This paper examines how government data regulations influence advanced vehicle firms’ pricing and product quality decisions, consumers’ data-sharing behavior, and social welfare. We develop a game-theoretic framework involving the government, an advanced vehicle firm, and consumers under three scenarios: (i) consumers do not share data, (ii) consumers share data without regulation, and (iii) consumers share data with government regulation. The analysis yields several insights. First, while government regulation enhances privacy protection, it also restricts the advanced vehicle firm’s ability to utilize consumer data, relative to the unregulated data-sharing scenario, resulting in lower product quality and price. Second, when either the cost coefficient or the privacy loss is very low or very high, data sharing leads to higher profits and greater consumer surplus than no sharing. However, at intermediate levels, the increasing cost of personalization outweighs its benefits, making non-sharing more preferable. Third, moderate regulation can alleviate privacy concerns while preserving data value. Although stronger data regulation initially enhances consumer trust, excessive regulation increases technological costs, thereby reducing overall social welfare. Based on this, we further extend our model by allowing consumers to partially share data. A counterintuitive finding is that more consumer data does not necessarily increase firm profit. This indicates that, compared with mandatory full data sharing, data-sharing flexibility can improve consumers’ participation incentives by enabling them to better trade off personalization benefits against privacy loss, thereby expanding effective demand and potentially increasing firm profit despite reduced data collection.
Traffic forecasting, as a cornerstone task in intelligent transportation systems, plays a vital role in alleviating urban congestion, optimizing traffic management, and improving travel efficiency. However, in real-world traffic systems, smooth and continuous variations driven by routine commuting patterns often coexist with abrupt disturbances caused by accidents, signal transitions, and unexpected events. Although capturing such mixed dynamics is of significant practical importance, it has not been fully explored in existing studies. To address this challenge, we propose a spatial-temporal graph learning framework (ST-GINDE) incorporating an impulse-like modulation neural ordinary differential equation module. The model formulates traffic dynamics as a continuous-time evolution process on spatial-temporal graphs, enabling unified modeling of temporal evolution, spatial dependencies, and adaptive responses to significant traffic state variations. To capture abrupt disturbances, we introduce an impulse-like modulation mechanism that adaptively adjusts the continuous dynamics when significant traffic variations occur, without introducing discontinuous state jumps or explicit hybrid system assumptions. Furthermore, ST-GINDE integrates multi-scale temporal feature extraction and hybrid dynamic graph learning to jointly model short-term fluctuations, long-term periodic patterns, and time-varying spatial dependencies across traffic networks. We further analyze the mathematical properties of the constructed dynamic system to verify the rationality of the proposed modeling framework. Extensive experiments on multiple real-world traffic datasets demonstrate that ST-GINDE achieves highly competitive performance in traffic forecasting tasks, with notable improvements in prediction accuracy and robustness under complex traffic conditions and sudden disturbances.
To better satisfy the uneven spatiotemporal distribution of passenger demand on the Y-type metro line, the concept of flexible train composition is explored, where the train compositions can be changed flexibly at a joint station and terminal stations connected to depots. A mixed-integer nonlinear programming model is developed to jointly optimize train timetables, rolling stock circulation plans, and flexible train compositions on a Y-type metro line. Using standard linearization techniques, the model is reformulated as a mixed-integer linear programming (MILP) model with an objective that balances operating cost and passenger service quality. To incorporate short-term demand variations and support real-time operations, the integrated MILP is embedded into a Model Predictive Control (MPC) framework, which updates timetabling decisions dynamically based on demand forecasts. However, solving the resulting MPC-based MILP in real time is computationally challenging due to the large number of binary decision variables. To overcome this limitation, we propose a learning-based optimization framework that integrates offline learning with online mixed-integer optimization. The learning module is trained offline on historical MPC solutions and is designed to selectively predict a subset of high-impact binary decisions, including train composition and routing choices, thereby reducing the combinatorial complexity of the online optimization problem. The predicted decisions are then fixed in an online MILP to optimize the remaining variables while preserving feasibility and consistency across rolling horizons. In addition, a feasibility-efficiency balancing strategy is introduced through selective decision prediction and feasibility-aware penalized training, which reduces infeasible learning outputs. Numerical experiments based on real-world operational data from Guangzhou Metro Line 14 show that the proposed learning-based MPC framework achieves real-time timetabling, while maintaining solution quality comparable to a full MPC-based optimization benchmark, with only a marginal loss in feasibility. Computational results demonstrate the effectiveness of integrating learning and MPC for real-time timetable optimization.
