Zhenyu Li
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Zhenyu Li has a total of 25 co-authors. Here are the top co-authors by number of shared publications:
- Pengjie Xu
Qingdao University
5 shared publications - Tianyi Shang
Department of Maynooth International Engineering College, Fuzhou University, Fuzhou, China
5 shared publications - Zhaojun Deng
College of Surveying and Geo-Informatics, Tongji University, Shanghai, China
5 shared publications Pengjie Xu
School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai, China
4 shared publicationsTianyi Shang
School of Mechanical Engineering, Qilu University of Technology (Shandong Academy of Sciences), Jinan, China
3 shared publications- Pengjie Xu
Shanghai Jiao Tong University
2 shared publications
Latest Publications
- Open Access
A Comprehensive Review of Robotic Grinding Technology
Integrated die-cast components reduce machining/assembly steps and improve mechanical dynamic characteristics, eliminating joint loosening/fracture risks after long-term use. However, the highly variable geometries and random spatial distributions of burrs, flash, parting lines, and risers in castings invalidate pre-programmed or teach-in robotic grinding methods. This paper reviews recent progress and future trends in robotic grinding, analyzing four core aspects: force control stability/adaptability (e.g., adaptive impedance control can reduce average force-tracking error to 0.38 N), trajectory planning/path generation (e.g., error-driven compensation can lower contour error by 34.2–55.1%), process parameter optimization, and challenges of sensing latency/quality evaluation (e.g., deep learning models achieve 97.64% accuracy in identifying abrasive belt wear states). The key enabling technologies are summarized, including active/passive compliant force control, model-/data-driven adaptive trajectory planning, intelligent process parameter optimization integrating physical mechanisms and data-driven approaches, and multi-modal state monitoring with online quality assessment. Representative applications (metal castings, aero-engine blades, thin-walled components, weld seams) are presented, and prospective research directions are proposed. This paper provides a comprehensive reference for theoretical research and engineering practice in this field.
Open AccessA Comprehensive Review of Robotic Grinding Technology
Integrated die-cast components reduce machining/assembly steps and improve mechanical dynamic characteristics, eliminating joint loosening/fracture risks after long-term use. However, the highly variable geometries and random spatial distributions of burrs, flash, parting lines, and risers in castings invalidate pre-programmed or teach-in robotic grinding methods. This paper reviews recent progress and future trends in robotic grinding, analyzing four core aspects: force control stability/adaptability (e.g., adaptive impedance control can reduce average force-tracking error to 0.38 N), trajectory planning/path generation (e.g., error-driven compensation can lower contour error by 34.2–55.1%), process parameter optimization, and challenges of sensing latency/quality evaluation (e.g., deep learning models achieve 97.64% accuracy in identifying abrasive belt wear states). The key enabling technologies are summarized, including active/passive compliant force control, model-/data-driven adaptive trajectory planning, intelligent process parameter optimization integrating physical mechanisms and data-driven approaches, and multi-modal state monitoring with online quality assessment. Representative applications (metal castings, aero-engine blades, thin-walled components, weld seams) are presented, and prospective research directions are proposed. This paper provides a comprehensive reference for theoretical research and engineering practice in this field.
- 30 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Informatics
- Vol. 22 (7) , 5975-5986
Seeing Through the Rain: Multistage Attention With Depth-Guided Restoration for Robust Localization
Rainy conditions significantly degrade visual perception for autonomous vehicles, thereby degrading the performance of vision-based localization and navigation systems. To enhance robust vehicle place recognition in rainy conditions, this article proposes a rain-removal-enhanced visual place recognition network, connected end-to-end with a multiscale pyramid transformer. To enhance feature representation on rainy days for high-level tasks, a novel multiscale sampling network is proposed to achieve lightweight operation by utilizing efficient residual connections. To further strengthen the visual representation and achieve discriminative features, a multiscale feature pyramid network is designed to aggregate multistage visual attention. Furthermore, to provide a comprehensive evaluation, we develop two challenging benchmarks for rainy day place recognition: RainPlace and RainSim, the former is built on real scenes, while RainSim is based on virtual scenes. Through extensive experiments on benchmarks, we demonstrate that the proposed enhanced net achieves superior place recognition performance in rainy scenes for mobile vehicle applications and outperforms existing SOTA methods that rely on discriminative attention features or various aggregation algorithms. Our.
