Mixed-Fleet VRP under Dynamic Demand

This line of work studies a mixed-fleet vehicle routing problem where electric vehicles and conventional vehicles are used together.

The problem setting includes customer demand, vehicle capacity, EV energy limits, charging infrastructure, time-related constraints, operating costs, and carbon-related cost terms. The experiment is not only about finding a low-cost route. It is also about checking whether the route is actually feasible and whether the reported result is traceable.

Method

The current implementation uses heuristic and metaheuristic components, including simulated annealing, genetic algorithm operators, and hybrid search logic. The exact algorithm description must match the repository implementation. I do not describe the algorithm as a novel method unless the code and paper explicitly support that claim.

My role

My specific contribution is auditing the system's internal consistency at three levels.

1. Model layer
   • formula validity
   • dimensional consistency
   • objective-function terms
   • constraint completeness
   • vehicle-specific assumptions
2. Code layer
   • whether MATLAB functions compute what the model specifies
   • whether data dimensions match the mathematical notation
   • whether decoding, repair, and feasibility checks are consistent
   • whether EV and conventional-vehicle logic are separated correctly
3. Experiment layer
   • whether results are reproducible
   • whether algorithm comparisons are fair
   • whether initialization, stopping conditions, penalty structures, and random settings are controlled
   • whether route figures, convergence curves, and result tables come from the same solution

Repository

github.com/zlxshu/dynamic-demand-mixed-fleet-vrp-sa-ga-reproduction

I describe it as a research codebase and working reproduction system, not as mature commercial software.