Lele Wang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Lele Wang has a total of 34 co-authors. Here are the top co-authors by number of shared publications:
- Qirong Xiao
Tsinghua University
9 shared publications Dan Li
Department of Precision Instrument, Tsinghua University, Beijing 100084, China
7 shared publications- Yuan Meng
Department of Mechanical Engineering and Materials Science, Washington University in St. Louis, St. Louis, USA
6 shared publications Ping Yan
Department of Precision Instrument, Tsinghua University, Beijing 100084, China
5 shared publicationsMali Gong
State Key Laboratory of Precision Measurement Technology and Instruments, Department of Precision Instrument, Tsinghua University 1 , Beijing 100084, China
4 shared publicationsHailong Zhang
Department of Precision Instrument, Tsinghua University, Beijing, China
3 shared publications
Latest Publications
- Open Access
- 12 June 2026
- Published by American Association for the Advancement of Science (AAAS) in Science Advances
- Vol. 12 (24) , eaef6263
Deep learning–enabled versatile shape perception for soft robots via single-ended multimode fiber
The evolution of soft robots into embodied intelligent systems relies fundamentally on precise proprioception. However, a universal solution for capturing continuous deformations during diverse interactions, particularly in spatially confined interventional scenarios, remains lacking. Here, we introduce a deep learning–enabled versatile shape perception method based on a single-ended multimode fiber (MMF). By leveraging the intrinsic integration advantages of optics, our minimalist reflective architecture physically eliminates the dependence on complex demodulation units and distal devices. Furthermore, treating chaotic optical speckle fields as data streams encoding high-dimensional shape information, reconfigurable neural decoders resolve a single physical channel into versatile perception modes tailored to heterogeneous tasks: discrete state confirmation on soft grippers (>99% accuracy), continuous shape tracking on bionic dexterous hands (~5-fold spatial resolution enhancement), and intuitive 3D morphological reconstruction of soft surgical robots (IoU>0.93). Overall, our work establishes a versatile framework for breaking hardware adaptability limits via computation, laying a solid foundation for closed-loop control in digital twins of soft robots.
- 12 June 2026
- Published by American Association for the Advancement of Science (AAAS) in Science Advances
- Vol. 12 (24) , eaef6263
Open AccessDeep learning–enabled versatile shape perception for soft robots via single-ended multimode fiber
The evolution of soft robots into embodied intelligent systems relies fundamentally on precise proprioception. However, a universal solution for capturing continuous deformations during diverse interactions, particularly in spatially confined interventional scenarios, remains lacking. Here, we introduce a deep learning–enabled versatile shape perception method based on a single-ended multimode fiber (MMF). By leveraging the intrinsic integration advantages of optics, our minimalist reflective architecture physically eliminates the dependence on complex demodulation units and distal devices. Furthermore, treating chaotic optical speckle fields as data streams encoding high-dimensional shape information, reconfigurable neural decoders resolve a single physical channel into versatile perception modes tailored to heterogeneous tasks: discrete state confirmation on soft grippers (>99% accuracy), continuous shape tracking on bionic dexterous hands (~5-fold spatial resolution enhancement), and intuitive 3D morphological reconstruction of soft surgical robots (IoU>0.93). Overall, our work establishes a versatile framework for breaking hardware adaptability limits via computation, laying a solid foundation for closed-loop control in digital twins of soft robots.
