contributes to greater safety and efficiency during testing.
“InoNet’s customized computing platform enables us to bring cutting-edge research to public transport. The reliable, powerful and low-maintenance all-in-one solution covers all our use cases, from data processing and decision making during the journey, including data recording, to resimulation and evaluation after the journey,” said Florian Pfab, research, TUM School of Engineering and Design, Institute of Automotive Technology.
Efficiency improvements
The latest InoNet QuickTray v3 can be replaced even more efficiently thanks to the hot-plugging function, and its built-in data encryption ensures maximum protection of sensitive information from unauthorized access.
Check out what people had to say at the October 2024 launch.