Hybrid genetic search for the CVRP: Open-source implementation and SWAP* neighborhood

The vehicle routing problem is one of the most studied combinatorial optimization topics, due to its practical importance and methodological interest. Yet, despite extensive progress, many recent studies are hampered by limited access simple efficient open-source solution methods. Given sophistication current algorithms, reimplementation becoming a difficult time-consuming exercise that requires care for details be truly successful. Against this background, we use opportunity short paper introduce – implementation hybrid genetic search (HGS) specialized capacitated (CVRP). This state-of-the-art algorithm uses same general methodology as Vidal et al. (2012) but also includes additional improvements lessons learned over past decade research. In particular, it an neighborhood called SWAP* which consists in exchanging two customers between different routes without insertion place. As highlighted our study, exploration moves significantly contributes performance local searches. Moreover, observed experimental comparisons with other approaches on classical instances Uchoa (2017), HGS still stands leading metaheuristic regarding quality, convergence speed, conceptual simplicity.