Zhaowei Zhang
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
Zhaowei Zhang has a total of 61 co-authors. Here are the top co-authors by number of shared publications:
- Yaodong Yang
Peking University, Beijing, China
5 shared publications - Hui Liu
School of Software Engineering, Jinling Institute of Technology, Nanjing, China
4 shared publications - Jiakuo Zuo
School of Internet of Things, Nanjing University of Posts and Telecommunications, Nanjing, China
3 shared publications - Su Pan
Nanjing University of Posts and Telecommunications
3 shared publications - Xiaoye Shi
College of Telecommunications and Information Engineering, Nanjing University of Posts and Telecommunications, Nanjing 210003, China
3 shared publications Chun Tian
Optoelectron. Res. Centre, Southampton Univ
2 shared publications
Latest Publications
- 1 October 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (19) , 45537-45561
Integrated UAV-Enabled MEC Network and STAR-RISs for IoT Communications: A Computation Rate Maximization Approach
- 1 October 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (19) , 45537-45561
Integrated UAV-Enabled MEC Network and STAR-RISs for IoT Communications: A Computation Rate Maximization Approach
- 1 April 2026
- Published by Elsevier in Computer Networks
- Vol. 282
Optimizing 3D trajectory and task offloading in collaborative UAV-Enabled mobile edge computing networks
- 1 April 2026
- Published by Elsevier in Computer Networks
- Vol. 282
Optimizing 3D trajectory and task offloading in collaborative UAV-Enabled mobile edge computing networks
- Open Access
- 1 March 2026
- Published by MDPI AG in Mathematics
- Vol. 14 (5) , 841
A Two-Stage Algorithm for Time Series Compression: ARIMA-Based Pre-Compression and Reinforcement Learning Optimized Chunking
The explosive growth of time series gives rise to a large amount of data, which emphasizes the importance of data compression. The data compression not only reduces storage costs but also enhances data transmission efficiency and processing speed. However, traditional compression algorithms usually suffer an insufficient compression ratio and an excessive computational cost. To address these problems above, in this paper, we propose a two-stage compression algorithm for the large-scale time series data. In the first stage, we transform the time series data into low-volatility residual data by using Autoregressive Integrated Moving Average (ARIMA) modeling and apply adaptive precision quantization to improve compressibility. In the second stage, we implement a reinforcement learning-based compression strategy, which utilizes the Q-learning to select the number of blocks to divide the quantized data segment and achieves compression by storing the same content between the divided data blocks only once and storing the different content separately; and we incorporate the Upper Confidence Bound (UCB) to balance exploration and exploitation in order to track changes in data patterns and improve compression performance. Experimental results demonstrate that our algorithm achieves a higher compression ratio while maintaining a low computational complexity compared with traditional compression algorithms.
- 1 March 2026
- Published by MDPI AG in Mathematics
- Vol. 14 (5) , 841
Open AccessA Two-Stage Algorithm for Time Series Compression: ARIMA-Based Pre-Compression and Reinforcement Learning Optimized Chunking
The explosive growth of time series gives rise to a large amount of data, which emphasizes the importance of data compression. The data compression not only reduces storage costs but also enhances data transmission efficiency and processing speed. However, traditional compression algorithms usually suffer an insufficient compression ratio and an excessive computational cost. To address these problems above, in this paper, we propose a two-stage compression algorithm for the large-scale time series data. In the first stage, we transform the time series data into low-volatility residual data by using Autoregressive Integrated Moving Average (ARIMA) modeling and apply adaptive precision quantization to improve compressibility. In the second stage, we implement a reinforcement learning-based compression strategy, which utilizes the Q-learning to select the number of blocks to divide the quantized data segment and achieves compression by storing the same content between the divided data blocks only once and storing the different content separately; and we incorporate the Upper Confidence Bound (UCB) to balance exploration and exploitation in order to track changes in data patterns and improve compression performance. Experimental results demonstrate that our algorithm achieves a higher compression ratio while maintaining a low computational complexity compared with traditional compression algorithms.
- 30 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. 75 (5) , 8669-8674
Sum-Weighted-Comb-Search Based Doppler-Shift Acquisition for Space Communications
In space communications, the signal's long-distance transmission between the flying transmitter-receiver pair suffers both a low signal-to-noise ratio (SNR) and a dynamic Doppler-shift. For the Doppler-shift acquisition, the low SNR requires many symbols accumulated in a long-time period, during which the dynamic Doppler-shift disperses these symbols' total energy over a wide frequency range (i.e., the energy dispersion problem). To address the problem above, we propose a sum-weighted-comb-search (SWCS) scheme to increase the acquisition probability. The SWCS scheme first selects some large elements (called comb-elements) with their quantified weights, then constructs the comb-search-range and sums several weights on every frequency point to strengthen the signal element, and finally searches the largest sum-weighted element to obtain the acquisition result. Moreover, the simulation results have verified our proposed SWCS scheme's significant increase in acquisition probability.
