DGWO-F2OPT
Dynamic Grey Wolf Optimizer
with Floating 2-opt Local Search
A multi-objective optimization approach for humanitarian vehicle routing under dynamic flood disaster conditions — balancing distance, time-windows, flood risk, and critical-node coverage.
Humanitarian Routing
Under Uncertainty
During flood disasters, aid vehicles must reach affected communities as quickly as possible — but roads may be flooded, node urgency varies by victim count, and strict time-windows determine whether aid arrives in time.
Classical routing methods fail here. DGWO-F2OPT was designed to simultaneously optimize distance, time-window feasibility, flood-risk exposure, and critical-node coverage using a multi-objective framework.
Mamdani Fuzzy Inference
Node priority is not deterministic — it depends on multiple overlapping indicators. A Mamdani fuzzy inference system fuses victim count, damage severity, and flood risk into a single actionable priority score.
Three Fuzzy Inputs
Number of people affected at the node. Higher counts increase priority.
Infrastructure and property damage level. Severe damage signals urgent routing.
Real-world flood index from BNPB 2016–2022 data, calibrated stochastically.
Inside DGWO-F2OPT
Seven-stage optimization pipeline combining global metaheuristic search with local search intensification and multi-objective Pareto preservation.
Initialize Population
A population of candidate solutions is initialized using equal-division random key encoding, ensuring diverse and balanced initial route distributions across the search space.
Social Hierarchy
as Optimization Engine
The Grey Wolf Optimizer mimics the leadership hierarchy of wolf packs. The three best solutions (Alpha, Beta, Delta) guide the search direction for the remaining population (Omega), naturally balancing exploration and exploitation.
Local Search Refinement
Floating 2-opt refines candidate routes by restructuring route segments after the global optimization step, finding improvements that the global search alone would miss.
Floating 2-opt applies this restructuring flexibly across all route segments — not just adjacent pairs — allowing deeper local optima to be found within each iteration.
Research Dashboard
Performance Comparison
Results across four humanitarian objectives. Only values supported by the project data are shown. Qualitative labels indicate relative performance where exact figures are unavailable.
Pareto Front
Multi-objective optimization produces a set of non-dominated solutions rather than a single answer. Each point on the Pareto front represents a different trade-off between competing objectives.
Hover over the points to explore individual solutions. Non-dominated solutions (on the front) cannot be improved on one objective without sacrificing another.
Humanitarian routing must simultaneously balance travel distance, time-window feasibility, flood-risk exposure, and critical-node coverage. A route that is shortest may be infeasible, high-risk, or leave critical communities unserved. DGWO-F2OPT is designed to find routes that are not just fast — but safe, feasible, and effective.
Built With
What I Built
Algorithm Design
Designed the full DGWO-F2OPT algorithm combining Grey Wolf, Floating 2-opt, and dynamic Pareto archive into a unified optimization engine.
Mathematical Modeling
Formulated the humanitarian CVRPTW with multi-objective constraints including adjusted distance, time-window penalty, risk exposure, and priority coverage.
Fuzzy Logic Integration
Implemented Mamdani fuzzy inference for node priority scoring using real-world BNPB flood data as the risk foundation.
Pareto Archive
Built a dynamic non-dominated Pareto archive that preserves trade-off diversity across multiple objectives throughout optimization.
Experimental Evaluation
Conducted 30-run experiments against Classic GA, Adaptive GA, and MPSO baselines. Analyzed results across all four objective dimensions.
Multi-Objective Analysis
Analyzed trade-offs on Pareto fronts, demonstrating that minimizing distance alone does not yield the best humanitarian routing outcome.
Why DGWO-F2OPT?
| Feature | Classic GA | Adaptive GA | MPSO | DGWO-F2OPT |
|---|---|---|---|---|
| Multi-objective | ✕ | ✓ | ✓ | ✓ |
| Time-window feasibility | ✕ | ✓ | ✓ | ✓ |
| Flood risk awareness | ✕ | ✓ | ✓ | ✓ |
| Priority node coverage | ✕ | ✕ | ✓ | ✓ |
| Local search (2-opt) | ✕ | ✕ | ✕ | ✓ |
| Dynamic Pareto archive | ✕ | ✕ | ✕ | ✓ |
| Fuzzy priority scoring | ✕ | ✕ | ✕ | ✓ |
Explore the DGWO-F2OPT Optimization Codebase
View the multi-objective Grey Wolf Optimizer, floating 2-opt local search engine, and CVRPTW benchmark evaluation.