Publications

Joint Vehicle Pricing and Repositioning for Shared Micromobility: Prediction, Robustness and Decision-dependency

Working Paper, 2025

This research proposes a data-driven RS model to address a multi-period, joint vehicle pricing and repositioning problem for shared micromobility. Specifically, historical data and side information are leveraged to predict uncertain demand with a linear prediction model. Based on the estimated distribution, a residual-based RS model addresses distributional ambiguity of random residuals, and an estimation-fortified RS model is developed to control prediction error at the expense of target loss.

Target-Oriented Robust Inventory Management in Electric Vehicle Battery Swapping Networks

Under Review in Transportation Science, 2025

This research aims to address the battery inventory management problem with uncertain demand that arises in an EV battery swapping-charging network. In this problem, decision-makers determine initial battery stocks and periodic network operations, including battery replenishment, fulfillment, and lateral transshipment. To tackle the curse of dimensionality and distributional ambiguity, we propose a utility-based Service Level Measure (SLM) and incorporate this measure to derive a target-oriented Robust Satisficing (RS) model. A rigorous mathematical proof is provided for the salient properties of the SLM and RS models. This work pioneers the examination of multi-stage robust decisions under novel RS framework.

Download Paper

Multi-objective energy-efficient hybrid flow shop scheduling using Q-learning and GVNS driven NSGA-II

Published in Computers & Operations Research, 2023

This paper proposes a novel model for multi-objective energy-efficient hybrid flow shop. Time-of-use electricity prices, power down strategy, and uniform parallel machines are considered. To solve the model, we propose a Q-learning and general variable neighborhood search (GVNS) driven non-dominated sorting genetic algorithm II (QVNS-NSGA-II). The novelty of the algorithm is that we incorporate Q-learning into GVNS to guide premium adaptive operator selection throughout the shaking and local search processes.

Recommended citation: Li, Peize, et al. "Multi-objective energy-efficient hybrid flow shop scheduling using Q-learning and GVNS driven NSGA-II." Computers & Operations Research 159 (2023): 106360.
Download Paper | Download Slides