A Hybrid (1+1)-Evolutionary Strategy for the Open Vehicle Routing Problem

Abstract

The Open Vehicle Routing Problem (OVRP) is a small but practically relevant modification of the well known Capacitated Vehicle Routing Problem, in which each route ends at the last served customer. We present how the OVRP and its constraints are modeled in the super customer framework, an integrated modeling and optimization framework for general vehicle routing problems. The framework provides several standard neighborhoods that are based on the path exchange operator and allows the use of accelerated delta function evaluations. With predefined templates, these neighborhoods can be used in mutation operators, evolutionary algorithms, neighborhood search, multiple neighborhood search and iterated variation selection procedures. For the OVRP, we used this framework to implement a hybrid combination out of a stochastic multiple neighborhood search and an (1+1)-evolutionary strategy. The performance of this hybrid (1+1)-evolutionary strategy was evaluated on 16 standard benchmark instances from literature. Best known solutions were computed for all 16 instances and new best solutions were found for seven instances.