Wenzhen Huang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Wenzhen Huang has a total of 61 co-authors. Here are the top co-authors by number of shared publications:
- Yong Li
Tsinghua University
12 shared publications - Haoqiang Liu
Department of Electronic Engineering, Tsinghua University, Beijing, China
8 shared publications - Tong Li
Department of Electronic Engineering, Beijing National Research Center for Information Science and Technology (BNRist), Tsinghua University, Beijing, China
8 shared publications - Qianyue Hao
Tsinghua University
3 shared publications - Weikang Su
Department of Electronic Engineering, Tsinghua University, Beijing, China
3 shared publications - Bin Wang
Huawei Noah's Ark Lab, Beijing, China
2 shared publications
Latest Publications
- 6 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. PP (00189545) , 1-16
L2S: Learning to Optimize Sleep Strategy for Energy-Efficient Mobile Networks
With surging wireless traffic and massive base station deployments, periodic demand fluctuations often cause energy waste during low-traffic periods. To address this issue, we consider minimizing energy consumption by dynamically adjusting the operational statuses of base station components in response to traffic demand. We formulate an optimization problem for base station energy consumption by modeling the power consumption of energy-consuming components (Base Band Units, Remote Radio Units, and air conditioners) according to their operational statuses (on/sleep/off) and traffic load while maintaining required service coverage. We then transform the optimization problem into the management of cell status, traffic load, and base station status. Our proposed method, Learning to Optimize Sleep strategy (L2S), employs deep reinforcement learning to refine the operational statuses generated by a diffusion model, and trains the diffusion model with these refined statuses. Starting from random operational statuses, this bidirectional iteration process progressively enhances the performance. The diffusion model ensures control smoothness by generating cell statuses conditioned on historical decisions. Evaluation of L2S under realistic base station distribution and traffic demand patterns demonstrates 58.6% energy savings with minimal switching frequencies (average frequencies of 0.009 and 0.018 per half hour for base station and cell switching, respectively).
- 6 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. PP (00189545) , 1-16
L2S: Learning to Optimize Sleep Strategy for Energy-Efficient Mobile Networks
With surging wireless traffic and massive base station deployments, periodic demand fluctuations often cause energy waste during low-traffic periods. To address this issue, we consider minimizing energy consumption by dynamically adjusting the operational statuses of base station components in response to traffic demand. We formulate an optimization problem for base station energy consumption by modeling the power consumption of energy-consuming components (Base Band Units, Remote Radio Units, and air conditioners) according to their operational statuses (on/sleep/off) and traffic load while maintaining required service coverage. We then transform the optimization problem into the management of cell status, traffic load, and base station status. Our proposed method, Learning to Optimize Sleep strategy (L2S), employs deep reinforcement learning to refine the operational statuses generated by a diffusion model, and trains the diffusion model with these refined statuses. Starting from random operational statuses, this bidirectional iteration process progressively enhances the performance. The diffusion model ensures control smoothness by generating cell statuses conditioned on historical decisions. Evaluation of L2S under realistic base station distribution and traffic demand patterns demonstrates 58.6% energy savings with minimal switching frequencies (average frequencies of 0.009 and 0.018 per half hour for base station and cell switching, respectively).
- 25 March 2026
- Published by Elsevier in Neurocomputing
- Vol. 683
Collaborative route optimization for efficient and constraint-aware express delivery using deep reinforcement learning
- 25 March 2026
- Published by Elsevier in Neurocomputing
- Vol. 683
Collaborative route optimization for efficient and constraint-aware express delivery using deep reinforcement learning
- 16 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Wireless Communications
- Vol. 25 (15361276) , 12112-12128
Joint Energy-Efficient and Throughput Optimization in Large-Scale Mobile Networks via Safe Hierarchical MARL
The rapid advancement of 5G networks has significantly increased the demand for high throughput and low latency, particularly for enhanced Mobile Broadband (eMBB) services. Optimizing user rates in large-scale base station (BS) collaborative scenarios is crucial, ensuring high throughput and low energy consumption while meeting the needs of users. The optimization process faces two key challenges: how to efficiently explore and utilize the vast action space for collaborative decision-making in large-scale networks with numerous BSs, and how to ensure optimal decision-making across numerous parameters while satisfying constraints. To tackle these challenges, we propose a novel solution that integrates a generative AI-powered Digital Twin platform with a multi-agent reinforcement learning (MARL) decision algorithm. An important tool is the high-fidelity DT-SimNet platform, enabling seamless integration with AI optimizers. Moreover, we propose a novel algorithm called Safe Hierarchical Multi-Agent Proximal Policy Optimization with Adaptive Prediction mechanism (Safe HMAPPO-AP). This algorithm integrates Safe MARL to ensure decision-making remains within constraint boundaries and incorporates a hierarchical structure to decompose the action space. Moreover, it leverages Graph Neural Networks (GNNs) for efficient feature extraction and employs the Adaptive Prediction mechanism to accelerate learning and improve network performance. Comprehensive experiments using the DT-SimNet platform demonstrate the superior performance of the proposed algorithm compared to competing methods. Within the target area, Safe HMAPPO-AP achieved a 6.73% improvement in average throughput for regular users and an 11.48% increase in system utility, while meeting demands of newly added eMBB users.