- 30 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. 75 (5) , 8669-8674
Sum-Weighted-Comb-Search Based Doppler-Shift Acquisition for Space Communications
In space communications, the signal's long-distance transmission between the flying transmitter-receiver pair suffers both a low signal-to-noise ratio (SNR) and a dynamic Doppler-shift. For the Doppler-shift acquisition, the low SNR requires many symbols accumulated in a long-time period, during which the dynamic Doppler-shift disperses these symbols' total energy over a wide frequency range (i.e., the energy dispersion problem). To address the problem above, we propose a sum-weighted-comb-search (SWCS) scheme to increase the acquisition probability. The SWCS scheme first selects some large elements (called comb-elements) with their quantified weights, then constructs the comb-search-range and sums several weights on every frequency point to strengthen the signal element, and finally searches the largest sum-weighted element to obtain the acquisition result. Moreover, the simulation results have verified our proposed SWCS scheme's significant increase in acquisition probability.
- Open Access
Make an Offer They Can't Refuse: Grounding Bayesian Persuasion in Real-World Dialogues without Pre-Commitment
Persuasion, a fundamental social capability for humans, remains a challenge for AI systems such as large language models (LLMs). Current studies often overlook the strategic use of information asymmetry in message design or rely on strong assumptions regarding pre-commitment. In this work, we explore the application of Bayesian Persuasion (BP) in natural language within single-turn dialogue settings, to enhance the strategic persuasion capabilities of LLMs. Our framework incorporates a commitment-communication mechanism, where the persuader explicitly outlines an information schema by narrating their potential types (e.g., honest or dishonest), thereby guiding the persuadee in performing the intended Bayesian belief update. We evaluate two variants of our approach: Semi-Formal-Natural-Language (SFNL) BP and Fully-Natural-Language (FNL) BP, benchmarking them against both naive and strong non-BP (NBP) baselines within a comprehensive evaluation framework. This framework covers a diverse set of persuadees -- including LLM instances with varying prompts and fine-tuning and human participants -- across tasks ranging from specially designed persuasion scenarios to general everyday situations. Experimental results on LLM-based agents reveal three main findings: (1) LLMs guided by BP strategies consistently achieve higher persuasion success rates than NBP baselines; (2) SFNL exhibits greater credibility and logical coherence, while FNL shows stronger emotional resonance and robustness in naturalistic conversations; (3) with supervised fine-tuning, smaller models can attain BP performance comparable to that of larger models.
Open AccessMake an Offer They Can't Refuse: Grounding Bayesian Persuasion in Real-World Dialogues without Pre-Commitment
Persuasion, a fundamental social capability for humans, remains a challenge for AI systems such as large language models (LLMs). Current studies often overlook the strategic use of information asymmetry in message design or rely on strong assumptions regarding pre-commitment. In this work, we explore the application of Bayesian Persuasion (BP) in natural language within single-turn dialogue settings, to enhance the strategic persuasion capabilities of LLMs. Our framework incorporates a commitment-communication mechanism, where the persuader explicitly outlines an information schema by narrating their potential types (e.g., honest or dishonest), thereby guiding the persuadee in performing the intended Bayesian belief update. We evaluate two variants of our approach: Semi-Formal-Natural-Language (SFNL) BP and Fully-Natural-Language (FNL) BP, benchmarking them against both naive and strong non-BP (NBP) baselines within a comprehensive evaluation framework. This framework covers a diverse set of persuadees -- including LLM instances with varying prompts and fine-tuning and human participants -- across tasks ranging from specially designed persuasion scenarios to general everyday situations. Experimental results on LLM-based agents reveal three main findings: (1) LLMs guided by BP strategies consistently achieve higher persuasion success rates than NBP baselines; (2) SFNL exhibits greater credibility and logical coherence, while FNL shows stronger emotional resonance and robustness in naturalistic conversations; (3) with supervised fine-tuning, smaller models can attain BP performance comparable to that of larger models.
