High Computing Power and Bandwidth for Time-Critical Functions

When developing ADAS/AD functionalities, a number of aspects have to be taken into account because ADAS/AD functionalities intervene directly and immediately in vehicle control. Even if many of these aspects can be processed to a high degree in simulations in the lab, real test drives are still indispensable. After all, as plausible as the trajectory planning in the simulation may look, the real driving feel and driving experience can differ significantly and cannot be replaced by a simulation in the laboratory.      
However, the prototyping of ADAS/AD applications requires high computing power, high data throughput, and real-time capability, as large amounts of data have to be processed and merged for environment recognition and the evaluation of the current driving situation. Driving situations can change very suddenly and very quickly. The relevant vehicle components have to detect these changes and the new situation extremely quickly so that the vehicle can respond with an appropriate driving maneuver. All of this requires a great deal of computing power.     
 

Requirements for the Prototyping System

The requirements for a powerful prototyping system are as follows: 

• High flexibility to adapt and expand sensors, namely camera, lidar, and radar systems  
• Support of all common interfaces such as CAN, Ethernet, etc.
• Direct ECU access to develop new add-on, overlay, or replacement functionalities on the dSPACE system, while the original ECU algorithms remain unchanged
• Time synchronization of data and process sequences
• Ability to process very high data rates in order to make data available for further use or later analysis
• A simulation environment that supports development and testing in a simple way
• Transfer of a model to a framework that can be efficiently executed on the prototyping system

Solution: Combination of AUTERA and RTMaps

The challenges described above can be overcome with a working environment which consists of the AUTERA in-vehicle data acquisition and prototyping system as hardware and RTMaps (Real-Time Multisensor Applications) software specifically for real-time multisensor applications as software. 

AUTERA: Hardware for Data Logging and Prototyping

AUTERA is a dSPACE product family for the data-driven development of ADAS/AD applications. The centerpiece is the AUTERA AutoBox – a modular and robust in-vehicle data logging and prototyping system that can read in raw data, for example, from lidar, radar, and camera sensors, automotive buses, and networks, with very high bandwidth and a precise time stamp. AUTERA also has extensive, hot-swappable storage to record large amounts of data during test drives for validation purposes. This data can be analyzed at a later date and, if necessary, used to generate new test scenarios or train neural networks. 
In addition, the AUTERA AutoBox can be equipped with powerful hardware accelerators for the rapid prototyping of perception and sensor fusion algorithms or direct preprocessing of data during data logging.
Thanks to its high computing power and optional GPU extension, AUTERA offers optimum conditions for processing large volumes of data. The system is delivered with a ready-to-use setup based on RTMaps. However, it is also possible to use a customer-specific software environment for the application based on a Linux operating system.
 

Technical details about AUTERA:

• Up to 50 Gbit/s recording speed with hot-swappable SSDs
• High performance of the server class: Intel Xeon CPU with 12 cores (12x 2.0 GHz)
• 32 GB RAM – expandable up to 512 GB
• Comprehensive bus and network support incl. CAN FD and automotive Ethernet
• Integration of up to 8x GMSL2/FPD Link III cameras in 1 slot
• Connectivity with Wi-Fi, LTE, and GNSS support
• Support of HW accelerators, for example, GPU, for edge AI inferencing
• Flexibility and openness: Expandable with your own PCIe boards in the open Linux OS  

RTMaps: Software for Multisensor Applications

RTMaps (Real-Time Multi-sensor Applications) from Intempora, a dSPACE group company, is a component-based software development and run-time environment that allows users to acquire, time-stamp, synchronize, and process data from different sensors and vehicle buses. With the intuitive graphical user interface of RTMaps, I/O interfaces, for example, for visual sensors, filters, and other components, can simply be dragged from extensive libraries to define inputs, outputs, and data processing steps in an application. User-specific functions can be easily integrated. RTMaps offers interfaces for the integration of C++, Python, Simulink®, or QML-based code, or for connection to ROS and DDS. The individual data processing steps can be coordinated and simulated in the development environment before the entire model can be tested in a highly efficient real-time framework on the prototyping system. 

Technical details about RTMaps:

• Typical applications: ADAS/AD, mobile robotics, data logging, data replay
• Development, testing, validation, and benchmarking of processing and data fusion algorithms
• Log and replay data faster than in real time for offline development and validation
• Graphical programming using block diagrams and simple integration of C++, Python, and Simulink code
• Optimized multi-threaded run-time engine and dedicated real-time functions
• Time-stamping of data and latency measurement for data processing and data synchronization, even on multiple distributed platforms
• Extremely wide range of supported sensors, communication buses, and protocols. More than 600 ready-to-use components for your development tasks, support for over 150 sensor models from the robotics and automotive sectors (cameras, radar, lidar, GPS, IMU, CAN, biometrics, etc.).
  (For complete list see: https://intempora.com/products/rtmaps/supported-technologies/)

Conclusion

When you develop functions for autonomous driving, you use many different sensors that have to be synchronized and processed in real time. High-performance hardware is just as important as a functional and efficient software framework. AUTERA and RTMaps offer precisely these features, making them a perfect couple.