Harbour berth assignment with reinforcement learning
Cross-disciplinary project SV420 with Kristiansand harbour — sole technical contributor.
Context
Berths are limited; vessels arrive under uncertainty. Manual planning works but is hard to stress-test and optimise under changing traffic.
What we built
- A reinforcement-learning prototype that explores policies for assigning ships to berths under simplified rules.
- Visualisation: Plotly dashboards to communicate schedules and utilisation to non-RL stakeholders.
Illustration placeholder
Harbour schematic, Gantt-style chart from Plotly, or training curve — good eye-candy for this page later.
Why it matters
Bridges classic OR-style scheduling discourse with modern ML experimentation: even when production schedulers stay heuristic or MIP-based, RL prototypes can explore edge cases and spark discussion with domain experts.