- 7 August 2025
- Published by Springer Nature
- p. 381-392
Heterogeneous Value Alignment Evaluation for Large Language Models
- 7 August 2025
- Published by Springer Nature
- p. 381-392
Heterogeneous Value Alignment Evaluation for Large Language Models
- 7 August 2025
- Published by Springer Nature
- p. 370-380
Roadmap on Incentive Compatibility for AI Alignment and Governance in Sociotechnical Systems
- 7 August 2025
- Published by Springer Nature
- p. 370-380
Roadmap on Incentive Compatibility for AI Alignment and Governance in Sociotechnical Systems
- 14 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Communications
- Vol. 73 (10) , 9338-9355
Probabilistic-Search and Neighbor-Density Based Doppler-Shift Acquisition in Space Communications
In space communications, the signal’s long-distance transmission between a flying transmitter-receiver pair encounters two obstacles: a huge path-loss and a high-speed movement. The huge path-loss results in a low signal-to-noise ratio (SNR) and the high-speed movement brings a dynamic Doppler-shift, thus posing a great challenge for Doppler-shift acquisition. Under the low SNR, the Doppler-shift acquisition has to accumulate many symbols in a long-time period. However, during this period, the dynamic Doppler-shift disperses all these symbols’ total energy over a wide range, thus causing the energy dispersion problem. To address this problem, we propose a probabilistic-search and neighbor-density (PSND) scheme, where the probabilistic-search considers the element’s amplitude and sacrifices some redundant signal elements for a narrower search-range, while the neighbordensity sums the signal element’s all neighbors for a larger signal energy. The PSND scheme includes two algorithms: the Generic PSND and Iterative PSND. The Generic PSND algorithm first selects some search-elements with their probabilities to construct a probabilistic-search-range, then selects some density-elements with their neighbor-elements to derive the neighbor-densities, and finally searches the largest neighbor-density to obtain the acquisition result. Built upon the Generic PSND algorithm above, the Iterative PSND algorithm iteratively updates the searchelements to further strengthen the neighbor-density in a narrower probabilistic-search-range. Moreover, the simulations results have demonstrated our PSND scheme’s higher acquisition probability, as compared with the existing schemes.
- 14 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Communications
- Vol. 73 (10) , 9338-9355
Probabilistic-Search and Neighbor-Density Based Doppler-Shift Acquisition in Space Communications
In space communications, the signal’s long-distance transmission between a flying transmitter-receiver pair encounters two obstacles: a huge path-loss and a high-speed movement. The huge path-loss results in a low signal-to-noise ratio (SNR) and the high-speed movement brings a dynamic Doppler-shift, thus posing a great challenge for Doppler-shift acquisition. Under the low SNR, the Doppler-shift acquisition has to accumulate many symbols in a long-time period. However, during this period, the dynamic Doppler-shift disperses all these symbols’ total energy over a wide range, thus causing the energy dispersion problem. To address this problem, we propose a probabilistic-search and neighbor-density (PSND) scheme, where the probabilistic-search considers the element’s amplitude and sacrifices some redundant signal elements for a narrower search-range, while the neighbordensity sums the signal element’s all neighbors for a larger signal energy. The PSND scheme includes two algorithms: the Generic PSND and Iterative PSND. The Generic PSND algorithm first selects some search-elements with their probabilities to construct a probabilistic-search-range, then selects some density-elements with their neighbor-elements to derive the neighbor-densities, and finally searches the largest neighbor-density to obtain the acquisition result. Built upon the Generic PSND algorithm above, the Iterative PSND algorithm iteratively updates the searchelements to further strengthen the neighbor-density in a narrower probabilistic-search-range. Moreover, the simulations results have demonstrated our PSND scheme’s higher acquisition probability, as compared with the existing schemes.
- 22 May 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. 73 (10) , 14923-14935
Max-Min-Sliding-Window Based Doppler-Shift Acquisition for Space Communications
In space scenarios, the communications between the flying transmitters and receivers generally face two difficulties: a long-distance transmission and a high-speed movement. The long-distance transmission decreases the signal-to-noise ratio (SNR) at the receiver and the high-speed movement imposes a dynamic Doppler-shift on the transmitter's carrier frequency. Under the low SNR, the Doppler-shift acquisition requires a long-time accumulation of many received symbols. However, during this long-time period, the dynamic Doppler-shift disperses all symbols' total energy over a wide frequency range, thus causing the energy dispersion problem. To address this problem, we propose a max-min-sliding-window (MMSW) scheme by less noise elements and all signal elements. The large elements (also called max-elements) and small elements (also called min-elements) provide us the cues of possible ranges and impossible ranges, respectively. Also, the sliding-window is adopted to collect all signal elements. The MMSW scheme includes two algorithms (i.e., Generic MMSW and Iterative MMSW). The Generic MMSW algorithm first selects the max-elements and min-elements to construct a narrow max-min-range, then slides a window over this range to derive the signal and noise windows, and finally searches the largest window to obtain the acquisition result. By updating the max-elements and min-elements, the Iterative MMSW algorithm iteratively slides a window over a much narrower max-min-range to further strengthen the signal window. Moreover, the simulations results have demonstrated our MMSW scheme's significant advantages over the existing schemes.
