About

Zhi-Li Zhang is a Professor, Distinguished McKnight University Professor, and the Qwest Land Grant Chair in Telecommunications at the University of Minnesota's Department of Computer Science and Engineering. His research interests broadly encompass networking, edge and cloud computing, cyber-physical systems, and the application of artificial intelligence and machine learning (AI/ML) support for networked systems, as well as networked systems support for AI/ML. His past research focused on the analysis, design, and development of scalable Internet quality-of-service (QoS), content distribution networks, Internet measurement, resilient routing systems, and network security. His current research primarily aims at building service-oriented, application-aware, highly scalable, resilient, and secure 5G and next-generation networked systems. He is also dedicated to developing and integrating novel AI/ML algorithms to enable intelligent software-defined networking infrastructures, edge and cloud systems, and emerging applications such as collaborative autonomous driving, Digital Twins, the metaverse, and the Internet of Things (IoT). Zhang has made significant contributions to the field through his leadership in research projects, including those funded by the NSF, and has been recognized with numerous awards, including the McKnight Professorship, IEEE Fellow, and the AWS Machine Learning Research Award.

Research topics

• Computer Science
• Operating system
• Computer network
• Distributed computing
• Embedded system
• Parallel computing
• Telecommunications

Selected publications

• Formation mechanism of transfer films between PEEK-based materials and silicon nitride ceramic balls under varying loads

  Industrial Lubrication and Tribology · 2025-12-04

  article

  Purpose This study aims to further investigate the tribological behavior of polyether ether ketone (PEEK) and its composites against silicon nitride (Si3N4), with a focus on analyzing the transfer film formation mechanism and the lubricating roles of various fillers in the PEEK composites during the film-forming process. Design/methodology/approach The friction coefficients of three PEEK composite discs sliding against Si3N4 balls under dry friction at various loads were investigated using a tribometer. The wear profiles of the three discs were measured using a Taylor surface profilometer, and the specific wear rates were calculated. The surface morphology, infrared spectra and Raman spectra of the transfer films on the Si3N4 balls were characterized using scanning electron microscopy (SEM), Fourier transform infrared (FTIR) microscopy and confocal Raman microscopy, respectively. Finally, the wear resistance of the transfer films was evaluated through additional durability tests. Findings Among the tested composites, PEEK matrix filled with 10 Vol.% graphite, 10 Vol.% carbon fiber (CF) and 10 Vol.% polytetrafluoroethylene (PTFE) (PVX) exhibited the lowest friction coefficient against Si3N4, whereas graphite-filled PEEK (CF30) showed the lowest specific wear rate. The formation of locally graphitized carbon (sp2) films within the polymer transfer layer plays a crucial role in solid lubrication during sliding. The transfer film of neat PEEK mainly consists of organic debris or degradation products of the matrix. In contrast, for CF30 and PVX, the incorporated CFs can generate sp2 carbon films, while in PVX, the presence of PTFE further provides synergistic lubrication with the PEEK matrix. Originality/value The findings provide a theoretical foundation for the improved application of PEEK composites in full ceramic ball bearing cages.

  PublisherDOI

• A Peek into 5G NSA vs. SA Control Plane Performance

  2025-02-12

  articleSenior author

  The Stand-Alone (SA) 5G deployment mode promises many benefits over the Non-Stand-Alone (NSA) 5G mode, such as improved throughput, lower latency, and more architectural flexibility, to better support future emerging applications such as AR/VR, IoT, and teleoperated driving. These promised improvements also extend to the control plane operations of 5G-SA, such as attachment/registration procedures to mobile networks, mobility, and security management to provide better user quality-of-experience (QoE). Most of the existing work explores the data plane and end-to-end performance of 5G. In this paper, we investigate and quantify the performance differences in the control plane of 5G-SA compared to 5G-NSA. Our results indicate that 5G-SA mostly has a worse (i.e., slower) control performance (by 16.6% for attachment/registration, PDU session establishment and 64.3% RRC procedure), unlike expectations, raising questions about current (virtualized) 5G-SA deployment and core network functionality placement.

  PublisherDOI

• Rule-adaptive lane-changing trajectory planning method for autonomous vehicles driven by dynamic risk information

  Scientific Reports · 2025-08-26 · 3 citations

  articleOpen accessSenior author

  Rapid advancements in autonomous driving technology have highlighted the challenges of ensuring vehicle safety and driving efficiency in complex dynamic traffic environments. Current approaches typically define potential risks as safety constraints for compliance and use them in trajectory planning. However, the risks predefined in these constraints are often fixed, reducing driving efficiency. To address this limitation, we proposed a dynamic risk-information-driven adaptive trajectory planning method for autonomous vehicles (AVs). This study dynamically adjusted safety constraints using risk assessment results to improve driving efficiency without compromising safety. Firstly, considering the influence of vehicle suspension characteristics on driving safety, collision, and instability risk assessment indices were designed using a three-way-coupled dynamic model to assess driving safety risks. Next, we used the safety risk assessment module to evaluate specific potential risks and adaptively adjusted the safety constraints for constraint-based adaptive trajectory planning. Furthermore, considering trajectory traversal constraints, trajectory selection and optimization were performed on pre-planned trajectories using the cost function to determine the optimal driving trajectory. Lane-changing trajectory planning experiments showed that the method adaptively adjusts safety constraints based on risk assessment results. Under the premise of ensuring driving safety, driving efficiency improved by 55.9% in the preset instability constraint scenario and 27.86% in the preset collision constraint scenario.

