The dynamic electric vehicle routing problem (DEVRP) is a variant of the classical dynamic vehicle routing problem, where electric vehicles (EVs) are employed to provide on-demand logistics services. We investigate the DEVRP with stochastic requests and proactive charging decisions, where an EV must optimize its route and energy management to maximize the number of served requests while adhering to constraints on battery capacity and depot deadline. We formulate the DEVRP as a route-based Markov decision process (MDP) model and develop two complementary solution approaches: (1) a perfect-information rollout (PIR) policy based on stochastic lookahead and (2) a deep Q-learning (DQN) policy based on neural networks. Compared with conventional stochastic lookahead policies, such as rollout, the PIR policy improves the quality of reward-to-go estimation by better utilizing stochastic information, and we design two adjustable acceleration mechanisms to improve its balance between efficiency and accuracy. Although the PIR policy performs all computation online and adapts well to various scenarios, the DQN policy relies on offline training of neural networks to achieve faster online decision making. We also propose enhancement techniques to improve the performance of DQN. Through extensive numerical experiments, we demonstrate the advantages of the proposed policies in terms of solution quality and computational efficiency over multiple state-of-the-art approaches (such as value function approximation and rollout). Compared with the best-performing benchmark approaches, our proposed policies improve the decision quality by up to 5.3% while reducing the online computation time by more than half. Moreover, the merits of proactively and partially charging the EV in the DEVRP are investigated and validated. Robustness experiments further demonstrate the effectiveness of the proposed policies when demand prediction is inaccurate. Funding: This research is supported by the National Natural Science Foundation of China (NSFC) [Grant 72301109]. Supplemental Material: The online appendices are available at https://doi.org/10.1287/trsc.2025.0476 .
Accurate forecasting of origin-destination (O-D) demand is critical for transportation network design, planning, and operational management. An underexplored challenge is inductive forecasting: predicting O-D flows for new or unobserved locations, which is essential for evaluating network expansions and new facility placements. However, most existing approaches either are transductive (limited to fixed observed networks) or are deep learning models that lack interpretability, fail to provide uncertainty quantification, and often overlook critical data constraints such as structural zeros. To this end, we develop a Bayesian spatiotemporal hierarchical model for probabilistic O-D demand estimation, designed specifically for inductive forecasting. We assume total trip generation is given and model the destination-choice counts using a multinomial distribution. To handle the high-dimensional probability tensor, which features both simplex and structural zero constraints, we parameterize it via a masked-centered softmax transformation of a latent utility tensor. We then model this latent utility tensor using a CANDECOMP/PARAFAC (CP) tensor factorization to parsimoniously capture spatiotemporal patterns. Crucially, we impose Gaussian process (GP) priors on the spatial and temporal latent factors; the GP over the network provides the principled statistical mechanism to make inductive predictions for new locations. For posterior inference, we propose an Markov chain Monte Carlo algorithm. We validate the proposed model on synthetic data and two real-world O-D data sets. Results confirm our model's ability to accurately estimate O-D flows and provide robust inductive forecasts with full uncertainty quantification, which is essential for robust decision making in downstream applications, such as stochastic network optimization and facility location problems.
Large Language Models (LLMs) offer significant potential by serving as human proxies to advance travel demand modeling, but their behavioral misalignment with human travelers remains a critical obstacle. Furthermore, existing alignment methods are often impractical or inefficient when applied to the sparse data sets typically available for travel choices, limiting the adoption of these powerful new tools. We introduce a novel framework to align LLMs with travel choice behavior. Our method first infers a set of traveler personas from empirical data and then estimates a persona loading function that uses learned embeddings to select the appropriate persona for an individual based on their sociodemographics. Validated on the Swissmetro mode choice data set, our approach significantly outperforms established benchmarks in predicting both aggregate and individual choice outcomes. Our research offers a more adaptable, interpretable, and resource-efficient pathway to robust LLM-based travel behavior simulation, paving the way to integrate LLMs into transportation modeling practice in the future.
We introduce a new variant of the Undirected Team Orienteering Arc Routing Problem (UTOARP) that incorporates three key features: required edges, capacitated vehicles, and multiple services. These features have been investigated individually in the literature but have not been considered simultaneously. In this problem, demand is placed at some edges of a given undirected graph, and served demand edges produce a profit. Feasible routes must start and end at a given depot, and there is a time limit on the maximum duration of each route and a capacity limit on the demand served by each vehicle. The problem asks for a set of up to |K|maximum profit routes while ensuring all required edges are served. We exploit optimality conditions for this problem and propose a new unified, undirected formulation with binary variables. We also introduce a logic-based Benders decomposition derived from this formulation, resulting in a new problem reformulation, and show how to strengthen the logic-based Benders cuts. Crucially, the structure of the Benders subproblems remains unchanged regardless of which of the above features are enabled, highlighting the modularity and flexibility of the approach. Furthermore, we design several new families of valid inequalities, where some of them are derived from conflict graphs. Extensive computational tests are conducted to examine the performance of the proposed formulations and valid inequalities under various settings. We further analyze the solution structure of a real-world instance to illustrate the practical impact of the different features. Our findings highlight the pivotal role of logic-based Benders decomposition and conflict graphs in solving the UTOARP, marking their first application in the context of arc routing problems to the best of our knowledge. Moreover, these techniques hold promise for advancing solution approaches in broader arc routing contexts.
This article may be used only for the purposes of research, teaching, and/or private study. Commercial use or systematic downloading (by robots or other automatic processes) is prohibited without explicit Publisher approval, unless otherwise noted. For more information, contact permissions@informs.org. The Publisher does not warrant or guarantee the article's accuracy, completeness, merchantability, fitness inclusion of an advertisement in this article, neither constitutes nor implies a guarantee, endorsement, or support
