Abstract
When Wacker Neuson develops new excavator models, the prototypes must undergo extensive testing during the development phase. One standard method is the completion of digging cycles, which were previously carried out manually by excavator operators. In order to reduce costs, but also to do preliminary work in the field of autonomous excavation, such a cycle should be automated. This master's thesis therefore deals with the development of an autonomous digging cycle for the EW100 mobile excavator from Wacker Neuson using the Robot Operating System ROS2. Based on the master's thesis by Sebastian Buchner [2], the control of a digital twin is first implemented in the simulation environment Nvidia Isaac Sim. This is helpful for testing the general architecture of the ROS2 system. A simulation also has the advantage that algorithms can be tested more easily and quickly than on the real excavator.
Subsequently, a PD joint control system is designed for the excavator. A setpoint trajectory including feedforward control is specified, which is obtained by storing the path and the associated joystick signals of a manual digging cycle. PD control of the joint angles during operation makes it in principle possible to continue a periodic, constant digging cycle for any length of time. However, the problem can arise that the shape of the material cone changes during operation and that the excavator therefore no longer excavates material but only air from a certain point in time.
Due to the changing environment, material measurement of the bucket is realized using a lidar sensor, but also using pressure sensors on the boom cylinder. In addition, the former is also used to scan the environment. The RANSAC algorithm is used to detect an inclined plane in the point cloud of the lidar sensor. If the material content of the bucket falls below a certain limit, the path is dynamically adjusted by selecting the lowest point of the detected plane as the new digging point. In addition, the position of the under carriage is also checked using the rearview camera of the excavator and an ArUco-Marker attached behind it, as it is possible that the under carriage may move unintentionally.
Subsequently, a PD joint control system is designed for the excavator. A setpoint trajectory including feedforward control is specified, which is obtained by storing the path and the associated joystick signals of a manual digging cycle. PD control of the joint angles during operation makes it in principle possible to continue a periodic, constant digging cycle for any length of time. However, the problem can arise that the shape of the material cone changes during operation and that the excavator therefore no longer excavates material but only air from a certain point in time.
Due to the changing environment, material measurement of the bucket is realized using a lidar sensor, but also using pressure sensors on the boom cylinder. In addition, the former is also used to scan the environment. The RANSAC algorithm is used to detect an inclined plane in the point cloud of the lidar sensor. If the material content of the bucket falls below a certain limit, the path is dynamically adjusted by selecting the lowest point of the detected plane as the new digging point. In addition, the position of the under carriage is also checked using the rearview camera of the excavator and an ArUco-Marker attached behind it, as it is possible that the under carriage may move unintentionally.
| Original language | English |
|---|---|
| Qualification | Master |
| Awarding Institution |
|
| Supervisors/Reviewers |
|
| Award date | 07 Jul 2025 |
| Publication status | Published - Jul 2025 |
Fields of science
- 202027 Mechatronics
- 203013 Mechanical engineering
- 202 Electrical Engineering, Electronics, Information Engineering
- 202035 Robotics
- 203022 Technical mechanics
- 203015 Mechatronics
JKU Focus areas
- Digital Transformation