Iterated Greedy Algorithm for Vehicle Routing Problem With Parcel Lockers

The growth in e-commerce sales has been constant and, driven by the Covid-19 pandemic, has exceeded previous expectations. Owing to social isolation, parcel lockers - systems for goods collection - have become widely used to reduce physical contact between people involved in the delivery process. This resource has also become an important strategy for reducing costs, as it enables deliveries from multiple customers to be consolidated in the same location. In the context of smart cities, in which the optimization of urban transport plays a key role in building more sustainable and efficient environments, the inclusion of parcel lockers is essential. Thus, research in this area that considers their use aligns with the concept of smart cities, promoting the use of innovative technologies to improve mobility and urban logistics. Specifically in this context, the Vehicle Routing Problem With Parcel Lockers (VRPPL) is an extension of the Vehicle Routing Problem with TimeWindows (VRPTW), introducing the use of lockers, with the objective to reduce the total distance traveled. In this study, an algorithm based on Iterated Greedy (IG) and Random Variable Neighborhood Descent (RVND) was developed to reduce the costs and computational times presented in the literature. We selected a set of 168 well-known instances for the problem. Computational experiments and comparative analyses were performed. The developed algorithm achieved better results in 167 instances, with an average gap of −1.315%. The best execution times were obtained for all instances, thereby reducing the mean execution time from 240.293s to 0.104s.