- 22 May 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. 73 (10) , 14923-14935
Max-Min-Sliding-Window Based Doppler-Shift Acquisition for Space Communications
In space scenarios, the communications between the flying transmitters and receivers generally face two difficulties: a long-distance transmission and a high-speed movement. The long-distance transmission decreases the signal-to-noise ratio (SNR) at the receiver and the high-speed movement imposes a dynamic Doppler-shift on the transmitter's carrier frequency. Under the low SNR, the Doppler-shift acquisition requires a long-time accumulation of many received symbols. However, during this long-time period, the dynamic Doppler-shift disperses all symbols' total energy over a wide frequency range, thus causing the energy dispersion problem. To address this problem, we propose a max-min-sliding-window (MMSW) scheme by less noise elements and all signal elements. The large elements (also called max-elements) and small elements (also called min-elements) provide us the cues of possible ranges and impossible ranges, respectively. Also, the sliding-window is adopted to collect all signal elements. The MMSW scheme includes two algorithms (i.e., Generic MMSW and Iterative MMSW). The Generic MMSW algorithm first selects the max-elements and min-elements to construct a narrow max-min-range, then slides a window over this range to derive the signal and noise windows, and finally searches the largest window to obtain the acquisition result. By updating the max-elements and min-elements, the Iterative MMSW algorithm iteratively slides a window over a much narrower max-min-range to further strengthen the signal window. Moreover, the simulations results have demonstrated our MMSW scheme's significant advantages over the existing schemes.
- Open Access
- 24 March 2024
- Published by Association for the Advancement of Artificial Intelligence (AAAI) in Proceedings of the AAAI Conference on Artificial Intelligence
- Vol. 38 (16) , 17337-17345
STAS: Spatial-Temporal Return Decomposition for Solving Sparse Rewards Problems in Multi-agent Reinforcement Learning
Centralized Training with Decentralized Execution (CTDE) has been proven to be an effective paradigm in cooperative multi-agent reinforcement learning (MARL). One of the major challenges is credit assignment, which aims to credit agents by their contributions. They lack the functionality to model complicated relations of the delayed global reward in the temporal dimension and suffer from inefficiencies. To tackle this, we introduce Spatial-Temporal Attention with Shapley (STAS), a novel method that learns credit assignment in both temporal and spatial dimensions. It first decomposes the global return back to each time step, then utilizes the Shapley Value to redistribute the individual payoff from the decomposed global reward. To mitigate the computational complexity of the Shapley Value, we introduce an approximation of marginal contribution and utilize Monte Carlo sampling to estimate it. We evaluate our method on an Alice & Bob example and MPE environments across different scenarios. Our results demonstrate that our method effectively assigns spatial-temporal credit, outperforming all state-of-the-art baselines.
- 24 March 2024
- Published by Association for the Advancement of Artificial Intelligence (AAAI) in Proceedings of the AAAI Conference on Artificial Intelligence
- Vol. 38 (16) , 17337-17345
Open AccessSTAS: Spatial-Temporal Return Decomposition for Solving Sparse Rewards Problems in Multi-agent Reinforcement Learning
Centralized Training with Decentralized Execution (CTDE) has been proven to be an effective paradigm in cooperative multi-agent reinforcement learning (MARL). One of the major challenges is credit assignment, which aims to credit agents by their contributions. They lack the functionality to model complicated relations of the delayed global reward in the temporal dimension and suffer from inefficiencies. To tackle this, we introduce Spatial-Temporal Attention with Shapley (STAS), a novel method that learns credit assignment in both temporal and spatial dimensions. It first decomposes the global return back to each time step, then utilizes the Shapley Value to redistribute the individual payoff from the decomposed global reward. To mitigate the computational complexity of the Shapley Value, we introduce an approximation of marginal contribution and utilize Monte Carlo sampling to estimate it. We evaluate our method on an Alice & Bob example and MPE environments across different scenarios. Our results demonstrate that our method effectively assigns spatial-temporal credit, outperforming all state-of-the-art baselines.