Cross-border supply chain disruptions are increasingly posing significant challenges to global operations. However, their impact on firm resilience, particularly for small and medium-sized enterprises (SMEs), has been insufficiently investigated. This study examines how SMEs react to cross-border supply chain disruptions. Drawing on a panel sample of 10,501 firm-quarter observations of SMEs listed on China’s ChiNext market, we find that cross-border supply chain disruptions negatively affect SMEs’ resilience. Furthermore, this effect is significantly mitigated when SMEs effectively orchestrate three key resources: alternative suppliers (relational resources), governmental financial support (financial resources), and digital technology (technological resources). Our study advances logistics and transportation literature by uncovering the specific effects of cross-border supply chain disruptions on SMEs’ resilience. Moreover, our study extends resource orchestration theory into the context of cross-border supply chain disruption by showing that SMEs build resilience through orchestrating relational, financial, and technological resources.
Efficient delivery of humanitarian supplies is crucial for disaster loss reduction. The Truck-Drone Collaborative Routing Problem (TDRP) has mostly been studied under static demand and ideal road networks, neglecting the dynamic revelation of demand points and nonlinear road capacity degradation from disaster damage. This paper introduces the TDRP with Road Damage and Dynamic Demand (TDRP-RD-DD) to minimize global delivery completion time. We build a full-chain model linking disaster intensity, road capacity, and truck travel speed, and develop an event-driven two-layer rolling optimization framework with locally optimal selection criteria for supply and replenishment modes. To solve this NP-hard problem, we design a Rolling Replenishment-aware Adaptive Large Neighborhood Search (RR-ALNS) with three specialized destroy-repair operators, a capacity- and mode-aware Warm-start strategy, and a replenishment-prioritized repair mechanism for infeasible solutions. Experiments on Solomon instances and a real-world disaster-relief case in Hubei Province show that RR-ALNS closely matches CPLEX on CPLEX-proven small-scale instances, with a 0.47% average gap for the default simulated-annealing setting, while scaling effectively to larger dynamic settings. Compared with three baseline algorithms, RR-ALNS achieves stronger feasibility and overall lower makespan in both benchmark and real-world tests. Ablation and sensitivity analyses further support the value of truck-drone collaboration, collaborative replenishment, and explicit road-damage-to-speed modeling. This work extends dynamic truck-drone routing to post-disaster settings with road damage and demand revelation, and provides quantitative support for emergency logistics decision-making.
In the transmodal seafood cold chain, seafood perishability and uncertainty in consumers’ freshness perceptions create a trade-off between freshness preservation and sales revenue. This study considers the channel choice behaviors influenced by consumers’ perception of freshness and develops a joint optimization model for fresh-keeping transportation and pricing based on freshness-decay dynamics to determine the fresh-keeping strategy, transportation routing, and pricing scheme with the objective to maximize retailers’ sales profit. Seafood freshness indicators are first calculated using kinetic models parameterized with experimental data obtained under controlled time–temperature conditions in existing studies. A back propagation neural network is then adopted to establish the nonlinear mapping between these physicochemical indicators and perception of freshness based on corresponding sensory evaluation data, while a nested logit model is used to capture the price-freshness trade-offs during selling. It is revealed through numerical analysis that: (1) E-commerce platforms will adopt higher transportation temperatures (3.05°C) to reduce costs, enabling aggressive pricing (54.76 CNY/kg), but contribute only about 5.72% of the retailer’s total profit; (2) Brick-and-mortar stores will implement stricter temperature control (0.37°C) leveraging location advantages, supporting premium pricing despite incurring fresh-keeping costs about 4.30 times as high as those of the e-commerce platform; (3) AVS is priced higher than SM (129.05 vs. 109.54 CNY/kg), reflecting its added service value as a store-affiliated online-to-home channel; (4) Ambient temperature changes induce coordinated operational adjustments, with total system profit changing by only 0.06% under both low- and high-temperature scenarios, while higher ambient temperatures impose greater fresh-keeping pressure. The model provides actionable transportation protocols: temperature-driven cargo allocation, refrigeration investment benchmarks, and dynamic pricing tariffs aligned with real-time freshness metrics.