- Open Access
- 5 January 2026
- Published by Springer Nature in Nature Communications
- Vol. 17 (1)
Deep learning and superoscillatory speckles empowered multimode fiber probe for in situ nano-displacement detection and micro-imaging
High-precision metrology has laid the foundation for semiconductor fabrication and life sciences. However, existing displacement measurement approaches are incapable of performing flexible probing within complex equipment interiors. Here, we present a in situ, non-contact nano-displacement measurement approach. Leveraging a multimode fiber probe empowered by deep learning, fine feature information can be efficiently extracted from superoscillatory speckles, achieving single-ended detection with 10 nm resolution and 99.95% accuracy. A physical model is established to correlate the displacement with higher-order modes proportion in the fiber. Sub-millimeter-sized probe enables detecting targets with different structures in confined spaces. Robust recognition is achieved through joint learning, under varying fiber bending conditions and different metal materials. With extreme compression ratios of less than 0.1%, the system delivers high accuracy, low training costs, and high-speed processing. The imaging capability of the probe is also experimentally validated, proving potential as a powerful tool in applications such as lithography, weak force sensing, and super-resolution micro-endoscopy.
- 5 January 2026
- Published by Springer Nature in Nature Communications
- Vol. 17 (1)
Open AccessDeep learning and superoscillatory speckles empowered multimode fiber probe for in situ nano-displacement detection and micro-imaging
High-precision metrology has laid the foundation for semiconductor fabrication and life sciences. However, existing displacement measurement approaches are incapable of performing flexible probing within complex equipment interiors. Here, we present a in situ, non-contact nano-displacement measurement approach. Leveraging a multimode fiber probe empowered by deep learning, fine feature information can be efficiently extracted from superoscillatory speckles, achieving single-ended detection with 10 nm resolution and 99.95% accuracy. A physical model is established to correlate the displacement with higher-order modes proportion in the fiber. Sub-millimeter-sized probe enables detecting targets with different structures in confined spaces. Robust recognition is achieved through joint learning, under varying fiber bending conditions and different metal materials. With extreme compression ratios of less than 0.1%, the system delivers high accuracy, low training costs, and high-speed processing. The imaging capability of the probe is also experimentally validated, proving potential as a powerful tool in applications such as lithography, weak force sensing, and super-resolution micro-endoscopy.
- 30 August 2025
- Published by Wiley in Laser & Photonics Reviews
- Vol. 20 (2)
Multiplexing‐Enhanced Computational Imaging: High‐Fidelity Reconstruction and Color Imaging in Multimode Fibers
The advancement of computational imaging has propelled multimode fiber systems beyond conventional endoscopy by synergizing optical physics and deep learning to reconstruct images from disordered speckle patterns. However, existing multimode fiber (MMF) imaging predominantly relies on single‐channel transmission, limiting information capacity, imaging quality, and color fidelity. Here, multiplexing‐enhanced computational imaging is proposed to overcome these constraints through wavelength‐division multiplexing and multi‐channel fusion. By integrating inverse transmission matrix optimization, multi‐input Pix2Pix generative adversarial networks, and image fusion algorithms, the system reconstructs high‐fidelity color images from multiplexed speckle patterns encoded by RGB wavelengths in a single MMF. Experimental results demonstrate a structural similarity improvement from 0.649 to 0.799 and a 26.4% reduction in color deviation, achieving parallel data transmission and enhance imaging quality. A wavelength‐speckle relationship model and analyze multiplexing capacity limits are established, showing its extensibility to time‐, space‐, and polarization‐division multiplexing. This framework not only enables ultra‐thin full‐color micro‐endoscopy with superior information density, but also expands to hyperspectral imaging, optical neural networks, and quantum communication.
- 30 August 2025
- Published by Wiley in Laser & Photonics Reviews
- Vol. 20 (2)
Multiplexing‐Enhanced Computational Imaging: High‐Fidelity Reconstruction and Color Imaging in Multimode Fibers
The advancement of computational imaging has propelled multimode fiber systems beyond conventional endoscopy by synergizing optical physics and deep learning to reconstruct images from disordered speckle patterns. However, existing multimode fiber (MMF) imaging predominantly relies on single‐channel transmission, limiting information capacity, imaging quality, and color fidelity. Here, multiplexing‐enhanced computational imaging is proposed to overcome these constraints through wavelength‐division multiplexing and multi‐channel fusion. By integrating inverse transmission matrix optimization, multi‐input Pix2Pix generative adversarial networks, and image fusion algorithms, the system reconstructs high‐fidelity color images from multiplexed speckle patterns encoded by RGB wavelengths in a single MMF. Experimental results demonstrate a structural similarity improvement from 0.649 to 0.799 and a 26.4% reduction in color deviation, achieving parallel data transmission and enhance imaging quality. A wavelength‐speckle relationship model and analyze multiplexing capacity limits are established, showing its extensibility to time‐, space‐, and polarization‐division multiplexing. This framework not only enables ultra‐thin full‐color micro‐endoscopy with superior information density, but also expands to hyperspectral imaging, optical neural networks, and quantum communication.