- 16 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Wireless Communications
- Vol. 25 (15361276) , 12112-12128
Joint Energy-Efficient and Throughput Optimization in Large-Scale Mobile Networks via Safe Hierarchical MARL
The rapid advancement of 5G networks has significantly increased the demand for high throughput and low latency, particularly for enhanced Mobile Broadband (eMBB) services. Optimizing user rates in large-scale base station (BS) collaborative scenarios is crucial, ensuring high throughput and low energy consumption while meeting the needs of users. The optimization process faces two key challenges: how to efficiently explore and utilize the vast action space for collaborative decision-making in large-scale networks with numerous BSs, and how to ensure optimal decision-making across numerous parameters while satisfying constraints. To tackle these challenges, we propose a novel solution that integrates a generative AI-powered Digital Twin platform with a multi-agent reinforcement learning (MARL) decision algorithm. An important tool is the high-fidelity DT-SimNet platform, enabling seamless integration with AI optimizers. Moreover, we propose a novel algorithm called Safe Hierarchical Multi-Agent Proximal Policy Optimization with Adaptive Prediction mechanism (Safe HMAPPO-AP). This algorithm integrates Safe MARL to ensure decision-making remains within constraint boundaries and incorporates a hierarchical structure to decompose the action space. Moreover, it leverages Graph Neural Networks (GNNs) for efficient feature extraction and employs the Adaptive Prediction mechanism to accelerate learning and improve network performance. Comprehensive experiments using the DT-SimNet platform demonstrate the superior performance of the proposed algorithm compared to competing methods. Within the target area, Safe HMAPPO-AP achieved a 6.73% improvement in average throughput for regular users and an 11.48% increase in system utility, while meeting demands of newly added eMBB users.
- 26 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Mobile Computing
- Vol. 25 (5) , 6472-6483
GenNet: A Generative AI-Driven Mobile Network Simulator for Multi-Objective Network Optimization With RL
Simulation-based optimization has emerged as a crucial methodology in the field of mobile network optimization, addressing the need for dynamic and predictive network management. To address the scarcity of open-source mobile network simulators for advanced research, we developed GenNet—a generative AI-driven mobile network simulator. GenNet can create virtual replicas of mobile users, base stations, and wireless environments, utilizing generative AI methods to simulate the behaviors of these entities under various network settings with high accuracy. GenNet features a tailor-made API explicitly designed for reinforcement learning environments, enabling researchers to finely adjust network parameters such as tilts, azimuth, and transmitting power. Extensive experiments have utilized GenNet to benchmark multi-objective optimization algorithms, with a focus on enhancing network coverage, throughput, and energy efficiency, thereby validating its effectiveness as a robust platform for advancing network optimization techniques. Through this innovative tool, we aim to empower researchers and practitioners to identify and implement the most effective approaches for network optimization, paving the way for future advancements in mobile network management.
