ELECTRO: Embedding-guided large neighbourhood search for the electric vehicle routing problem with time windows

Open data of the article "ELECTRO: Embedding-guided large neighbourhood search for the electric vehicle routing problem with time windows" (Yuksel and Gudul). Every value reported in the article can be checked with this deposit: the solution files hold the routes, and the independent validators recompute feasibility, number of vehicles and distance from the instance files. Contents:- Solutions with their routes for every benchmark of the article: the 56 Schneider et al. (2014) instances with full and partial recharging, the comparison set with HMA, the six JD-derived instances in Beijing (150-900 customers), synthetic instances with 2,000 to 100,000 customers, the WCCI-2020 E-CVRP instances and the 115 instances with proven optima of Desaulniers et al. (2016).- New best known solutions for eight E-CVRP instances (also in the competition's text format) and the first known solutions of the JD-derived instances, full and partial recharging.- Every run summary (instance, seed, feasibility, vehicles, distance, time, configuration), the multi-objective fronts and the article's tables.- Validators (Python, MIT licence): evrp_validate for all benchmarks and ecvrp_check for the E-CVRP files.- Visualisation: three static web pages that play the plans as delivery days (a 3D route scene, a delivery day in Istanbul on our own 3D city built from OpenStreetMap, and the JD plans on the Beijing road network), with the scripts and Blender files that built them. README.md describes the contents, how to check a solution and how to view the pages; CREDITS.md lists third-party sources and their licences. The solver itself is not part of this deposit.