- Open Access
- 14 May 2025
- Published by Springer Nature in Research Square
Deep learning and superoscillatory speckles empowered multimode fiber probe for in situ nano-displacement detection and micro-imaging
High-precision metrology has laid the foundation for semiconductor fabrication and life sciences. However, existing displacement measurement approaches are incapable of performing flexible probing within complex equipment interiors. Here, we present a novel in situ, non-contact nano-displacement measurement approach. Leveraging a multimode fiber probe empowered by deep learning, fine feature information can be efficiently extracted from superoscillatory speckles, achieving single-ended detection with 10 nm resolution and 99.95% accuracy. A physical model is established to correlate the displacement with higher-order modes proportion in the fiber. Sub-millimeter-sized probe enables detecting targets with different structures in confined spaces. Robust recognition is achieved through joint learning, under varying fiber bending conditions and different metal materials. With extreme compression ratios of less than 0.1%, the system delivers high accuracy, low training costs, and high speed processing. The imaging capability of the probe is also experimentally validated, proving great potential as a powerful tool in applications such as lithography, weak force sensing, and super-resolution micro-endoscopy.
- 14 May 2025
- Published by Springer Nature in Research Square
Open AccessDeep learning and superoscillatory speckles empowered multimode fiber probe for in situ nano-displacement detection and micro-imaging
High-precision metrology has laid the foundation for semiconductor fabrication and life sciences. However, existing displacement measurement approaches are incapable of performing flexible probing within complex equipment interiors. Here, we present a novel in situ, non-contact nano-displacement measurement approach. Leveraging a multimode fiber probe empowered by deep learning, fine feature information can be efficiently extracted from superoscillatory speckles, achieving single-ended detection with 10 nm resolution and 99.95% accuracy. A physical model is established to correlate the displacement with higher-order modes proportion in the fiber. Sub-millimeter-sized probe enables detecting targets with different structures in confined spaces. Robust recognition is achieved through joint learning, under varying fiber bending conditions and different metal materials. With extreme compression ratios of less than 0.1%, the system delivers high accuracy, low training costs, and high speed processing. The imaging capability of the probe is also experimentally validated, proving great potential as a powerful tool in applications such as lithography, weak force sensing, and super-resolution micro-endoscopy.
- Open Access
- 13 June 2024
- Published by Springer Nature in Research Square
Versatile multimode fiber network with high capacity enabled by deep learning
In the realm of spatial information transmission in multimode fiber (MMF), the MMF-based endoscopes and information encryption technologies have garnered considerable attention. However, existing designs are limited to establishing a single connection between one input node and one output node, thus constraining the capacity and application scenarios of MMF spatial information transmission. Here, we demonstrate a new concept of MMF-based physical networking for spatial information transmission, and develop a physical model and implementation method for establishing multi-node networking with various topological structures via cascading MMFs.We experimentally verify the feasibility of parallel transmission of spatial information at multiple nodes in an exemplary three-node MMF network with chain topology, showcasing its capability in transmitting color images through "node multiplexing" with significantly enhanced communication security through long-distance reprogrammable optical encryption. Designing MMF networks based on different node quantities and topological structures can significantly expand the scenarios for MMF spatial information transmission, providing valuable paradigms for various applications such as minimally invasive panoramic endoscopy, low-cost distributed sensing, and scaling optical reservoir computing.
