AI Self Sailing Boats
Thursday 25 April, MARIN welcomed an enthusiastic audience that joined the AI Sail workshop. This audience was a nice mixture of student and people from companies and other fields of work. After Hannes Bogaert kicked off the workshop with MARIN’s strategy about the use of AI, Fanny Rebiffe continued the workshop. She shared her knowledge from MARIN’s AI Sail project and provided use cases to allow the participants to train and tune their AI agent to learn to sail a sailing boat for kids (Optimist) against the wind. The end result of their tuning effort showed new insights to everyone that participated, including MARIN. This intuitive case is a stepping stone to real applications. For example, we are now working on ship emissions reduction by combining AI and on-board human experience. AI sail project is an accelerator of MARIN AI developments.
In November 2023, MARIN gave a public demonstration of an AI remotely sailing an Optimist upwind. We invite you to participate to a workshop in which you will use MARIN toolbox developed for this demonstration.
Reinforcement Learning (RL) is a branch of Artificial Intelligence (AI) that can optimise a strategy through interactions with a dynamic environment. The RL agent interacts with the environment by observing the state of the environment, deciding of an action and receiving a feedback called a reward. The agent is trained to optimize the cumulative reward over a sailing course. The reward is a transcription of a broad range of skills and knowledge such as education, sailing, data science, physics… thus everyone can contribute to the design of the reward, key element of this AI. This workshop will allow you to train your own RL agent without programming skills.
The aim of AI sail is to explore the possibilities and limitation of RL on an intuitive case study: a small sailboat intend for children training.
During the first half of this workshop, the trainees will have an overview of AI sail project, Reinforcement Learning and MARIN RL toolbox. During the second half of the workshop, the trainees will design in group the training of their own RL agent with the help of a RL specialist and using MARIN RL toolbox. A short analysis of the results of the training and their RL agent performance will be send afterward to the trainees.
Program
9:00 - 9:30am Welcome and overview of our AI developments
9:30 - 10:00am Summary of AI SAIL project
10:00 - 12:00pm Hands-on workshop – Part 1
12:00 - 1:00pm Lunch and Tour
1:00 - 1:45pm Second round tuning
1:45 – 2:30pm More details about AI
2:30 - 3:15pm Third round tuning
3:15 - 4:00pm Benchmark results & Closure of Day
… if you want to keep improving your AI-agents, no problem, there is time until 5:00pm.
Details
* Registration for KIVI members opens at 15 March 2024.
* For non-members -if seats are still available- registration opens at 18 April 2024.