- 26 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Mobile Computing
- Vol. 25 (5) , 6472-6483
GenNet: A Generative AI-Driven Mobile Network Simulator for Multi-Objective Network Optimization With RL
Simulation-based optimization has emerged as a crucial methodology in the field of mobile network optimization, addressing the need for dynamic and predictive network management. To address the scarcity of open-source mobile network simulators for advanced research, we developed GenNet—a generative AI-driven mobile network simulator. GenNet can create virtual replicas of mobile users, base stations, and wireless environments, utilizing generative AI methods to simulate the behaviors of these entities under various network settings with high accuracy. GenNet features a tailor-made API explicitly designed for reinforcement learning environments, enabling researchers to finely adjust network parameters such as tilts, azimuth, and transmitting power. Extensive experiments have utilized GenNet to benchmark multi-objective optimization algorithms, with a focus on enhancing network coverage, throughput, and energy efficiency, thereby validating its effectiveness as a robust platform for advancing network optimization techniques. Through this innovative tool, we aim to empower researchers and practitioners to identify and implement the most effective approaches for network optimization, paving the way for future advancements in mobile network management.
- 21 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Wireless Communications
- Vol. 25 (15361276) , 7617-7634
Coordinated Downlink Beamforming in Multi-Cell MIMO Networks: A Diffusion Model-Enhanced Multi-Agent Reinforcement Learning Perspective
To address the surge in wireless traffic, multiple-input multiple-output (MIMO) technology has become essential for advancing communication systems. Within MIMO cellular networks, coordinated beamforming (CBF), achieved through the collaborative design of beamformers across multiple base stations (BSs), presents a promising strategy for enhancing network performance. However, in large-scale multi-cell, multi-user MIMO environments, optimizing beamforming coordination remains challenging due to high-dimensional spaces and dynamic conditions. While centralized, optimization-based CBF algorithms provide near-optimal solutions, their dependence on real-time global channel state information (CSI) and high computational complexity renders them impractical for dynamic networks. To overcome these limitations, we introduce a diffusion model-enhanced multi-agent reinforcement learning (MARL) coordinated beamforming framework, termed Diffusion-enhanced MACBF, which enables BSs to determine optimal beamformers independently based on partial observations. Key innovations of this framework include: firstly, a novel limited-information exchange protocol that facilitates effective coordination among BSs with minimal communication overhead; secondly, the introduction of a diffusion model-enhanced Multi-agent Soft Actor-Critic (MASAC) algorithm that enables BSs to efficiently manage spatial and temporal variations in large-scale scenarios; and thirdly, an optimized network architecture combining an Encoder, Convolutional Neural Network (CNN), and Bidirectional Long Short-Term Memory (Bi-LSTM) layers to accurately map partial observations to optimal actions. Extensive simulations demonstrate the effectiveness and efficiency of the proposed algorithm, achieving an average rate of 7.5502 bps/Hz with up to 10.66% improvement over existing methods while significantly reducing information exchange requirements.
- 21 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Wireless Communications
- Vol. 25 (15361276) , 7617-7634
Coordinated Downlink Beamforming in Multi-Cell MIMO Networks: A Diffusion Model-Enhanced Multi-Agent Reinforcement Learning Perspective
To address the surge in wireless traffic, multiple-input multiple-output (MIMO) technology has become essential for advancing communication systems. Within MIMO cellular networks, coordinated beamforming (CBF), achieved through the collaborative design of beamformers across multiple base stations (BSs), presents a promising strategy for enhancing network performance. However, in large-scale multi-cell, multi-user MIMO environments, optimizing beamforming coordination remains challenging due to high-dimensional spaces and dynamic conditions. While centralized, optimization-based CBF algorithms provide near-optimal solutions, their dependence on real-time global channel state information (CSI) and high computational complexity renders them impractical for dynamic networks. To overcome these limitations, we introduce a diffusion model-enhanced multi-agent reinforcement learning (MARL) coordinated beamforming framework, termed Diffusion-enhanced MACBF, which enables BSs to determine optimal beamformers independently based on partial observations. Key innovations of this framework include: firstly, a novel limited-information exchange protocol that facilitates effective coordination among BSs with minimal communication overhead; secondly, the introduction of a diffusion model-enhanced Multi-agent Soft Actor-Critic (MASAC) algorithm that enables BSs to efficiently manage spatial and temporal variations in large-scale scenarios; and thirdly, an optimized network architecture combining an Encoder, Convolutional Neural Network (CNN), and Bidirectional Long Short-Term Memory (Bi-LSTM) layers to accurately map partial observations to optimal actions. Extensive simulations demonstrate the effectiveness and efficiency of the proposed algorithm, achieving an average rate of 7.5502 bps/Hz with up to 10.66% improvement over existing methods while significantly reducing information exchange requirements.