This study proposes a reverse logistics framework by introducing a value-preserving repurchase (VPR) policy for electric vehicle (EV) makers. Under VPR, users pay a service fee and retail price and can return EVs after a specified period. To motivate returns, EV makers guarantee refunds exceeding typical secondhand prices. EV makers can offer repurchase, exchange (crediting the refund toward a new EV), or mixed options. However, market uncertainty can inflate secondhand prices beyond contract refunds, driving users to sell directly in secondary markets instead of returning. This undermines EV makers’ value recovery and new-vehicle demand. While the existing literature focuses on post-purchase returns from valuation discrepancies, we integrate volatile secondhand market prices into the VPR design. Methodologically, we develop an analytical framework for the EV reverse logistics channel, characterize the participants’ utilities through three VPR modes, and derive how secondhand prices reshape optimal VPR terms under different market structures. Furthermore, the EV makers’ mode preferences across monopoly and competitive settings are identified, thereby providing a rigorous analytical foundation for when and why different VPR options are chosen. From a managerial perspective, the results offer actionable guidance for EV makers designing VPR contracts under uncertain resale values: in a monopoly, the exchange mode emerges as dominant, and the mixed mode exhibits subtle efficiency; whereas the mixed mode prevails under competition, even it does not dominate in a monopolistic market. Overall, the findings extend beyond EVs by offering a transferable framework for value-preserving trade-in policy design in other industries facing uncertain secondhand prices.
Freight transport demand emerges from interdependent sequences of activities unfolding across space and time, yet most large-scale models continue to represent freight as independent trips or exogenous origin-destination flows. This paper proposes a spatially explicit dynamic Bayesian network (DBN) that generates complete, synthetic freight activity chains while embedding location choice directly within the generative process. Rather than assigning destinations in post-processing, spatial behaviour arises endogenously through learned dependencies between vehicle attributes, activity context, land use, and prior locations.A central contribution is the hierarchical spatial encoding strategy, integrating a multi-resolution discrete global grid system (H3) directly into the probabilistic graphical model. By representing relative positions within nested hexagons as interdependent categorical variables, the approach captures spatial dependence across scales while keeping conditional probability tables tractable. This enables scalable, privacy-preserving synthesis of high-resolution activity chains without reproducing identifiable trajectories.The framework is demonstrated using nationwide GNSS freight telemetry from Belgium. Out-of-sample evaluation shows that the model reproduces key structural, temporal, and spatial characteristics of observed freight behaviour, including chain length and duration, origin-destination flows, border-crossing dynamics, and both global and local spatial autocorrelation. By jointly modelling sequencing, timing, and spatial structure in a transparent probabilistic framework, the proposed approach advances generative freight demand modelling and provides a flexible foundation for policy and scenario analysis and agent-based simulation in data-sensitive environments.
This study examines how CCR schemes targeting ride-sourcing services influence multi-modal travel behavior in systems with public transit, ride-sourcing, and private vehicles to promote green transportation and reduce congestion and emissions. By designing Credit-Charge-cum-Reward scheme targeting ride-sourcing services, the study seeks to balance government fiscal revenue, intervention objectives, and ride-sourcing platform profitability. To rigorously capture the policy-constrained interaction between the ride-sourcing platform and travelers, a bi-level Stackelberg game-theoretic framework is developed. Within this framework, the government establishes the CCR policy environment as an exogenous intervenor. Under these policy constraints, the ride-sourcing platform acts as the leader in the upper-level problem by optimizing pricing and driver supply strategies, while travelers act as followers in the lower-level problem by making equilibrium mode choices to minimize generalized travel costs. The results demonstrate that CCR schemes effectively shift travelers toward green modes and reduce system-wide emissions while maintaining overall traffic efficiency. Interventions targeting public transit and private cars can incentivize a modal shift from high-emission private driving to ride-sourcing services. Furthermore, the study reveals the complementary functions of different credit mechanisms under fiscal neutrality, where interventions on private vehicles and transit directly determine emission reduction while those on ride-sourcing ensure fiscal neutrality in the regular equilibrium regimes, and incentive-based strategies further enhance platform profitability. Consequently, the proposed mechanism effectively reconciles the conflicts among government fiscal balance, platform viability, and traffic efficiency improvements.
