Optimisation problems can be tackled by multiple metaheuristic approaches, but individual methods have distinct strengths and weaknesses. Could combining Ant Colony Optimisation (ACO) and Genetic Algorithms (GA) outperform either alone?
- Implemented and benchmarked standalone ACO and GA solutions
- Designed a hybrid method leveraging both search strategies
- Compared performance across test problems with rigorous evaluation
Demonstrated that the hybrid approach could combine the exploratory power of genetic algorithms with the constructive search behaviour of ant colony methods, with documented trade-offs across problem types.
- Python
- Custom ACO & GA implementations
- Statistical comparison & reporting
Background
This was my undergraduate final year dissertation project. Metaheuristic algorithms are widely used when exact solutions are impractical. ACO mimics ant foraging behaviour to explore solution spaces, while GA evolves populations of candidate solutions through selection, crossover, and mutation.
What I learned
Designing a fair comparison framework is as important as the algorithms themselves. Normalising performance metrics, choosing representative test instances, and documenting parameter choices were critical to producing credible results — lessons I still apply when building and evaluating models in industry.