3D Test Environment
for Autonomy Controllers
Developers of autonomous robots or tractors typically test their path-following controllers, collision avoidance systems, and implements on real machines under real-world conditions—a process that entails significant time and cost. HYMATS—a 3D test environment for autonomy controllers—offers a way to conduct more extensive testing faster and under reproducible conditions. It consists of a real-time simulation computer that generates a realistic, interactive environmental simulation, combined with the actual control unit running the original autonomy software.
Intuitive Configuration of the 3D Test Environment
An intuitive user interface is provided for operating the 3D test environment. Users can select from various machines, interface configurations, and environmental settings. The 3D test environment simulates both the prime mover (or carrier vehicle) and the implement. Operators can currently choose between two carrier vehicles—the EIDAM TK100 (articulated steering) and the InnoTRAC Caesar (Ackermann steering)—but can also create custom vehicle models using CAD data, as well as various generic types or classes. To facilitate this, users are offered a range of machine configurations (number of axles, steering type, mounting positions, dimensions). Just as with the carrier vehicles—which feature multiple mounting positions (front, center, rear)—implements can also be inserted as generic models. These allow users to quickly define a custom implement (specifying coupling type, number of axles, steering type, and dimensions) and attach it to the carrier vehicles at the designated interfaces.
The machine models include the necessary driving and steering functions. Kinematics, mass distribution, and inertia are accounted for in the model, as is wheel slip. This is particularly necessary when driving on slopes or maneuvering on headlands, as the autonomy controllers must compensate for the associated disturbances.
Models of LiDAR sensors (solid-state or rotating), GNSS sensors, and cameras can be selected for the environmental sensing system. The sensor models are parameterizable, allowing for the investigation of various sensor types. All details of the virtual environment are captured by the sensor models and transmitted to the autonomy controller—either as raw data or via ROS2 topics over Ethernet, or through CAN messages (J1939 or ISOBUS).
Users can modify the virtual environment using a scenario editor, which allows for the definition of static and dynamic obstacles. Static obstacles might include parked implements or fruit crates, while dynamic obstacles could be people or other machines. Users have full control over when and where these obstacles appear or disappear. Furthermore, the speed and path of dynamic obstacles can be defined via the user interface, as can the day-night cycle and weather conditions.
To automate testing, an interface was developed and implemented to connect test management software—exemplified by TraceTronic’s *ecu.test* software. This interface enables the definition, automated execution, and automated evaluation of complex test scenarios. The entire Continuous Integration (CI) process was implemented in Jenkins. It is also possible to feed test results back into the corresponding Application Lifecycle Management (ALM) tool, such as PTC RV&S.
German Presentation for Download
3D Test Environment
Funded by the Federal Ministry for Research, Technology and Space
