Genetic Algorithm vs Firefly for a Single-Truck, Single-Drone Vehicle Routing Problem
Abstract
Janith Thenura Akuratiya Gamge
We study a practically motivated variant of the Vehicle Routing Problem with Drones (VRP-D): a single truck collaborates with a single drone to serve customers. On each truck edge (i → j), at most one drone sortie may be launched at i, visit exactly one customer k, and rendezvous at j. The edge time is the maximum of the truck’s travel and service time versus the drone’s sortie time (including launch and recovery). We compare two popular metaheuristics—Genetic Algorithm (GA) and Firefly Algorithm (FF)—against a simple baseline (Nearest-Neighbor + 2-opt with greedy drone assignment). We implement a reproducible experiment suite, parse parameters and a 20-customer instance from provided files, and report best-time (makespan) performance over multiple random seeds. On the 20-customer instance, FF tends to achieve lower total time than GA (Mann–Whitney: z ≈ −1.73, p ≈ 0.083; Cliff’s δ ≈ −0.531, large, negative favors FF). The baseline offers a fast, competitive reference. We release code, datasets, and plots to support replication.