- Open Access
- 21 November 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Knowledge Discovery From Data
- Vol. 20 (1) , 1-25
Jointly Optimizing Deployment and Antenna of Base Stations Using Hierarchical Reinforcement Learning
The coordinated deployment of multiple Base Stations (BS) and tuning of antenna configuration plays a crucial role in ensuring high-quality communication services, especially in the context of dense 5G BS deployment in megacities. However, traditional optimization methods, such as heuristics and Reinforcement Learning (RL), face challenges in addressing such problems involving the coordination of hundreds of BSs due to their limitations in handling the complexity and scale of large-scale scenarios. To address these challenges, this article proposes the Hierarchical Multi-Agent Proximal Policy Optimization with Representation Learning (HMAPPO-RL). By employing a hierarchical structure, we effectively decouple the optimization problem into two sub-problems: BS deployment and antenna parameter tuning. Different from the step-by-step method of optimizing the BS location and antenna, HMAPPO-RL achieves joint optimization of the two problems through an ingenious interactive mechanism, fully considering the mutual influence of the BS location and antenna. To address the large-scale challenge posed by hundreds of BSs, we utilize the upsampling and downsampling mechanisms of the UNet network to integrate global and local information from large-scale state information for performance enhancement. Since complex environmental information will cause great difficulties for the agent to evaluate the state value in large-scale scenarios, we add a representation learning module to enhance the accuracy of the agent’s state value estimation. The experiments using a precise mobile network simulator demonstrate the superiority of the proposed HMAPPO-RL, offering a comparative analysis with existing state-of-the-art methods. HMAPPO-RL achieves a coverage rate of 91.66% and an average throughput of 4,983,537 bit/s. These results represent improvements of 3.62% and 6.75% in coverage rate and throughput, respectively, when compared with the MAPPO algorithm.
- 21 November 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Knowledge Discovery From Data
- Vol. 20 (1) , 1-25
Open AccessJointly Optimizing Deployment and Antenna of Base Stations Using Hierarchical Reinforcement Learning
The coordinated deployment of multiple Base Stations (BS) and tuning of antenna configuration plays a crucial role in ensuring high-quality communication services, especially in the context of dense 5G BS deployment in megacities. However, traditional optimization methods, such as heuristics and Reinforcement Learning (RL), face challenges in addressing such problems involving the coordination of hundreds of BSs due to their limitations in handling the complexity and scale of large-scale scenarios. To address these challenges, this article proposes the Hierarchical Multi-Agent Proximal Policy Optimization with Representation Learning (HMAPPO-RL). By employing a hierarchical structure, we effectively decouple the optimization problem into two sub-problems: BS deployment and antenna parameter tuning. Different from the step-by-step method of optimizing the BS location and antenna, HMAPPO-RL achieves joint optimization of the two problems through an ingenious interactive mechanism, fully considering the mutual influence of the BS location and antenna. To address the large-scale challenge posed by hundreds of BSs, we utilize the upsampling and downsampling mechanisms of the UNet network to integrate global and local information from large-scale state information for performance enhancement. Since complex environmental information will cause great difficulties for the agent to evaluate the state value in large-scale scenarios, we add a representation learning module to enhance the accuracy of the agent’s state value estimation. The experiments using a precise mobile network simulator demonstrate the superiority of the proposed HMAPPO-RL, offering a comparative analysis with existing state-of-the-art methods. HMAPPO-RL achieves a coverage rate of 91.66% and an average throughput of 4,983,537 bit/s. These results represent improvements of 3.62% and 6.75% in coverage rate and throughput, respectively, when compared with the MAPPO algorithm.
