Roadmapping a long-term AI initiative

FORCE Technology

Roadmapping a long-term AI initiative

"Halfspace saved us months of research trying to find the right approach. Halfspace pointed out the way forward both in the short and long term, giving a more efficient way forward and no doubt a better end product", Bugge Jensen, Snr. Project Manager

Summary

Accelerated project progress

Saved months of research and piloting

Vetted approach for AI adoption

Outlined a vetted and actionable roadmap to start and operationalise the AI initiative

“Halfspace delivered invaluable insights in which AI methods are suited to extract human expert knowledge. Very early on, Halfspace forced us to attack the problem in general terms, instead of thinking of a specific case. This meant that, independent of the formulation, the suggested method and approach could be reused for other ships in other ports. We found a way of reusing this formulation independently of the ship type, and eventually on the usage of tug. Halfspace’s mantra of “Thinking big, but starting small” was certainly put into practice, and we will live by these words in the coming sprints of the development of our AI navigator"

Bugge Jensen, Snr. Project Manager, FORCE Technology

Challenge

FORCE Technology is an internationally leading non-profit technological consulting and service company that, among many other things, operates a ship simulation center. FORCE Technology has recently begun a long-term project building a ship simulator that can navigate without a human captain but using AI. The vision is to build and train an AI model that is able to:

  • berth a ship in a given port in a ship simulator under varying weather conditions, using various ship types.
  • emulate how human captains would behave using a large dataset of historical simulations.

FORCE Technology wanted support to break down this complex vision and guidance for the technical approach to build the AI agent. The main challenge was the lack of a clear roadmap for AI adoption: Knowing where and how to start a long-term AI initiative

Solution

FORCE Technology and Halfspace worked together over a period of five weeks to combine expert knowledge within ship simulations and AI methodologies, respectively. The overall goal was to define the best possible technical roadmap for the long-term AI initiative: How to teach an AI model to navigate using Inverse Reinforcement Learning (RL). Inverse RL allows AI agents to learn the objectives that human experts optimise for by simply observing and emulating expert behaviour.

The Key Takeaway

Halfspace advised FORCE Technology on how to make a complex AI vision more actionable, by validating the approach and accelerating the project by reducing time-consuming methodology research and piloting. 

Starting off right is crucial for succeeding with major AI initiatives. However, at these early stages, many long-term factors need to be considered carefully to ensure an efficient process and the right approach to AI that can be operationalised. Now, FORCE Technology is well-equipped to start their AI initiative with a thoroughly vetted and actionable roadmap.