With the rapid development of global trade, ports and terminals are playing an increasingly important role, and automatic guided vehicles (AGVs) have been used as the main carriers performing the loading/unloading operations in automated container terminals. In this paper, we investigate a multi-AGV dynamic scheduling problem to improve the terminal operational efficiency, considering the sophisticated complexity and uncertainty involved in the port terminal operation. We propose to model the dynamic scheduling of AGVs as a Markov decision process (MDP) with mixed decision rules. Then, we develop a novel adaptive learning algorithm based on a deep Q-network (DQN) to generate the optimal policy. The proposed algorithm is trained based on data obtained from interactions with a simulation environment that reflects the real-world operation of an automated in Shanghai, China. The simulation studies show that, compared with conventional scheduling methods using a heuristic algorithm, i.e., genetic algorithm (GA) and rule-based scheduling, terminal the proposed approach performs better in terms of effectiveness and efficiency.

As consumers’ green awareness continues to grow, the level of a product’s eco-friendliness and the quality of its marketing have become significant factors in shaping consumers’ purchasing decisions. The power structures within the supply chain, as well as corresponding government subsidy policies, are also key elements influencing sustainable strategies for the green supply chain. In a green supply chain comprising one manufacturer and one e-commerce platform, two sales models exist within the online e-commerce platforms: reselling and agency selling. This paper establishes and analyzes three distinct Stackelberg game models, namely: manufacturer-led model without subsidy (bm), manufacturer-led model with subsidy (sm), and platform-led model with subsidy (sp). The results are shown as follows, with the rise of consumers’ environmental awareness, more consumers opt for green products, inspiring the manufacturer to increase its optimal greenness and platform to enhance its optimal marketing efforts level. Notably, government subsidies provide a significant stimulus. An increase in the green technology cost coefficient leads to a decline in the manufacturer’s profits across all three modes. Intriguingly, the manufacturer’s profits are always highest in the sp mode. As the marketing efforts cost coefficient increases, the platform’s profits decrease in the bm and sm modes. In contrast, in the sp mode, the platform’s profits increase rather than a decrease. The choice of mode primarily depends on the platform’s marketing efforts cost coefficient. When this coefficient exceeds a threshold, the platform chooses the sp mode. However, due to the relatively low marketing efficiency in this scenario, the manufacturer prefers the sm mode. For the government, the sp mode involves agency selling, serves as an effective mechanism to redistribute subsidies, thereby yielding the maximum social welfare benefits. Management insights are provided for the manufacturer and platform managers to make decisions about the degree of greenness and marketing efforts level, along with insights for governments to optimize subsidy policies.

E-commerce companies generate massive orders daily, and efficiently fulfilling them is a critical challenge. In the “parts-to-picker” order fulfillment system, the joint optimization of order allocation and rack selection is a crucial problem. Previous research has primarily focused on these two aspects separately and has yet to consider the issue of workload balancing across multiple picking stations, which can significantly impact picking efficiency. Therefore, this paper studies a joint optimization problem of order allocation and rack selection for a “parts-to-picker” order picking system with multiple picking stations to improve order picking efficiency and avoid uneven workload distribution. An integer programming model of order allocation and rack selection joint optimization is formulated to minimize the racks’ total moving distance and to balance the orders allocated to each picking station. The problem is decomposed into three sub-problems: order batching, batch allocation, and rack selection, and an improved simulated annealing (SA) algorithm is designed to solve the problem. Two workload comparing operators and two random operators are developed and introduced to the SA iterations. Random instances of different scales are generated for experiments. The algorithm solutions are compared with those generated by solving the IP model directly in a commercial solver, CPLEX, and applying the first-come-first-serve strategy (FCFS), respectively. The numerical results show that the proposed algorithm can generate order allocation and rack selection solutions much more efficiently, where the moving distances of the racks are effectively reduced and the workloads are balanced among the picking stations simultaneously. The model and algorithm proposed in this paper can provide a scientific decision-making basis for e-commerce companies to improve their picking efficiency.

Due to external environmental factors, the layout of compound channels in estuarine ports is restricted. Moreover, with the opposing distribution of anchorages and terminals within the port, vessels navigating between these areas must cross the channel, severely affecting channel navigation safety and efficiency. In order to improve the efficiency of vessel scheduling, we analyze the layout characteristics of an estuarine port and its compound channel, summarize vessel navigation modes and traffic flow conflicts, and identify five key conflict areas. On this basis, we develop a multi-objective optimization model aimed at minimizing vessel waiting times and the total channel occupancy time ratio. This model incorporates constraints such as navigation rules, traffic flow conflicts, tidal effects, and traffic control. To solve the model, we propose an adaptive non-dominated sorting genetic algorithm, ANSGA-NS-SA, which integrates neighborhood search (NS) and Simulated Annealing (SA). The entropy-weighted technique for order preference by similarity to ideal solution (TOPSIS) is used to calculate the objective weights of the Pareto frontier and identify the optimal solution. Experimental results show that the proposed model and algorithm yield optimal port entry and exit scheduling solutions. In terms of port scheduling performance, the proposed model and algorithm outperform the traditional First-Come-First-Served (FCFS) strategy and the Non-Dominated Sorting Genetic Algorithm II (NSGA-II), reducing total vessel waiting time by 33.8% and improving total channel occupancy ratio by 8.8%. This study provides a practical and effective decision support tool for estuarine port scheduling, enhancing overall port operational efficiency.