- Open Access
- 20 July 2025
- Published by Association for Computing Machinery (ACM)
- p. 1457-1468
CoopRide: Cooperate All Grids in City-Scale Ride-Hailing Dispatching with Multi-Agent Reinforcement Learning
- 20 July 2025
- Published by Association for Computing Machinery (ACM)
- p. 1457-1468
Open AccessCoopRide: Cooperate All Grids in City-Scale Ride-Hailing Dispatching with Multi-Agent Reinforcement Learning
- Top Cited PapersOpen Access
- 10 February 2025
- Published by Association for Computing Machinery (ACM) in ACM Computing Surveys
- Vol. 57 (6) , 1-43
Generative AI Empowered Network Digital Twins: Architecture, Technologies, and Applications
The rapid advancement of mobile networks highlights the limitations of traditional network planning and optimization methods, particularly in modeling, evaluation, and application. Network Digital Twins, which simulate networks in the digital domain for evaluation, offer a solution to these challenges. This concept is further enhanced by generative AI technology, which promises more efficient and accurate AI-driven data generation for network simulation and optimization. This survey provides insights into generative AI-empowered network digital twins. We begin by outlining the architecture of a network digital twin, which encompasses both digital and physical domains. This architecture involves four key steps: data processing and network monitoring, digital replication and network simulation, designing and training network optimizers, Sim2Real, and network control. Next, we systematically discuss the related studies in each step and make a detailed taxonomy of the problem studied, the methods used, and the key designs leveraged. Each step is examined with a focus on the role of generative AI, from estimating missing data and simulating network behaviors to designing control strategies and bridging the gap between digital and physical domains. Finally, we discuss the open issues and challenges of generative AI-based network digital twins.
- 10 February 2025
- Published by Association for Computing Machinery (ACM) in ACM Computing Surveys
- Vol. 57 (6) , 1-43
Top Cited PapersOpen AccessGenerative AI Empowered Network Digital Twins: Architecture, Technologies, and Applications
The rapid advancement of mobile networks highlights the limitations of traditional network planning and optimization methods, particularly in modeling, evaluation, and application. Network Digital Twins, which simulate networks in the digital domain for evaluation, offer a solution to these challenges. This concept is further enhanced by generative AI technology, which promises more efficient and accurate AI-driven data generation for network simulation and optimization. This survey provides insights into generative AI-empowered network digital twins. We begin by outlining the architecture of a network digital twin, which encompasses both digital and physical domains. This architecture involves four key steps: data processing and network monitoring, digital replication and network simulation, designing and training network optimizers, Sim2Real, and network control. Next, we systematically discuss the related studies in each step and make a detailed taxonomy of the problem studied, the methods used, and the key designs leveraged. Each step is examined with a focus on the role of generative AI, from estimating missing data and simulating network behaviors to designing control strategies and bridging the gap between digital and physical domains. Finally, we discuss the open issues and challenges of generative AI-based network digital twins.
- Open Access
- 12 December 2024
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Knowledge Discovery From Data
- Vol. 19 (1) , 1-23
Digital Twin Enhanced Multi-Agent Reinforcement Learning for Large-Scale Mobile Network Coverage Optimization
With the rapid advancement of communication technology and the exponential growth of mobile users, improving network coverage quality and throughput has become increasingly important. In particular, large-scale Base Station (BS) cooperative optimization has become a highly significant topic. BSs can adjust various parameters for high-quality communication, but automating this optimization remains challenging due to environmental sensitivity and interdependencies. Traditional methods for network optimization are constrained by the intricate nature of real-world environments. Further, Reinforcement Learning (RL) techniques, which are effective for configuration policies, encounter difficulties in intricate, high-dimensional wireless communication networks, especially in multi-agent cooperative optimization. To overcome these challenges, this article proposes the Enhanced Multi-Agent Proximal Policy Optimization (EMAPPO), which utilizes the capabilities of the UNet network to extract multi-spatial relationships among a massive number of network elements and employs the DiffPool network to efficiently depict the impact of large-scale action coordination among massive agents on coverage performance. To facilitate evaluation in communication optimization, we further introduce a high-fidelity digital twin-driven mobile network. Extensive experiments validate the effectiveness and superior performance of EMAPPO by utilizing the network digital twin. The results demonstrate significant improvements in signal coverage rate and network throughput compared to the competing methods.