The rapid development of autonomous vehicle (AV) technology is ushering in a new era of transportation characterized by the mixed-autonomy traffic (or mixed traffic for short), where AVs coexist and interact with human-driven vehicles (HVs). To harness the full potential of this transition, substantial research efforts over the past decade have been devoted to characterizing and managing the mixed traffic in road networks. This study provides a timely, comprehensive, and critical review of the state-of-the-art in mixed traffic network modeling and optimization. Specifically, we focus on studies that address human mobility in general congestible road networks, with an emphasis on macroscopic modeling or management of network-wide mixed traffic involving privately owned AVs and HVs. The literature search covers 136 academic journal articles published up to December 31, 2025. We classify the models developed in these studies into two main categories: mixed traffic analysis models and mixed traffic management & control models. Mixed traffic analysis models aim to investigate and predict how mixed traffic operates, evolves, or distributes over a network. We investigate these studies from three temporal dimensions: static, day-to-day dynamic and within-day dynamic. We examine the heterogeneity between AVs and HVs at the network level, translate it into ten AV-related behavioral assumptions, and highlight how these assumptions challenge and reshape modeling paradigms. Mixed traffic management & control models, on the other hand, focus on the determination of optimal management and control strategies. We overview cutting-edge strategies, including infrastructure retrofitting, participatory AV control, pricing and subsidization, among others. We elucidate how the strategy design problems are formulated and solved, and further discuss the application scenarios, advantages, limitations, and implementation barriers of these strategies. Finally, we identify critical research gaps and shed light on several inspiring directions for future investigation. This study aims to offer valuable insights for both academic research and practical decision-making regarding the deployment of AVs.
Generative Engine Optimization (GEO) is emerging as a strategic issue in AI-mediated retail supply chains. As conversational shopping systems increasingly mediate product search, comparison, and recommendation, firms compete not only for ranked-list positions but also for being surfaced, cited, or highlighted within compressed answer interfaces. We develop a two-stage analytical Hotelling duopoly in which retailers first invest in machine-legible GEO signals and then compete in prices, while a fraction of consumers purchase through a GenAI answer engine. The model endogenizes how probabilistic answer inclusion shapes demand allocation, pricing, welfare, fulfillment-capacity payoffs, and platform governance. The analysis yields several major findings. A higher highlighting probability raises both demand share and pricing power. Yet when retailers are initially symmetric, GEO creates a defensive arms race: both firms spend to preserve answer visibility, but relative visibility, prices, allocations, and consumer surplus remain unchanged, while total welfare declines. When retailers differ in inherited algorithmic visibility capital, the content-rich retailer’s advantage may persist and be reinforced under the maintained equilibrium conditions. Extensions show that fulfillment congestion attenuates the profit value of AI-steered demand, and that quality-weighted inclusion rules can reduce low-signal escalation by tying marginal GEO effectiveness to verifiable retail and operational quality. The paper contributes to research on retail supply-chain competition by showing that GenAI answer interfaces are upstream demand-allocation mechanisms rather than merely digital-marketing channels; their economic value depends on logistics execution, service reliability, and platform scoring design.
