A hybrid evolutionary algorithm for general routing problems

Although evolutionary algorithms have often been used to address VRPs and VRPTWs [1], applications to other routing problems such as period routing and inventory routing are less common. In this paper, a versatile hybrid evolutionary algorithm is developed that can be used for cyclic inventory routing and a variety of capacitated and periodic vehicle routing problems. First, a random insertion heuristic and three well-known local search (LS) operators are used to generate a pool of high quality solutions. An evolutionary algorithm (EA) then manages solution quality and diversity in the pool by using disruptive crossover and mutation operators combined with the same three local search operators. The versatility and performance of the approach is illustrated on a set of benchmark problems on the cyclic inventory routing problem (CIRP), the periodic vehicle routing problem (PVRP) and the classical capacitated vehicle routing problem (CVRP).