Constructive heuristics for weather-type models of capacitated arc routing problems in waste collection

Abstract

In this thesis, a new problem of arc routing for a fleet of vehicles in waste collection is studied. Objectively, this study investigates the effect of external factor, precisely the rainy weightage to expect the optimum routing cost and the number of trips for a fleet of vehicles in domestic waste collection. Although the rain is considered a crucial factor in Malaysia due to its ability to change the demand behaviour, consequently increased the quantity and trips, this factor has never been studied within the context of routing problems. Thus, two problems namely capacitated arc routing problem with stochastic demand (CARPSD) and capacitated arc routing problem with delivery time window (CARPDTW) were formulated to present the vehicle operations in waste collection. In general, conventional CARP in waste collection lies on undirected network graph, where a vehicle starts with empty capacity at a depot, performs pick-up of the customer demands on a set of required edges and ends at the depot without exceeding its capacity. In order to produce optimum or near-optimum solutions, constructive heuristics (CH) with several techniques such as modified path-scanning, switching rule, route compactness rule and shortest route rule are implemented. The performance of CH is tested on four real-life instances and a set of established benchmark dataset. In conclusion, the CH is able to produce optimum or near-optimum results in terms of routing cost and trips for a fleet of vehicles within a very fast computation time and is stable without any variation when compared to other methods, such as tabu search, reactive tabu search and column generation.