  PublisherOA PDFDOI

• High-Voltage-Resistant Highly Stable Solid Polymer Electrolyte via In Situ Integrated Construction with Fast Ion Migration

  ACS Applied Materials & Interfaces · 2025-01-22 · 5 citations

  article

  Electric aircraft such as electric aircraft and electric vehicles play a key role in the future electric aviation industry, but they put forward huge requirements for battery energy density. However, the current high-energy-density lithium battery technology still needs to be broken through. Herein, through the molecular structure design of the polymer electrolyte, a strategy of a fast migration channel and wide electrochemical window is proposed to fabricate high-voltage-resistant solid polymer electrolyte (HVPE) via in situ polymerization. Thus, HVPE exhibits an ultrahigh Li+ transfer number (tLi+) of 0.92 and an excellent electrochemical window of 5.1 V to match with a high-voltage lithium cobalt oxide (LCO) cathode. This fast conduction of Li+ allows for stable and uniform lithium plating and stripping deposition for more than 1000 h, which also reveals a well-defined dual interfacial stabilization mechanism. These results endow the assembled LCO|HVPE|Li cell cycles steadily for 500 cycles at 4.5 V and 0.5C with a superior capacity retention of 89.93%. Moreover, the assembled LCO|HVPE|Li pouch cell exhibits a capacity retention rate of up to 94.01% after 50 cycles. More importantly, our proposed HVPE provides new insights into structural design and fabrication strategies for high-energy-density solid-state polymer batteries.

  PublisherDOI

• A novel trend threshold range control strategy for enhancing energy efficiency and water temperature control performance in industrial-scale greenhouse aquaculture

  Energy · 2025-10-11

  article
  PublisherDOI

• Research on temperature compensation technology for fiber optic gyroscopes based on FISCSO-BPNN model and temperature field modeling

  2025-06-11

  article

  Fiber optic gyroscopes (FOG) have been widely used in aerospace, aviation, navigation, and other fields due to their excellent performance, such as small size, light weight, fast response, and high precision. However, changes in environmental temperature significantly affect the output accuracy of fiber optic gyroscopes, limiting their application precision in complex environments. To effectively address this issue ， a fusion temperature-based ISCSO-BPNN temperature compensation model (FISCSO-BPNN) was constructed by integrating the key temperature characteristic points, experimental set temperatures, and measured temperatures as inputs, with the gyroscope's bias data under temperature fluctuations as the output. Experimental results showed that the bias stability of the FISCSO-BPNN model improved by approximately 68.6% compared to the non-compensated scenario, and it outperformed the ISCSO-BPNN model in terms of compensation effectiveness.

  PublisherDOI

• OPCM: Opportunistic Performance-driven Connectivity Management for 5G/xG Networks

  Proceedings of the ACM on Networking · 2025-11-24

  articleOpen access

  5G and future 6G networks deploy cells with diverse combinations of access technologies, architectures, and radio frequency bands/channels. Cellular operators also employ carrier aggregation for higher data access speeds. We investigate the fundamental question of how to intelligently and dynamically configure and reconfigure a user equipment's serving cells to deliver the best network performance. Through comprehensive measurements across 12 cities in 5 countries, we experimentally show the wide availability, heterogeneity, and untapped performance gains of today's cell deployments. We then present a principled, performance-driven connectivity management framework, dubbed OPCM. It is a centralized solution deployed at the base station, allowing it to coordinate multiple UEs, enforce operator policies, and facilitate user fairness. Extensive evaluations show that OPCM improves the application QoE by up to 65.2%.

  PublisherDOI

• A Modified pBRDF Model Considering the Influence of Relative Azimuth Angle for Coating Targets

  IEEE photonics journal · 2025-10-14

  articleOpen access

  The key to the researches on the polarization characteristics of targets lies in the precise establishment of the polarization model—the polarized bidirectional reflectance distribution function (pBRDF). Currently, the researches on the pBRDF model for coating targets have the problem of inaccurately describing hemispherical distribution of the degree of linear polarization (DoLP), especially the insufficiency in describing the influence of the azimuth angle on DoLP. In this thesis, firstly, with the aim of minimizing the error in linear polarization degree, through simulation comparative experiments on common microfacet distribution function (NDF), geometric attenuation factor (GAF) and multiple reflection function, it is determined that the Gaussian NDF, the modified integral GAF and the Minnaert model considering roughness are more suitable for coating targets. Secondly, the combined model for coating targets was established by combining the three functions, and it is found that the combined model was insensitive to the azimuth angle. Finally, a pBRDF model incorporating a high-order polynomial of relative azimuth angle is proposed, which improved the problem of excessive error caused by the insensitivity to the azimuth angle. The results of experiments show that the relative errors when adopting the modified model have decreased by 45.8%, 66.7%, 10.5%, and 32.1% respectively. The determination coefficient has reached 0.948, 0.953, 0.917 and 0.930, and the performance indicators are superior to those of the existing models. The research results provide a reference for describing the hemispherical spatial distribution of DoLP for coating targets.