- 12 December 2024
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Knowledge Discovery From Data
- Vol. 19 (1) , 1-23
Open AccessDigital Twin Enhanced Multi-Agent Reinforcement Learning for Large-Scale Mobile Network Coverage Optimization
With the rapid advancement of communication technology and the exponential growth of mobile users, improving network coverage quality and throughput has become increasingly important. In particular, large-scale Base Station (BS) cooperative optimization has become a highly significant topic. BSs can adjust various parameters for high-quality communication, but automating this optimization remains challenging due to environmental sensitivity and interdependencies. Traditional methods for network optimization are constrained by the intricate nature of real-world environments. Further, Reinforcement Learning (RL) techniques, which are effective for configuration policies, encounter difficulties in intricate, high-dimensional wireless communication networks, especially in multi-agent cooperative optimization. To overcome these challenges, this article proposes the Enhanced Multi-Agent Proximal Policy Optimization (EMAPPO), which utilizes the capabilities of the UNet network to extract multi-spatial relationships among a massive number of network elements and employs the DiffPool network to efficiently depict the impact of large-scale action coordination among massive agents on coverage performance. To facilitate evaluation in communication optimization, we further introduce a high-fidelity digital twin-driven mobile network. Extensive experiments validate the effectiveness and superior performance of EMAPPO by utilizing the network digital twin. The results demonstrate significant improvements in signal coverage rate and network throughput compared to the competing methods.
- Open Access
- 24 August 2024
- Published by Association for Computing Machinery (ACM)
- p. 3128-3139
DyPS: Dynamic Parameter Sharing in Multi-Agent Reinforcement Learning for Spatio-Temporal Resource Allocation
In large-scale metropolis, it is critical to efficiently allocate various resources such as electricity, medical care, and transportation to meet the living demands of citizens, according to the spatio-temporal distributions of resources and demands. Previous researchers have done plentiful work on such problems by leveraging Multi-Agent Reinforcement Learning (MARL) methods, where multiple agents cooperatively regulate and allocate the resources to meet the demands. However, facing the great number of agents in large cities, existing MARL methods lack efficient parameter sharing strategies among agents to reduce computational complexity. There remain two primary challenges in efficient parameter sharing: (1) during the RL training process, the behavior of agents changes significantly, limiting the performance of group parameter sharing based on fixed role division decided before training; (2) the behavior of agents forms complicated action trajectories, where their role characteristics are implicit, adding difficulty to dynamically adjusting agent role divisions during the training process. In this paper, we propose Dynamic Parameter Sharing (DyPS) to solve the above challenges. We design self-supervised learning tasks to extract the implicit behavioral characteristics from the action trajectories of agents. Based on the obtained behavioral characteristics, we propose a hierarchical MARL framework capable of dynamically revising the agent role divisions during the training process and thus shares parameters among agents with the same role, reducing computational complexity. In addition, our framework can be combined with various typical MARL algorithms, including IPPO, MAPPO, etc. We conduct 7 experiments in 4 representative resource allocation scenarios, where extensive results demonstrate our method's superior performance, outperforming the state-of-the-art baseline methods by up to 31%. Our source codes are available at https://github.com/tsinghua-fib-lab/DyPS.
- 24 August 2024
- Published by Association for Computing Machinery (ACM)
- p. 3128-3139
Open AccessDyPS: Dynamic Parameter Sharing in Multi-Agent Reinforcement Learning for Spatio-Temporal Resource Allocation
In large-scale metropolis, it is critical to efficiently allocate various resources such as electricity, medical care, and transportation to meet the living demands of citizens, according to the spatio-temporal distributions of resources and demands. Previous researchers have done plentiful work on such problems by leveraging Multi-Agent Reinforcement Learning (MARL) methods, where multiple agents cooperatively regulate and allocate the resources to meet the demands. However, facing the great number of agents in large cities, existing MARL methods lack efficient parameter sharing strategies among agents to reduce computational complexity. There remain two primary challenges in efficient parameter sharing: (1) during the RL training process, the behavior of agents changes significantly, limiting the performance of group parameter sharing based on fixed role division decided before training; (2) the behavior of agents forms complicated action trajectories, where their role characteristics are implicit, adding difficulty to dynamically adjusting agent role divisions during the training process. In this paper, we propose Dynamic Parameter Sharing (DyPS) to solve the above challenges. We design self-supervised learning tasks to extract the implicit behavioral characteristics from the action trajectories of agents. Based on the obtained behavioral characteristics, we propose a hierarchical MARL framework capable of dynamically revising the agent role divisions during the training process and thus shares parameters among agents with the same role, reducing computational complexity. In addition, our framework can be combined with various typical MARL algorithms, including IPPO, MAPPO, etc. We conduct 7 experiments in 4 representative resource allocation scenarios, where extensive results demonstrate our method's superior performance, outperforming the state-of-the-art baseline methods by up to 31%. Our source codes are available at https://github.com/tsinghua-fib-lab/DyPS.