Local Search and Evolutionary Computation for Arc Routing in Garbage Collection

The paper studies hybrid solution methods for a general class of arc routing problems arising in the context of garbage collection. Important differences of basic local optimizers for arc oriented compared to node oriented problems are worked out. The initial problem is split into its routing (sequencing) and clustering part. For both problems meta-procedures that make use of the modified local search procedures are proposed. The routing part is a well defined problem called Mixed Rural Postman Problem with Turn Penalties. For this problem an Evolutionary Algorithm is implemented and compared to known solution methods. It is able to provide the best known solution quality at the expense of high computational effort. The clustering part is shown to be very application dependent. To offer a flexible modeling of the problem a multi-agent-system using the formerly presented local search operators is proposed.