Yard-Crane Scheduling Problem
The Yard-Crane Scheduling Problem (YCSP) decides how to sequence storage and retrieval jobs for one or more yard cranes (RTG/RMG) across yard blocks. Schedules must respect inter-crane interference, non-crossing/separation rules, travel and hoist times, and job ready times (truck/AGV arrivals). Typical goals include minimizing vessel-related makespan, reducing truck/AGV waiting, and balancing crane workloads.
Optimization Focus
Objective
- Minimize vessel/yard makespan and truck/AGV waiting.
- Reduce crane empty travel and interference delays.
- Balance workload; enable energy/throughput trade-offs.
Decision Variables
- Assignment/sequencing of jobs per crane with start/finish times.
- Feasible crane moves between bays/blocks with separation.
- Optional: dynamic I/O point selection and buffer use.
Constraints
- Inter-crane interference: minimum separation; no crossing on the same rail.
- Job ready times, bay travel and hoist times, precedence rules.
- Block boundaries, transfer zones, and safety margins.
Data Sources
- TOS logs (job timestamps, locations), bay-plan/stowage.
- Crane kinematics (speeds/accels), block layouts, rail distances.
- Truck/AGV arrival processes and I/O buffer status.
Main Assumptions
- Job set and locations known in advance (static YCSP variant).
- Cranes share a rail per block with fixed safety separation.
- Travel/handling times are deterministic or scenario-based.
- Precedence reflects stacking rules and equipment interfaces.
- Optional: I/O point selection included when buffers are modeled.
Modeling Approaches
YCSP is often formulated as a MILP (single/multi-crane) with interference constraints; realistic sizes frequently use matheuristics and metaheuristics.
- Exact/matheuristics: continuous-time MILP, branch-and-bound, decomposition, set-partitioning.
- Metaheuristics: GA/TS/SA, GRASP/IG, hybrid evolutionary; RL for dynamic variants.
- Integrated: joint planning with trucks/AGVs or with quay-crane operations for end-to-end flow.
Available Datasets
🧭 Real-World Datasets
| Dataset / Case | Description |
|---|---|
| Real Terminal Event Logs | timestamps, workers, costs, plus yard fields such as: yard block and yard slot. The user can extract storage/retrieval tasks and build YC schedules from the logs (License: CC BY-NC 4.0). |
| UCI ML Repository | Real crane controller signals, including speed, angle, and power. The dataset is a control-level (not a ready YC schedule), but useful for data-driven timing/energy models that can parameterize YC task durations. |
⚙️ Synthetic Benchmark Instances
| Dataset | Description |
|---|---|
| Multi-YC Scheduling Instances | Instances for 3 cranes / 2 blocks with 6–12 tasks. Provided as direct files for algorithm testing (License CC BY 4.0). |
| Test Data Generator | Generator for test instances of scheduling problems concerning cranes in transshipment terminals. |
Real-world datasets are typically restricted due to confidentiality, but the cited studies document realistic layouts and parameters. The public benchmarks (e.g., Mendeley multi-crane sets, I/O point selection instances) enable reproducible evaluation and method comparison.