Project Title: Intelligent Methods for the Electric Vehicle Routing Problem (EVRP)
Student Name: Sarthak Shalin Nanda
Student ID: 229007926
- 15 Jan: First experimentation with Mixed Integer Linear Programming (MILP) for solving EVRP considering basic constraints (battery, distance).
- 19 Jan: Initial implementation of a basic Genetic Algorithm (GA) for EVRP. Developed a basic auto-route generation to replace hard-coded instances.
- 28 Jan: Added initial vehicle route generation capabilities; foundational work on fitness evaluation began.
- 8 Feb: Developed initial versions of fitness functions; partial implementation of a MILP solver.
- 16 Feb: Implemented crossover and mutation with elitism in the GA to avoid premature convergence.
- 17 Feb: Minor edits to MILP model structure.
- Throughout Feb: Committed code regularly, focusing on population seeding heuristics.
- Early March: Modularisation of code base to improve structure and maintainability.
- 17-31 Mar: Faced and fixed issues related to infeasible heuristics—started pivoting towards a constructive solver approach using Clarke and Wright Savings (CWS) heuristic.
- 1 Apr: Completed final attempt at fixing feasibility with older methods; pivoted decisively towards Clarke and Wright constructive approaches.
- 6-7 Apr: Implemented Clarke and Wright heuristic; initial successful routes emerging, albeit needing optimisation.
- 8-9 Apr: Iterations led to improvements in feasibility; cleaned and clarified fitness functions.
- 10 Apr: Added intra-route 2-opt swap to improve individual routes.
- 16-17 Apr: Developed and integrated a custom evaluation module for feasibility and cost; debugging efforts included visual enhancement of plotted routes.
- 18 Apr: Created route image generation scripts; incremental improvements to feasibility via GA route validation.
- 17-23 Apr: Progressed evaluation modules; ensured evaluation pipelines were separated from main solvers for cleaner analysis.
- 23 Apr: Fixes applied for CWS result feasibility; developed route plotting tools.
- 24 Apr: Refactored main scripts to allow automated result export to CSV for easier benchmarking; extended GA scripts to handle multiple targeted instances.
| Date | Task | Status |
|---|---|---|
| Jan 2025 | Basic GA and MILP prototypes | Completed |
| Feb 2025 | Developed advanced GA mechanisms (crossover, elitism) | Completed |
| Mar 2025 | Realised limitations of heuristic approaches; pivoted to constructive CWS solver | Completed |
| Apr 2025 | Implemented CWS and integrated multi-instance evaluations; improved visualisation | Completed |
| Late Apr 2025 | Final refactoring of solvers and evaluation pipeline; extensive benchmarking setup | Completed |
- Feasibility issues: Early GA approaches struggled with battery and capacity constraints, leading to a necessary pivot towards constructive heuristics.
- Debugging and Evaluation: Designing a standalone evaluation module was challenging but critical for modular and scalable solution assessment.
- Visualisation Enhancements: Significantly improved graph plotting, aiding deeper route and solution quality analysis.
- System refinement: Final minor tweaks for visualisations and CSV output formats to streamline analysis.
- Prepare mini viva presentation: Summarise the year's work into a 5-minute technical presentation with supporting slides.