The capacitated arc routing problem (CARP) is a difficult combinatorial optimization problem that has been intensively studied in the last decades. We present a hybrid metaheuristic approach (HMA) to solve this problem which incorporates an effective local refinement procedure, coupling a randomized tabu thresholding procedure with an infeasible descent procedure, into the memetic framework. Other distinguishing features of HMA include a specially designed route-based crossover operator for solution recombination and a distance-and-quality based replacement criterion for pool updating. Extensive experimental studies show that HMA is highly scalable and is able to quickly identify either the best known results or improved best known results for almost all currently available CARP benchmark instances. In particular, it discovers an improved best known result for 15 benchmark instances (6 classical instances and 9 large-sized instances whose optima are unknown). Furthermore, we analyze some key elements and properties of the HMA-CARP algorithm to better understand its behavior.
Paper
Full text
A hybrid metaheuristic approach for the capacitated arc routing problem
Semantic Scholar · Computer Science · 2016
Abstract
The capacitated arc routing problem (CARP) is a difficult combinatorial optimization problem that has been intensively studied in the last decades. We present a hybrid metaheuristic approach (HMA) to solve this problem which incorporates an effective local refinement procedure, coupling a randomized tabu thresholding procedure with an infeasible descent procedure, into the memetic framework. Other distinguishing features of HMA include a specially designed route-based crossover operator for solution recombination and a distance-and-quality based replacement criterion for pool updating. Extensive experimental studies show that HMA is highly scalable and is able to quickly identify either the best known results or improved best known results for almost all currently available CARP benchmark instances. In particular, it discovers an improved best known result for 15 benchmark instances (6 classical instances and 9 large-sized instances whose optima are unknown). Furthermore, we analyze some key elements and properties of the HMA-CARP algorithm to better understand its behavior.