Offshore parcel delivery is essential for ensuring the sustained operation of offshore oil and gas platforms by providing timely access to essential supplies, spare parts, and emergency materials. We study a vessel routing problem emerging from a novel delivery paradigm in which a vessel and multiple unmanned aerial vehicles (UAVs) collaborate to complete delivery tasks. The studied problem extends the carrier–vehicle traveling salesman problem by allowing a vessel to carry multiple UAVs. We first formulate a mixed-integer second-order cone program that effectively captures the interdependent vessel–UAV operations by representing take-off and landing constraints using two operational points for each platform. To handle practical-sized instances, we design a tailored parallel adaptive large neighborhood search (PALNS) algorithm. The PALNS algorithm begins with a heuristic procedure to generate initial feasible solutions, incorporates problem-specific destroy and repair operators, and integrates local search mechanisms that explore various take-off and landing modes for multiple UAVs during the iterative process. Computational experiments on randomly generated instances show that CPLEX is effective for small-sized instances but fails to obtain feasible solutions for larger instances within the 3-hour time limit. Compared with the improved simulated annealing (ISA) baseline algorithm, PALNS reduces the average objective value from 27399.47 to 26286.25 on large-sized instances, corresponding to an average improvement of approximately 4.1%, while also requiring less computational time. In addition, we conduct a sensitivity analysis to evaluate the impacts of key parameters, including the number of UAVs, UAV endurance, and the speeds of both the vessel and UAVs. Finally, a case study is conducted to illustrate the effectiveness and practicality of the proposed method in offshore logistics.
Metro systems are central to urban mobility in dense cities, yet service disruptions often occur as partial capacity degradations whose impacts propagate through the network and depend on system design and operations. Existing studies mainly follow two perspectives: ex-post recovery, which reacts after disruptions but is constrained by fixed designs and resources; ex-ante planning, which embeds resilience in advance but abstracts from operational flexibility. This paper proposes an Event-Adaptive Resilience-Aware (EARA) framework for coordinated bus service design and response in multimodal transit systems. The framework provides a single implementable baseline design while enabling flexible adjustments once a disruption is identified. To avoid the computational burden of a full two-stage stochastic program, we embed disruption risk into planning through service-reliability-based effective capacity constraints and prepare event-specific frequency adjustments and shuttle deployment without changing the network structure. Disruption uncertainty is represented by a finite set of events specifying jointly affected metro segments, allowing heterogeneous disruption patterns to be captured without enumerating detailed capacity realizations. A Hong Kong Island case study compares four planning and response strategies. Results show that reactive adjustments cannot compensate for disruption-blind baseline designs, while disruption-aware planning without event differentiation leads to conservative capacity allocation. By combining disruption-aware planning with event-adaptive response, EARA substantially reduces unmet demand and improves overall system performance.
Robust navigation of automated guided vehicles (AGVs) in indoor logistics environments remains challenging, as AGVs must operate in confined, cluttered, and dynamically changing spaces while satisfying strict requirements for safety and efficiency. This paper focuses on three key problems: (1) balancing fast goal-directed motion with collision avoidance, (2) extracting transferable spatiotemporal representations from heterogeneous LiDAR and kinematic observations, and (3) improving the generalization of learned policies from simulation to unseen and real-world indoor logistics scenarios. To address these problems, an imitation-augmented hierarchical deep reinforcement learning framework with a Memory-Attention Generalized Network, termed IAHDRL-MAGNet, is proposed. It decomposes navigation into a low-level goal-directed motion policy and a high-level safety-aware option selector to balance rapid target approaching and obstacle avoidance. MAGNet integrates angle-aware multi-head attention, GRU-based temporal fusion, and D2RL feature propagation to extract compact spatiotemporal representations from multimodal LiDAR and kinematic observations. An augmented imitation learning mechanism further combines state-dependent expert guidance, prioritized experience replay, and limited randomization of LiDAR and kinematic observations to improve convergence efficiency and policy transferability. Simulation experiments in training and unseen indoor environments show that IAHDRL-MAGNet outperforms baseline methods by achieving higher success rates, fewer collisions, and shorter navigation times in successful trials. Ablation studies provide supporting evidence for the contributions of MAGNet and the augmented imitation learning mechanism. Real-world tests further show that the learned policy can be deployed without additional real-world fine-tuning under the tested settings. These results indicate that IAHDRL-MAGNet shows promise as a practical navigation framework for safe and efficient AGV navigation in indoor logistics environments.
