Due to the complexity and the amount of man-hours required to carry out field tests, the team developed a simulator for the N-boat/F-boat. The idea was to create a digital twin, so that all tests conducted in the virtual world could be replicated in practice.
The simulator, named YARA, was created using ROS (Robot Operating System) as the robotic operating system and Gazebo as the virtual simulation environment.
Thanks to this new environment, it was possible to start control tests using artificial intelligence, specifically reinforcement learning techniques (PPO). This allowed the sailboat to learn specialized behaviors for each situation.
Given that the F-boat has an electric outboard motor, various behaviors were created which utilize this feature extensively (consuming a lot of battery) to behaviors that do not use any electric propulsion (using only the wind as a means of locomotion). These specialized behaviors allow the sailboat to adapt to various conditions, consolidating itself as a more generalized multi-mission platform.
Here are the videos of the simulator:
Simulator demonstration:
Training(PPO) demonstration:
Track mission demonstration:
Perimeter coverage mission demonstration:
Complete videos playlist:
Source code repositories:
YARA Ocean Virtual Environment Github.YARA Ocean Virtual Environment ZIP.
Yara simulator with ia saling behaviors Github.
Yara simulator with ia saling behaviors ZIP.