Multiobjective car relocation problem in one-way carsharing system

Abstract

In this paper, we present a multiobjective approach for solving the one-way car relocation problem. We fix three objectives that include the number of remaining rejected demands, the number of jockeys used for the relocation operations, and the total time used by these jockeys. For this sake, we propose to apply two algorithms namely NSGA-II and an adapted memetic algorithm (MA) that we call MARPOCS which stands for memetic algorithm for the one-way carsharing system. The NSGA-II is used as a reference to compare the performance of MARPOCS. The comparison of the approximation sets obtained by both algorithms shows that the hybrid algorithm outperforms the classical NSGA-II and so solutions generated by the MARPOCS are much better than the solutions generated by NSGA-II. This observation is proved by the comparison of different quality indicators' values that are used to compare the performance of each algorithm. Results show that the MARPOCS is promising to generate very good solutions for the multiobjective car relocation problem in one-way carsharing system. It shows a good performance in exploring the search space and in finding solution with very good fitness values.

Modeling and optimization for carsharing services: A literature review

“…By incorporating a discrete choice model that depicts users' mode choice, Jian et al (2018) proposed a mixed-integer nonlinear programming model linking the supply and the demand to solve the vehicle relocation problem for stationbased CSSs. Zakaria et al (2018) presented a multi-objective integer linear programming model for solving the one-way carsharing relocation problem. In order to allow substantially longer reservation times while keeping the system profitable and achieving high service quality, Molnar and de Almeida Correia (2019) proposed a relocation-based reservation enforcement method combining vehicle locking and relocation movements.…”

Optimal Target Assignment with Seamless Handovers for Networked Radars

“…Many different MILP scheduling studies use the greedy algorithm together to compare performance [6,7]. Conversely, MILP formulations can sometimes be used to compare the performance of specially designed greedy algorithms [8,9,10].…”