- 30 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Informatics
- Vol. 22 (7) , 5975-5986
Seeing Through the Rain: Multistage Attention With Depth-Guided Restoration for Robust Localization
Rainy conditions significantly degrade visual perception for autonomous vehicles, thereby degrading the performance of vision-based localization and navigation systems. To enhance robust vehicle place recognition in rainy conditions, this article proposes a rain-removal-enhanced visual place recognition network, connected end-to-end with a multiscale pyramid transformer. To enhance feature representation on rainy days for high-level tasks, a novel multiscale sampling network is proposed to achieve lightweight operation by utilizing efficient residual connections. To further strengthen the visual representation and achieve discriminative features, a multiscale feature pyramid network is designed to aggregate multistage visual attention. Furthermore, to provide a comprehensive evaluation, we develop two challenging benchmarks for rainy day place recognition: RainPlace and RainSim, the former is built on real scenes, while RainSim is based on virtual scenes. Through extensive experiments on benchmarks, we demonstrate that the proposed enhanced net achieves superior place recognition performance in rainy scenes for mobile vehicle applications and outperforms existing SOTA methods that rely on discriminative attention features or various aggregation algorithms. Our.
- 26 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Transportation Systems
- Vol. PP (15249050) , 1-13
Vehicle-SceneInteraction: A Text-Driven 3-D Lidar Place Recognition Method for Autonomous Driving
Environment description-based vehicle-scene interaction for vehicle localization in large-scale point cloud maps constructed through multi-sensor systems is critically significant for the advancement of large-scale intelligent transportation systems, such as delivery vehicles operating in the ‘last mile’. However, current vehicle-scene interaction encounter challenges due to the inability of point cloud encoders to effectively capture local details and long-range spatial relationships and a significant modality gap between text and point cloud representations. To address these challenges, we present Des4Pos, a novel two-stage text-driven 3D Lidar place recognition framework. In the coarse stage, the point-cloud encoder utilizes the Multi-scale Fusion Attention Mechanism (MFAM) to enhance local geometric features, followed by a bidirectional Long Short-Term Memory (LSTM) module to strengthen global spatial relationships. Concurrently, the Stepped Text Encoder (STE) integrates cross-modal prior knowledge from CLIP and aligns text and point-cloud features using this prior knowledge, effectively bridging modality discrepancies. In the fine stage, we introduce a Cascaded Residual Attention (CRA) module to fuse cross-modal features and predict relative localization offsets, thereby achieving greater localization precision. Experiments on the KITTI360Pose test set demonstrate that Des4Pos achieves state-of-the-art performance in text-to-point-cloud place recognition at the expense of a slight increase in the runtime. Specifically, it attains a top-1 accuracy of 40% and a top-10 accuracy of 77% under a 5-meter radius threshold, surpassing the best previous sota method Text2Loc by 8% and 7%, respectively. Our code and datasets are publicly available at https://github.com/nuozimiaowu/Des4Pos
- 26 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Transportation Systems
- Vol. PP (15249050) , 1-13
Vehicle-SceneInteraction: A Text-Driven 3-D Lidar Place Recognition Method for Autonomous Driving
Environment description-based vehicle-scene interaction for vehicle localization in large-scale point cloud maps constructed through multi-sensor systems is critically significant for the advancement of large-scale intelligent transportation systems, such as delivery vehicles operating in the ‘last mile’. However, current vehicle-scene interaction encounter challenges due to the inability of point cloud encoders to effectively capture local details and long-range spatial relationships and a significant modality gap between text and point cloud representations. To address these challenges, we present Des4Pos, a novel two-stage text-driven 3D Lidar place recognition framework. In the coarse stage, the point-cloud encoder utilizes the Multi-scale Fusion Attention Mechanism (MFAM) to enhance local geometric features, followed by a bidirectional Long Short-Term Memory (LSTM) module to strengthen global spatial relationships. Concurrently, the Stepped Text Encoder (STE) integrates cross-modal prior knowledge from CLIP and aligns text and point-cloud features using this prior knowledge, effectively bridging modality discrepancies. In the fine stage, we introduce a Cascaded Residual Attention (CRA) module to fuse cross-modal features and predict relative localization offsets, thereby achieving greater localization precision. Experiments on the KITTI360Pose test set demonstrate that Des4Pos achieves state-of-the-art performance in text-to-point-cloud place recognition at the expense of a slight increase in the runtime. Specifically, it attains a top-1 accuracy of 40% and a top-10 accuracy of 77% under a 5-meter radius threshold, surpassing the best previous sota method Text2Loc by 8% and 7%, respectively. Our code and datasets are publicly available at https://github.com/nuozimiaowu/Des4Pos
- 6 February 2026
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 37 (5) , 055004
Orthogonal-Procrustes-generalized compact pose estimation using uncalibrated pseudo-camera clusters
Pose estimation capable of precise positioning and orientation description is a critical task in robotics, driverless, and aerospace. In this paper, an orthogonal-Procrustes (OP)-generalized compact pose estimation method is proposed using uncalibrated pseudo camera clusters. Based on the coplanarity constraints, the camera intrinsic parameters are estimated through local variable substitution. Taking into full account the anisotropic scaling, pseudo-camera pose estimation is proposed using generalized OP. Simulation analyses were conducted to investigate the influence of configuration parameters on measurement accuracy and stability. Real experiments on pose estimation were carried out to compare the commonly used algorithms and verify the feasibility. By multi-viewpoint prior information fusion, this method outperforms the typical calibrated and uncalibrated methods in terms of accuracy and stability, particularly under challenging conditions with high noise, a few points, and coplanar points. Specifically, the mean errors for rotation, translation, and focal length are 0.4°, 0.8%, and 13.6%. Moreover, it allows for application in space-constrained scenarios and zoomable visual systems without multi-station calibration.
