Introduction
The development, simulation, and validation of AI-controlled autonomous platforms, including vehicles, robots, UAVs, and marine platforms, are highly demanding processes that combine real-world testing with extensive laboratory-based simulations.
While real-world testing remains the ultimate validation step before deploying an autonomous platform, many companies increasingly replace a significant portion of physical testing with realistic Hardware-in-the-Loop (HIL) simulations. HIL testing is less costly, provides greater controllability, and improves overall test quality and validation coverage.
HIL Mode - Raw or Synthetic Data Injection
HIL Testing
Xylon offers complete HIL platforms optimized for testing AI systems based on video cameras, LiDARs, RADARs, and other low-bandwidth sensors. Xylon’s logiRECORDER Automotive Video HIL Logger and the more powerful XYLON QUATTRO Data Logging and HIL System, initially developed for the automotive industry, are successfully used across multiple markets.
These platforms can replay accurately reconstructed real-world test scenarios while simultaneously and optionally recording the responses of the Device Under Test (DUT).
The HIL logger replays vehicle sensor data in its original format and connects directly to the DUT through native sensor interfaces. Playback data is electrically and logically formatted to fully comply with the requirements of the target platform’s internal communication buses. In this way, the HIL logger efficiently replicates a complete test vehicle’s multi-camera video system and in-vehicle data infrastructure on a developer’s desk.
CLICK HERE to learn more about HIL testing with raw sensory data.
Data Acquisition in the Field
Raw Data Stimuli - Acquisition
HIL testing requires raw sensory data recorded in the field to properly stimulate the DUT. The acquisition of raw data is performed by dedicated devices called data loggers.
Data loggers are placed between the autonomous platform’s sensors and the AI processing computer (DUT). They non-intrusively capture large amounts of test data in an uncompressed format, with centralized timestamping and extremely low latency.
Xylon’s logiRECORDER and XYLON QUATTRO platforms combine advanced data logging and HIL capabilities, enabling the recording of up to 128 TB of valuable test data.
CLICK HERE to learn more about data logging.
Synthetic Data Stimuli - Generation
In addition to field-recorded raw sensory data, HIL simulations can also utilize synthetically generated data, providing improved controllability and more economical testing.
Xylon HIL platforms integrate with various third-party simulation environments, allowing developers to test challenging corner-case scenarios in a safe and repeatable way until autonomous control algorithms are fully optimized.
CLICK HERE to learn more about HIL testing with synthetic data.
The Demo
Xylon’s configurable logiBRICKS Vision AI Framework enables AI engineers to efficiently deploy AMD adaptive SoCs across a wide range of vision-based applications.
Explore our demo systems for airborne drone platforms, industrial warehouse environments, and automotive applications.
The demo system operates in an HIL setup using Xylon’s logiRECORDER to inject multiple high-resolution camera streams from real and synthetic video sources, together with additional sensor data (speed, angles, and more) through CAN and Ethernet networks.
CLICK HERE to learn more about Xylon’s logiBRICKS Vision AI Framework.
Click the image below to see the AI demo in action.