  PublisherDOI

• A Comprehensive Real-World Evaluation of 5G Improvements Over 4G in Low- and Mid-Bands

  IEEE Transactions on Cognitive Communications and Networking · 2025-04-04 · 6 citations

  article

  With the proliferation of 5G and developments of 6G technologies already underway, understanding the realworld performance of various 5G enhancements such as higher modulation, beamforming, and MIMO of deployed 5G over 4G is vital. This work addresses this knowledge gap by conducting extensive 4G/5G measurements in low-(<1 GHz) and mid-bands (1 to 6 GHz) across Chicago and Minneapolis. As both 4G and 5G utilize low-and mid-band channels, we carefully analyze their performance and signal parameters to reveal several key observations: (i) 5G’s throughput improvement today is mainly driven by wider channel bandwidth in the mid-bands, from both single channels and channel aggregation, (ii) realizing further throughput gains necessitates better signal conditions achievable through denser deployment and/or beamforming, (iii) channel rank analysis shows real-world channel conditions rarely support the full 4x4 MIMO, (iv) advanced features like MU-MIMO and higher order modulation like 1024-QAM have yet to be widely deployed, and (v) aggregated throughput performance in LTE can be enhanced by incorporating shared and unlicensed bands, resulting in a performance similar to single-channel NR. These observations and conclusions suggest that the next generation of cellular systems should prioritize wider channels, possibly with enhanced channel aggregation, and a denser deployment architecture utilizing more beams. This would ensure consistently better signal strength across the coverage area with up to 4 MIMO layers per user.

  PublisherDOI

• Understanding 5G Performance for Real-World Services: A Content Provider’s Perspective

  IEEE Transactions on Networking · 2025-03-11

  article

  Recent years have witnessed a rapid growth of both 5G coverage and 5G users, attracting several measurement studies on its coverage, reliability and quality of service. However, the capabilities and potential impacts of 5G, especially Standalone (SA) 5G, still remain to be fully understood from a content provider (CP)’s perspective. This paper fills this gap by studying 5G networks used by over 23 million users in one year in Kuaishou, a popular crowdsourced live streaming platform. With passive and active measurements, we have the following key findings: i) SA 5G generally provides end-to-end performance improvements compared to 4G or Non-Standalone (NSA) 5G, but its advantage depends on both the number of cellular users and CP-level configurations. ii) In the radio access networks, SA 5G is more sensitive to access density, but has better handover tolerance. iii) Controlled experiments with 29 mobile device models on energy consumption refute some “conventional wisdom,” including the notion that 5G always consumes more power. iv) Traceroute-based active experiments in over 300 cities show that although users are “closer” to the internet in SA 5G due to the control and user plane separation, whether end-to-end latency benefits from that partly depends on the routing policy at the gateways. Furthermore, we propose a 5G-aware rebuffer strategy tested by 9 million viewers in Kuaishou, showing a 7% reduction in rebuffer proportion. Finally, we also provide new design space for other 5G participants.

  PublisherDOI

Recent grants

• Collaborative Research: CISE: Large: Integrated Networking, Edge System and AI Support for Resilient and Safety-Critical Tele-Operations of Autonomous Vehicles

  NSF · $2.3M · 2023–2027

• NeTS: Small: Diverse and Resilient Beyond Paths

  NSF · $300k · 2016–2021

• CRI: IAD Research Infrastructure for Emerging Networked Systems and Applications

  NSF · $200k · 2007–2009

• NeTS: Small: VIRO: Highly Scalable, Robust and Namespace Independent Routing for Future Networks

  NSF · $436k · 2010–2015

• NeTS: Small: Exerting Logically Centralized Control over Legacy Switches via Incremental SDN Deployment

  NSF · $350k · 2016–2021

Frequent coauthors

• Eman Ramadan

  Twin Cities Orthopedics

  62 shared

• Wei Ye

  49 shared

• Feng Qian

  University of Southern California

  43 shared

• Rostand A. K. Fezeu

  University of Minnesota

  41 shared

• Arvind Narayanan

  AT&T (United States)

  36 shared

• Sourabh Jain

  31 shared

• Yanhua Li

  Hunan Institute of Technology

  31 shared

• Udhaya Kumar Dayalan

  University of Minnesota

  29 shared

Labs

• Zhi-Li Zhang LabPI

Awards & honors

• Qwest Land Grant Chair in Telecommunications (2007 - present…
• Amazon Web Services (AWS) Machine Learning Research Award (2…
• McKnight Professorships (2013)
• IEEE Fellow (2011)
• IBM Faculty Award (2007)