ASV NIRIIS
Autonomous Surface Vehicle "Niriis"
Technical Details
| Size: | 135 cm x 38 cm x 25 cm |
| Weight: | 4,5 kg (empty weight), 10,5 kg (with batteries) |
| Power supply: |
2x 22000 mAh 12V LiPo, 1x 5000 mAh 5V LiPo
|
| Speed: | approx. 5 m/s (9.72 kn) |
| Actuation/ Engine: |
Thruster: 720W, max. thrust: 25 N
Steering servo: ± 16°
|
| Sensors: |
1x 4K RGB Camera (Sensor: SONY078, angle of view: 170°)
1x Infrared camera (Spectrum: 7.5 – 13.5 µm, resolution: 336x256 px)
|
| Communication: |
WLAN-ac
Bluetooth 5
|
Organisational Details |
|
| Sponsor: | Federal Ministry for Economic Affairs and Climate Action |
| Grant number: | 03SX540D |
| Application Field: | Underwater Robotics |
System description
The ASV NIRIIS is used at DFKI RIC to evaluate algorithms for autonomous locomotion on the water plane. This includes problems as model identification, path-following control, trajectory tracking, and trajectory planning. Besides classical approaches, there is ongoing research on deep reinforcement learning -based approaches, which are evaluated on this vehicle.
The vehicle was built by the company Altus-LSA (https://altus-lsa.com/) and expanded by DFKI with more advanced electronics. This includes a Raspberry Pi 4B running Ubuntu 20.04, which is connected to the DFKI intranet via WLAN. Furthermore, a ROS (Robot Operating System) stack was developed for the system, providing basic control and navigation functions.
The position of the vehicle can be tracked by a ceiling camera system, which is fused with a 9-DOF inertial measurement unit using an extended Kalman Filter. In addition, two forward-facing cameras are mounted on the vehicle to test advanced algorithms for image-based navigation and collision-avoidance. The 4K RGB-camera has a SONY078 sensor and an angle of view of 170 degrees. The infrared camera installed in parallel has a VOx Microbolometer sensor, a focal length of 13 mm and a spectrum of 7.5 – 13.5 µm. The image resolution is 336x256 px.
For easy operation, a web interface was developed by DFKI, which connects to the underlying ROS stack. Besides manual control, it is possible to draw paths the vehicle has to follow. After a control command has been issued, the tracking accuracy is evaluated in the background.