A parallelised large neighbourhood search heuristic for the asymmetric two-echelon vehicle routing problem with swap containers for cargo-bicycles

A Variable Neighborhood Search Algorithm with Constraint Relaxation for the Two-Echelon Vehicle Routing Problem with Simultaneous Delivery and Pickup Demands

“…Note that the classic 2E-VRP assumes that each satellite can be visited by more than one first phase delivery route; however, in the 2E-VRPSDP, a satellite is limited to service from no more than one vehicle. Thus, we select twenty 2E-VRP benchmark instances in which the best-known solutions (BKS) of the 2E-VRP from the literature (Breunig et al 2016, andMühlbauer et al 2021) do not assign more than one vehicle to visit a satellite. To fully compare with other 2E-VRP methods, the results of the following existing state-of-the-art algorithms are presented, i.e., the large neighborhood search-based heuristic (LNS-2E) of Breunig et al (2016), the adaptive large neighborhood search heuristic (ALNS) of Hemmelmayr et al (2012), the hybrid heuristic (GRASP+VND) from Zeng et al (2014), and the parallelized large neighborhood search heuristic (PLNS) from Mühlbauer et al (2021) when applicable.…”

“…The algorithm of Zeng et al (2014) is carried out using a single core of an Intel Pentium Dual-Core E5500 processor 2.8 GHz and 2 GB of memory. The PLNS in Mühlbauer et al (2021) is performed on a computer with 8G RAM and an Intel i5-6200 processor with 2.4 GHz, two cores and four threads. The PLNS is a parallelized algorithm, which is executed on the four threads in parallel.…”

“…Thus, we select twenty 2E-VRP benchmark instances in which the best-known solutions (BKS) of the 2E-VRP from the literature (Breunig et al 2016, andMühlbauer et al 2021) do not assign more than one vehicle to visit a satellite. To fully compare with other 2E-VRP methods, the results of the following existing state-of-the-art algorithms are presented, i.e., the large neighborhood search-based heuristic (LNS-2E) of Breunig et al (2016), the adaptive large neighborhood search heuristic (ALNS) of Hemmelmayr et al (2012), the hybrid heuristic (GRASP+VND) from Zeng et al (2014), and the parallelized large neighborhood search heuristic (PLNS) from Mühlbauer et al (2021) when applicable. The LNS-2E of Breunig et al (2016) is insensitive to its parameters, which are presented in Table 2 of Breunig et al (2016).…”

Van-based robot hybrid pickup and delivery routing problem

“…The 2E-VRP involves a two-layered distribution network: In the first level, vans perform deliveries from distribution centers to urban consolidation centers, also called satellites. At the second level, customers' orders are consolidated into small vehicles that can travel along any street in the city center area, such as minivans, electric vans, and cargo bikes, and delivered to the final customers (Liu et al 2018, Dellaert et al 2019, Li et al 2020, Mühlbauer and Fontaine 2021, Perboli et al 2021.…”

“…Since our problem is composed of pickup and delivery, two-echelon, van no-go zones, and VRPD components, we draw on some operations used in (Ropke and Pisinger 2006, Mühlbauer and Fontaine 2021, Anderluh et al 2021, Sacramento et al 2019). In our problem, we have parking stations for the vehicle, battery, and cargo transshipment, and the parking station is an independent station that can be visited multiple times.…”