Berth Optimizer Heuristic | Products

Overview

Berth Optimizer (Heuristic) is the AI behind your berth-allocation desk. It takes a vessel queue, berth catalogue, crane pool and constraint pack, and returns an optimised berth-and-time plan that minimises total vessel wait while maximising berth and crane utilisation — subject to length, depth, tide, hazmat, line-priority and shift constraints. Built as a companion to the Berth Scheduling product, but works against any berth model.

The optimiser is heuristic-based (simulated annealing + genetic + local search) so it solves problems too big for pure MILP — a 14-day, 6-berth, 80-vessel plan in seconds. Planners stay in the driver's seat: every proposal is reviewable, tweak-able and explainable.

Typical impact — vessel wait ↓30–40%, berth utilisation ↑8–15%, throughput ↑10–18%
Constraint-aware — length, depth, tide, crane span, hazmat, gang continuity, line weights
Explainable — every proposed move tagged with cost delta and constraint reasoning
Reactive — re-optimises on ETA changes, breakdowns, weather closures

How It Works

Problem load — reads current schedule, vessel queue, berth + crane catalogue, constraint pack.
Cost model — weighted sum of (vessel wait) + (idle berth time) + (crane shifts) + (line-priority penalties).
Simulated annealing — explores broad solution space with cooling temperature.
Genetic crossover — takes top candidates, swaps assignment chromosomes, mutates.
Local search — final refinement with greedy swap-and-shift moves.
Tabu list — avoids cycling back to recently-tried bad assignments.
Publish — proposes assignments to the planner; on accept, pushes to Berth Scheduling / TOS.

Features

Hybrid solver — SA + GA + local search + tabu, with OR-Tools MILP fallback for small problems
Constraint pack — declarative YAML: length, depth, tide, crane, gang, hazmat, line priorities
Multi-objective — planner sets weights for wait vs util vs cost in a slider UI
What-if mode — A/B compare two solver runs side-by-side
Reactive re-optimisation — auto-trigger on ETA delta, breakdown, weather window
Explainability — every move comes with cost delta, alternatives considered, constraints satisfied
Audit trail — full history of solver runs, parameters, planner acceptance
Closed-loop learning — planner overrides feed back into cost weight tuning

Use Cases

Container terminals — 4–12 berth, 30–120 vessel/week problems
Bulk & multi-purpose ports — mixed-cargo with tight tide and bunker windows
Port authorities — central berthing optimiser across multiple operators
Greenfield design — what-if simulations for berth count, length, crane fleet sizing
Disruption recovery — storm, strike or breakdown re-plans in seconds
Decision support for planners — runs in background, proposes when better solutions found

Specifications

Solve time: 14-day / 6-berth / 80-vessel problem <15 s on 8 CPU; 21-day / 12-berth <60 s
Quality: typically within 1–3% of MILP optimum on tractable problems
Inputs: JSON / REST / direct DB; live ETAs from AIS or agency feeds
Constraints: 50+ built-in; custom Python plugin API
Engine: Python + Cython hot loops; multi-core parallel exploration
Deployment: Docker image, k8s helm chart, or managed SaaS
APIs: REST + WebSocket for live re-opt; Webhook for proposal delivery
Integrations: SGT Berth Scheduling, Navis N4, OPUS, CATOS, custom EDI