- 6 February 2026
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 37 (5) , 055004
Orthogonal-Procrustes-generalized compact pose estimation using uncalibrated pseudo-camera clusters
Pose estimation capable of precise positioning and orientation description is a critical task in robotics, driverless, and aerospace. In this paper, an orthogonal-Procrustes (OP)-generalized compact pose estimation method is proposed using uncalibrated pseudo camera clusters. Based on the coplanarity constraints, the camera intrinsic parameters are estimated through local variable substitution. Taking into full account the anisotropic scaling, pseudo-camera pose estimation is proposed using generalized OP. Simulation analyses were conducted to investigate the influence of configuration parameters on measurement accuracy and stability. Real experiments on pose estimation were carried out to compare the commonly used algorithms and verify the feasibility. By multi-viewpoint prior information fusion, this method outperforms the typical calibrated and uncalibrated methods in terms of accuracy and stability, particularly under challenging conditions with high noise, a few points, and coplanar points. Specifically, the mean errors for rotation, translation, and focal length are 0.4°, 0.8%, and 13.6%. Moreover, it allows for application in space-constrained scenarios and zoomable visual systems without multi-station calibration.
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Automation Science and Engineering
- Vol. 23, 13220-13232
Hybrid State Space Modeling for Sequence-Based Robot Localization Under Challenging Environments
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Automation Science and Engineering
- Vol. 23, 13220-13232
Hybrid State Space Modeling for Sequence-Based Robot Localization Under Challenging Environments
- 14 December 2025
- Published by Elsevier in Expert Systems with Applications
- Vol. 304
Quadruplet-attention transformer for scale-invariant robot place recognition
- 14 December 2025
- Published by Elsevier in Expert Systems with Applications
- Vol. 304
Quadruplet-attention transformer for scale-invariant robot place recognition
- 19 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 11 (1) , 642-649
FourierPlace: A Vision-Language Localization Framework Based on Frequency Domain Representations
Language-guided localization within 3D environments continues to pose a significant challenge for autonomous systems, primarily due to the need for precise alignment between sparse point cloud data and inherently ambiguous natural language descriptions. To address this, we present a novel vision-language localization framework, namely FourierPlace, that leverages frequency-domain representations to enhance the alignment of complex geometric features with imprecise linguistic cues. At the core of our approach is the Frequency Fusion Enhancement (FFE) module, which converts raw point cloud data into frequency-domain signals. Complementing this, the Fourier Gate Attention (FGA) mechanism operates on these frequency-domain features to strengthen cross-modal correspondence. Furthermore, we introduce the Hierarchical Language Understanding Network (HiLUNet), which progressively refines linguistic features through a multi-stage architecture. Additionally, the Multiscale Cascade Cross-Attention (MCCA) module incorporates geometric information at multiple scales in the fine stage. Experiments conducted on the KITTI360Pose benchmark demonstrate that FourierPlace achieves state-of-the-art performance, outperforming existing methods with 3.7% improvement in Top-1 coarse retrieval accuracy and 4.0% increase in Top-1 localization accuracy within a 15-meter threshold on the test set. Our proposed framework presents a robust and scalable solution for language-guided localization in large-scale autonomous applications, including delivery robotics and augmented reality (AR) navigation systems.