Accuracy and timeliness of forestry monitoring are critical for reliable resource assessment and early warning of pest and disease outbreaks. While drone technologies are increasingly used for forestry monitoring, current applications often lack systematic planning. Monitoring operations rely heavily on manual experience, resulting in low efficiency, redundant coverage, and coverage gaps. To address the trade-off between coverage completeness and operational timeliness in large and topographically complex forests, we investigate a truck-drone collaborative forestry monitoring framework that jointly optimizes vehicle routes, candidate docking points, and unmanned aerial vehicle (UAV or drone) coverage paths. The dual objectives of our model are to maximize coverage and minimize total operational time subject to road-network and engineering constraints. To this end, the monitoring region is gridded in a projected coordinate system. Non-monitoring units (e.g., villages, farmlands) identified through remote sensing and field surveys are excluded to reduce redundant searches and monitoring. Road network data serves as constraints for vehicle access and docking point selection. We propose a Fix-and-Optimize (FO) iterative decomposition to decouple and alternately solve two coupledsubproblems: docking-point combination selection and sub-region partitioning. Under fixed task region division, clustering-based sampling and variable neighborhood search optimization are performed for docking point combinations. Under fixed docking points, isolated block detection and local redistribution are conducted for sub-regions, and a genetic algorithm is used to optimize the heading angles of boustrophedon flight paths for efficient coverage routes. To improve the solution efficiency for large-scale cases, we devise a mechanism for maintaining and parallel evaluating the Pareto front. In the case study of Qingyuan, China, the proposed method reduces total operational time by 30.06 % at the same coverage level compared with the current manual planning scheme. The Pareto front also reveals a clear trade-off: coverage gains diminish sharply beyond approximately 55 UAV flights, and the 55-flight scheme achieves 96.6 % coverage while reducing the number of sorties by 16.7 % with only a 3.4 % coverage loss relative to the 66-flight full-coverage scheme. Our framework provides a quantitative tool for decision-makers to balance coverage goals with resource constraints, enabling more efficient and effective large-scale forestry monitoring.
This study investigates the integration of crowdshipping (CS) into e-commerce reverse logistics, specifically applied to the parcel transport stage between return points and transhipment facilities. Several barriers have limited CS adoption in forward logistics, such as trust, privacy, and security concerns due to direct customer–crowdshipper interaction, and the inability to offer bundled CS tasks that would enable attractive compensation schemes, but these constraints are mitigated in this transport-stage-focused reverse logistics setting. The proposed CS-integrated system operates in a business-to-business (B2B) context, eliminating customer contact and allowing bundled CS task assignment, thereby providing a more secure and operationally efficient setting for CS deployment. In this study, operational and external costs (i.e., per-kilometre social costs) for both conventional and CS-integrated reverse logistics systems are modelled and simulated across diverse spatial and demographic contexts, system scales, configurations, and levels of CS supply and behavioural dynamics. These simulations generate a large database that identifies the conditions under which the CS-integrated system can achieve operational and/or environmental advantages compared with the conventional system. Using a comprehensive simulation experiment based on systematically generated parameter combinations (45,000 simulation runs) and applying decision tree analysis, we find that the CS-integrated system reduces operational costs in nearly all cases, with savings reaching up to 70 %, particularly under conditions with sufficient CS supply and high participation willingness (i.e., low sensitivity to detour and compensation). However, environmental benefits are more context-dependent, emerging mainly in low-density, large-area settings where detours remain minimal. The study underscores the potential of CS to enhance the cost efficiency of reverse logistics while highlighting the importance of policy interventions that limit detours and induced travel demand in the CS-integrated system to ensure that sustainability gains are fully realised.
Unmanned surface vehicles (USVs) are increasingly deployed in modern logistics and transportation systems, supporting applications such as port surveillance, offshore logistics, and maritime infrastructure inspection. However, reliable navigation in dynamic marine environments remains challenging due to complex environmental disturbances and the conventional separation between path planning and motion control in existing approaches. To address these limitations, this paper proposes a marine-data-assisted distributed deep reinforcement learning (DRL) framework for two coupled USV navigation tasks: path planning and path following. Rather than assuming a single universal policy, the proposed framework introduces a task-aware shared learning architecture called DTD3-CPER. It formulates path planning and path following as task-specific Markov decision processes while coordinating their training, experience management, and environmental interaction within a common learning framework. Moreover, an ERA5-informed marine simulation environment is developed by integrating spatially varying ocean-current and wind-field data into a grid-based navigation model. A distributed Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm enhanced with categorical prioritized experience replay (CPER) is then designed to support task-aware experience management, improve sample utilization, and enhance training stability. Simulation experiments indicate that the proposed framework achieves competitive performance against mainstream DRL baselines and conventional control methods in terms of learning efficiency, navigation safety, tracking accuracy, and disturbance adaptability. The results suggest that integrating marine environmental data with task-aware distributed reinforcement learning is a promising direction for data-driven USV navigation. From a managerial perspective, the proposed framework provides insights for developing data-driven decision-support systems for autonomous maritime logistics and intelligent maritime transportation systems.
