Hybrid ML-ACO Route Optimisation
Summary: A practicum research project combining Ant Colony Optimisation with KNN, ANN and LSTM guidance models for last-mile delivery route optimisation.
KNN-ACObest performing model
VRProute optimisation problem
Pythonsimulation and analysis
Problem
Last-mile delivery routing is complex because vehicles must serve many delivery points while minimising distance, time and operational inefficiency. This problem can be modelled as a Vehicle Routing Problem.
Approach
- Used Ant Colony Optimisation as the core routing method.
- Compared KNN-ACO, ANN-ACO and LSTM-ACO hybrid approaches.
- Analysed spatial delivery patterns and route convergence.
- Evaluated route quality using distance and hit-rate style metrics.
Finding
KNN-ACO performed strongly because local spatial similarity is useful for static route optimisation. The project showed that simpler learning methods can be effective when the problem has strong spatial structure.
Outcome
The project demonstrates practical machine learning experimentation, optimisation logic, route simulation and research-style evaluation.