- 19 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 11 (1) , 642-649
FourierPlace: A Vision-Language Localization Framework Based on Frequency Domain Representations
Language-guided localization within 3D environments continues to pose a significant challenge for autonomous systems, primarily due to the need for precise alignment between sparse point cloud data and inherently ambiguous natural language descriptions. To address this, we present a novel vision-language localization framework, namely FourierPlace, that leverages frequency-domain representations to enhance the alignment of complex geometric features with imprecise linguistic cues. At the core of our approach is the Frequency Fusion Enhancement (FFE) module, which converts raw point cloud data into frequency-domain signals. Complementing this, the Fourier Gate Attention (FGA) mechanism operates on these frequency-domain features to strengthen cross-modal correspondence. Furthermore, we introduce the Hierarchical Language Understanding Network (HiLUNet), which progressively refines linguistic features through a multi-stage architecture. Additionally, the Multiscale Cascade Cross-Attention (MCCA) module incorporates geometric information at multiple scales in the fine stage. Experiments conducted on the KITTI360Pose benchmark demonstrate that FourierPlace achieves state-of-the-art performance, outperforming existing methods with 3.7% improvement in Top-1 coarse retrieval accuracy and 4.0% increase in Top-1 localization accuracy within a 15-meter threshold on the test set. Our proposed framework presents a robust and scalable solution for language-guided localization in large-scale autonomous applications, including delivery robotics and augmented reality (AR) navigation systems.
- Top Cited PapersOpen Access
- 30 August 2025
- Published by Springer Nature in Artificial Intelligence Review
- Vol. 58 (11) , 1-48
Place recognition meet multiple modalities: a comprehensive review, current challenges and future development
- 30 August 2025
- Published by Springer Nature in Artificial Intelligence Review
- Vol. 58 (11) , 1-48
Top Cited PapersOpen AccessPlace recognition meet multiple modalities: a comprehensive review, current challenges and future development
- Open Access
- 17 June 2025
- Published by Springer Nature in Research Square
Place Recognition Meet Multiple Modalities: A Comprehensive Review, Current Challenges and Future Development
Place recognition is a cornerstone of vehicle navigation and mapping, which is pivotal in enabling systems to determine whether a location has been previously visited. This capability is critical for tasks such as loop closure in Simultaneous Localization and Mapping (SLAM) and long-term navigation under varying environmental conditions. This survey comprehensively reviews recent advancements in place recognition, emphasizing three representative methodological paradigms: Convolutional Neural Network (CNN)-based approaches, Transformer-based frameworks, and cross-modal strategies. We begin by elucidating the significance of place recognition within the broader context of autonomous systems. Subsequently, we trace the evolution of CNN-based methods, highlighting their contributions to robust visual descriptor learning and scalability in large-scale environments. We then examine the emerging class of Transformer-based models, which leverage self-attention mechanisms to capture global dependencies and offer improved generalization across diverse scenes. Furthermore, we discuss cross-modal approaches that integrate heterogeneous data sources such as Lidar, vision, and text description, thereby enhancing resilience to viewpoint, illumination, and seasonal variations. We also summarize standard datasets and evaluation metrics widely adopted in the literature. Finally, we identify current research challenges and outline prospective directions, including domain adaptation, real-time performance, and lifelong learning, to inspire future advancements in this domain. The unified framework of leading-edge place recognition methods, i.e., code library, and the results of their experimental evaluations are available at https://github.com/CV4RA/SOTA-Place-Recognitioner.
- 17 June 2025
- Published by Springer Nature in Research Square
Open AccessPlace Recognition Meet Multiple Modalities: A Comprehensive Review, Current Challenges and Future Development
Place recognition is a cornerstone of vehicle navigation and mapping, which is pivotal in enabling systems to determine whether a location has been previously visited. This capability is critical for tasks such as loop closure in Simultaneous Localization and Mapping (SLAM) and long-term navigation under varying environmental conditions. This survey comprehensively reviews recent advancements in place recognition, emphasizing three representative methodological paradigms: Convolutional Neural Network (CNN)-based approaches, Transformer-based frameworks, and cross-modal strategies. We begin by elucidating the significance of place recognition within the broader context of autonomous systems. Subsequently, we trace the evolution of CNN-based methods, highlighting their contributions to robust visual descriptor learning and scalability in large-scale environments. We then examine the emerging class of Transformer-based models, which leverage self-attention mechanisms to capture global dependencies and offer improved generalization across diverse scenes. Furthermore, we discuss cross-modal approaches that integrate heterogeneous data sources such as Lidar, vision, and text description, thereby enhancing resilience to viewpoint, illumination, and seasonal variations. We also summarize standard datasets and evaluation metrics widely adopted in the literature. Finally, we identify current research challenges and outline prospective directions, including domain adaptation, real-time performance, and lifelong learning, to inspire future advancements in this domain. The unified framework of leading-edge place recognition methods, i.e., code library, and the results of their experimental evaluations are available at https://github.com/CV4RA/SOTA-Place-Recognitioner.