Precise modeling of electric vehicle (EV) energy consumption is fundamental to the efficient design and management of modern transportation systems. While physics-based models offer superior interpretability, they often struggle with limited adaptability to dynamic driving conditions and heterogeneous vehicle platforms. To bridge this gap, achieving real-time and accurate calibration of physical model parameters becomes essential. This paper proposes a novel two-stage Bayesian optimization framework that integrates Contextual Bayesian Optimization (CBO) and Transfer Bayesian Optimization (TBO). In the first stage, the CBO module learns a context-aware mapping between operating conditions and physical parameters within a source domain. In the second stage, the TBO module leverages the learned prior knowledge to achieve rapid adaptation to a target vehicle domain with minimal data requirements. We evaluate the proposed framework using real-world datasets from BMW i3 and Tesla Model 3. Experimental results demonstrate that the proposed framework achieves a per-second WMAPE of 17.27% in cross-condition scenarios. For cross-vehicle transfer, the primary out-of-sample evaluation on the held-out 70% of the Tesla trips yields a WMAPE of 24.82% and a total energy error of 12.65%. The CBO results further demonstrate rapid convergence under a limited online evaluation budget. This research provides a scalable and sample-efficient solution for high-fidelity energy modeling across diverse driving conditions and vehicle platforms.
This paper studies the multi-period line planning problem with vehicle rotation, a fundamental challenge in public transportation system design that integrates line selection, service frequency determination, and inter-period vehicle transfers. The temporal coupling induced by vehicle rotation generates strong interdependencies across planning periods, resulting in large-scale, highly combinatorial optimization models that are difficult to solve with existing mixed-integer programming methods. To address this complexity, we develop two exact decomposition-based solution frameworks. The first is a constraint satisfaction algorithm that iteratively relaxes and reinstates the vehicle flow-balance constraints, progressively reconstructing feasibility through a sequence of restricted master problems. The second is a logic-based Benders decomposition framework in which the feasibility verification stage is reformulated as a max-flow and min-cut problem. This reformulation enables the efficient identification of infeasibilities and the generation of strong feasibility cuts derived from the structure of the underlying network. Extensive computational experiments on three real public transportation networks, Istanbul Metrobüs, Athens Metro, and Quito Trolebús, demonstrate the efficiency and scalability of the proposed methods. The logic-based decomposition algorithm attains optimal solutions up to 99% faster than a state-of-the-art commercial solver and successfully solves instances with more than 300 lines that remain unsolved within 24 hours by conventional algorithms. The results confirm that embedding network-flow reasoning within a logic-based decomposition framework yields a computationally tractable and structurally coherent methodology for large-scale multi-period line planning with vehicle rotation.
Operational resilience is critical for suppliers to sustain competitive advantages amid volatile market environments. While prior research emphasizes firms’ own digital transformation as a driver of resilience, the spillover effects of customer-side digitalization on suppliers remain underexplored. Drawing on resource dependence and social network theories, this study examines the relationship between customer digital transformation and supplier operational resilience, as well as the moderating role of suppliers’ multiplex network structural embeddedness. Using data from Chinese listed firms and their supply chain relationships between 2018 and 2022, we find that customer digital transformation significantly undermines supplier operational resilience, indicating a negative downstream spillover in the digital transition. Further moderating effect analysis reveals that a supplier’s structural embeddedness in both the supply chain network and the executive interlock network can effectively mitigate this negative impact. Heterogeneity analysis shows that the adverse effect is more pronounced among non-state-owned enterprises, firms with highly substitutable products, and firms with lower digital transformation. Overall, this study advances research on digital transformation and operational resilience from a customer-driven perspective and highlights the role of network embeddedness in buffering external digital shocks.
