6533b851fe1ef96bd12a8c2f
RESEARCH PRODUCT
Automating the Parameter Selection in VRP: An Off-line Parameter Tuning Tool Comparison
Nysret MusliuJussi RaskuTommi Kärkkäinensubject
Set (abstract data type)Computer scienceVehicle routing problemCombinatorial optimizationTunerControl engineeringRouting (electronic design automation)AlgorithmTravelling salesman problemSelection (genetic algorithm)Task (project management)description
Vehicle route optimization is an important application of combinatorial optimization. Therefore, a variety of methods has been proposed to solve different challenging vehicle routing problems. An important step in adopting these methods to solve real-life problems is to find appropriate parameters for the routing algorithms. In this chapter, we show how this task can be automated using parameter tuning by presenting a set of comparative experiments on seven state-of-the-art tuning methods. We analyze the suitability of these methods in configuring routing algorithms, and give the first critical comparison of automated parameter tuners in vehicle routing. Our experimental results show that the tuning methods are able to effectively automate the task of parameter configuration of route optimization systems. Moreover, our comparison shows that while routing algorithms clearly benefit from parameter tuning, and while there is no single tuner which consistently outperforms others, the tuning performance can be clearly improved with careful choice of a tuning method.
| year | journal | country | edition | language |
|---|---|---|---|---|
| 2014-01-01 |