- 13 June 2024
- Published by Springer Nature in Research Square
Open AccessVersatile multimode fiber network with high capacity enabled by deep learning
In the realm of spatial information transmission in multimode fiber (MMF), the MMF-based endoscopes and information encryption technologies have garnered considerable attention. However, existing designs are limited to establishing a single connection between one input node and one output node, thus constraining the capacity and application scenarios of MMF spatial information transmission. Here, we demonstrate a new concept of MMF-based physical networking for spatial information transmission, and develop a physical model and implementation method for establishing multi-node networking with various topological structures via cascading MMFs.We experimentally verify the feasibility of parallel transmission of spatial information at multiple nodes in an exemplary three-node MMF network with chain topology, showcasing its capability in transmitting color images through "node multiplexing" with significantly enhanced communication security through long-distance reprogrammable optical encryption. Designing MMF networks based on different node quantities and topological structures can significantly expand the scenarios for MMF spatial information transmission, providing valuable paradigms for various applications such as minimally invasive panoramic endoscopy, low-cost distributed sensing, and scaling optical reservoir computing.
- Open Access
- 1 October 2022
- Published by AIP Publishing in APL Photonics
- Vol. 7 (10) , 106104
Complex pattern transmission through multimode fiber under diverse light sources
Multimode fibers with high information capacity and ultra-thin diameter offer new possibilities for non-invasive endoscopy and remote high-speed secure communication. However, due to their sensitivity, special demands are thus raised on the light sources, which act as both lighting and information carrier. We build a single-arm multimode fiber image transmission system. The impact of five different sources on transmission quality is systematically compared and analyzed. High-quality transmission of complex grayscale patterns is achieved with inverse transmission matrix. The measured structural similarity exceeds 0.65. Experimental results indicate that random fiber laser with high time–domain stability is suitable for single-fiber imaging. This work is conducive to further study of fiber imaging and provides guidance for novel fiber communication and flexible endoscopy.
- 1 October 2022
- Published by AIP Publishing in APL Photonics
- Vol. 7 (10) , 106104
Open AccessComplex pattern transmission through multimode fiber under diverse light sources
Multimode fibers with high information capacity and ultra-thin diameter offer new possibilities for non-invasive endoscopy and remote high-speed secure communication. However, due to their sensitivity, special demands are thus raised on the light sources, which act as both lighting and information carrier. We build a single-arm multimode fiber image transmission system. The impact of five different sources on transmission quality is systematically compared and analyzed. High-quality transmission of complex grayscale patterns is achieved with inverse transmission matrix. The measured structural similarity exceeds 0.65. Experimental results indicate that random fiber laser with high time–domain stability is suitable for single-fiber imaging. This work is conducive to further study of fiber imaging and provides guidance for novel fiber communication and flexible endoscopy.
- 3 July 2022
- Published by Wiley in Laser & Photonics Reviews
- Vol. 16 (9)
High‐Speed All‐Fiber Micro‐Imaging with Large Depth of Field
Single fiber imaging has evolved into a powerful method for detecting minute objects in narrow spaces. However, existing systems are not conducive to imaging dynamic objects at depth due to their bulky probes, time‐consuming scanning acquisition methods, and transmissive illumination mode. Minimally invasive reflection mode imaging with high spatial and temporal resolution remains an open challenge. Here, a precise and high‐speed imaging scheme without scanning is proposed. Multimode fiber imaging technology is incorporated into an all‐fiber aberration‐free precision detection system. High temporal resolution (5000 fps) detection of tiny natural scenes is experimentally realized by optimizing the approximation of the inverse transmission matrix in a simple and compact setup. The system can display the detected screen in real‐time and the computational imaging with a large depth of field (1 mm) is enabled by jointly learning. The recovery results are superior compared to typical deep neural networks. The demonstrated scheme offers a new possibility for many applications, for example, microendoscopy, all‐optical computing, and remote high‐speed video transmission.
