A bi-objective stochastic approach for stochastic CARP

The Capacitated Arc Routing Problem (CARP) occurs in applications like urban waste collection or winter gritting. It is usually defined in literature on an undirected graph ( ) E V G , = , with a set V of n nodes and a set E of m edges. A fleet of identical vehicles of capacity Q is based at a depot node. Each edge i has a cost (length) i c and a demand i q (e.g. an amount of waste), and it may be traversed any number of times. The edges with non-zero demands or tasks require service by a vehicle. The goal is to determine a set of vehicle trips (routes) of minimum total cost, such that each trip starts and ends at the depot, each task is serviced by one single trip, and the total demand handled by any vehicle does not exceed Q . To the best of our knowledge the best published method is a memetic algorithm first introduced in 2001. This article provides a new extension of the NSGA II (Non-dominated Sorting Genetic Algorithm) template to comply with the stochastic sight of the CARP. The main contribution is: • to introduce mathematical expression to evaluate both cost and duration of the longest trip and also standard deviation of these two criteria. • to use a NGA-II template to optimize simultaneously the cost and the duration of the longest trip including standard deviation. The numerical experiments managed on the thee well-known benchmark sets of DeArmon, Belenguer and Benavent and Eglese, prove it is possible to obtain robust solutions in four simultaneous criteria in rather short computation times.