- 3 July 2022
- Published by Wiley in Laser & Photonics Reviews
- Vol. 16 (9)
High‐Speed All‐Fiber Micro‐Imaging with Large Depth of Field
Single fiber imaging has evolved into a powerful method for detecting minute objects in narrow spaces. However, existing systems are not conducive to imaging dynamic objects at depth due to their bulky probes, time‐consuming scanning acquisition methods, and transmissive illumination mode. Minimally invasive reflection mode imaging with high spatial and temporal resolution remains an open challenge. Here, a precise and high‐speed imaging scheme without scanning is proposed. Multimode fiber imaging technology is incorporated into an all‐fiber aberration‐free precision detection system. High temporal resolution (5000 fps) detection of tiny natural scenes is experimentally realized by optimizing the approximation of the inverse transmission matrix in a simple and compact setup. The system can display the detected screen in real‐time and the computational imaging with a large depth of field (1 mm) is enabled by jointly learning. The recovery results are superior compared to typical deep neural networks. The demonstrated scheme offers a new possibility for many applications, for example, microendoscopy, all‐optical computing, and remote high‐speed video transmission.
- Top Cited PapersOpen Access
- 17 March 2022
- Published by Springer Nature in Nature Communications
- Vol. 13 (1) , 1-8
All-fiber high-speed image detection enabled by deep learning
Ultra-high-speed imaging serves as a foundation for modern science. While in biomedicine, optical-fiber-based endoscopy is often required for in vivo applications, the combination of high speed with the fiber endoscopy, which is vital for exploring transient biomedical phenomena, still confronts some challenges. We propose all-fiber imaging at high speeds, which is achieved based on the transformation of two-dimensional spatial information into one-dimensional temporal pulsed streams by leveraging high intermodal dispersion in a multimode fiber. Neural networks are trained to reconstruct images from the temporal waveforms. It can not only detect content-aware images with high quality, but also detect images of different kinds from the training images with slightly reduced quality. The fiber probe can detect micron-scale objects with a high frame rate (15.4 Mfps) and large frame depth (10,000). This scheme combines high speeds with high mechanical flexibility and integration and may stimulate future research exploring various phenomena in vivo.
- 17 March 2022
- Published by Springer Nature in Nature Communications
- Vol. 13 (1) , 1-8
Top Cited PapersOpen AccessAll-fiber high-speed image detection enabled by deep learning
Ultra-high-speed imaging serves as a foundation for modern science. While in biomedicine, optical-fiber-based endoscopy is often required for in vivo applications, the combination of high speed with the fiber endoscopy, which is vital for exploring transient biomedical phenomena, still confronts some challenges. We propose all-fiber imaging at high speeds, which is achieved based on the transformation of two-dimensional spatial information into one-dimensional temporal pulsed streams by leveraging high intermodal dispersion in a multimode fiber. Neural networks are trained to reconstruct images from the temporal waveforms. It can not only detect content-aware images with high quality, but also detect images of different kinds from the training images with slightly reduced quality. The fiber probe can detect micron-scale objects with a high frame rate (15.4 Mfps) and large frame depth (10,000). This scheme combines high speeds with high mechanical flexibility and integration and may stimulate future research exploring various phenomena in vivo.
- 1 January 2022
- Published by Optica Publishing Group
- Vol. 9, SM5O.6
All-fiber scan-free precision detection with large depth of field
We demonstrate an all-fiber detecting system based on a single multimode fiber. The system showcases the capability to detect precision natural scenes by its inverse transmission matrix. Integrated reflective illumination mode makes it more accessible.
- 1 January 2022
- Published by Optica Publishing Group
- Vol. 9, SM5O.6
All-fiber scan-free precision detection with large depth of field
We demonstrate an all-fiber detecting system based on a single multimode fiber. The system showcases the capability to detect precision natural scenes by its inverse transmission matrix. Integrated reflective illumination mode makes it more accessible.