Erbao Dong
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
Research Interests
Erbao Dong has a total of 166 co-authors. Here are the top co-authors by number of shared publications:
- Qiqiang Hu
Department of Precision Machinery and Precision Instrumentation, Institute of Humanoid Robots, University of Science and Technology of China, Hefei, China
9 shared publications - Dong Sun
Department of Biomedical Engineering, City University of Hong Kong, Kowloon, Hong Kong, China
7 shared publications Hu Jin
Department of Precision Machinery and Precision Instrumentation, Institute of Humanoid Robots, University of Science and Technology of China, Hefei, China
6 shared publicationsJie Yang
Department of Precision Machinery and Precision Instrumentation, University of Science and Technology of China, Hefei, China
5 shared publications- Shiwu Zhang
University of Science and Technology of China
5 shared publications - Guangrui Bai
University of Science and Technology of China
4 shared publications
Latest Publications
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Informatics
- Vol. 22 (11) , 9765-9776
Adaptive Mask-Guided Multiscale Contrastive Learning Framework for Novel Class Discovery of Industrial Anomalies
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Informatics
- Vol. 22 (11) , 9765-9776
Adaptive Mask-Guided Multiscale Contrastive Learning Framework for Novel Class Discovery of Industrial Anomalies
- 1 October 2026
- Published by IOP Publishing in Engineering Research Express
- Vol. 8 (19) , 195213
Low-light image enhancement via enhancement-guided adaptive refinement and progressive rational mapping
Low-light images suffer from coupled degradations of insufficient illumination, low contrast and amplified noise. Curve-based methods needing no paired supervision are efficient, but they correct luminance and rarely address the noise that amplification makes visible. This paper couples an iterative curve estimator with a refinement module conditioned on the enhancement process itself. The enhancement module predicts a per-pixel curve parameter and a gate for each of eight iterations; the refinement module consumes a twelve-channel prior built from those quantities and performs spatially adaptive restoration in the regions where amplification is largest. The refinement module carries the method. Over three independently trained seeds, removing it costs 0.162 ± 0.028 structural similarity index, several times its own spread, against 0.046 across the three main curve families (gamma, rational, polynomial), each measured over three seeds. Removing only its guidance prior costs 0.050 ± 0.049 , positive on every seed but not distinguishable from zero on three runs, so we claim the module without apportioning its value internally. The model uses 0.271 M parameters and 35.35 GFLOPs at 256 × 256 . Results are means over three seeds with their spread, since a different seed alone moves LOLv2-Real peak signal-to-noise ratio (PSNR) by up to 1.52 dB. Three findings qualify the design and we report them as measured. Gamma correction attains a higher PSNR than the rational mapping we adopt, so we present it as a characterised design choice, not a demonstrated optimum. The module removes the synthetic noise it is trained on but recovers only 0.33 ± 0.09 dB on real sensor noise, under two per cent of the measured noise level. And performance varies by more than ten decibels with input illumination. Model selection uses paired references from held-out training scenes, so the training objective is reference-free while the pipeline is not.
- 1 October 2026
- Published by IOP Publishing in Engineering Research Express
- Vol. 8 (19) , 195213
Low-light image enhancement via enhancement-guided adaptive refinement and progressive rational mapping
Low-light images suffer from coupled degradations of insufficient illumination, low contrast and amplified noise. Curve-based methods needing no paired supervision are efficient, but they correct luminance and rarely address the noise that amplification makes visible. This paper couples an iterative curve estimator with a refinement module conditioned on the enhancement process itself. The enhancement module predicts a per-pixel curve parameter and a gate for each of eight iterations; the refinement module consumes a twelve-channel prior built from those quantities and performs spatially adaptive restoration in the regions where amplification is largest. The refinement module carries the method. Over three independently trained seeds, removing it costs 0.162 ± 0.028 structural similarity index, several times its own spread, against 0.046 across the three main curve families (gamma, rational, polynomial), each measured over three seeds. Removing only its guidance prior costs 0.050 ± 0.049 , positive on every seed but not distinguishable from zero on three runs, so we claim the module without apportioning its value internally. The model uses 0.271 M parameters and 35.35 GFLOPs at 256 × 256 . Results are means over three seeds with their spread, since a different seed alone moves LOLv2-Real peak signal-to-noise ratio (PSNR) by up to 1.52 dB. Three findings qualify the design and we report them as measured. Gamma correction attains a higher PSNR than the rational mapping we adopt, so we present it as a characterised design choice, not a demonstrated optimum. The module removes the synthetic noise it is trained on but recovers only 0.33 ± 0.09 dB on real sensor noise, under two per cent of the measured noise level. And performance varies by more than ten decibels with input illumination. Model selection uses paired references from held-out training scenes, so the training objective is reference-free while the pipeline is not.
- 5 June 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Automation Science and Engineering
- Vol. 23 (15455955) , 11228-11244
Decoupling Control of a Multi-Segment Hybrid-Actuated Soft Origami Continuum Robot Through Variable Stiffness
Soft origami continuum robots have attracted considerable attention because of their flexibility and adaptability, but they face challenges in control accuracy. This study presents a novel design and control scheme for large-deformation hybrid-actuated soft origami continuum robots to improve their motion performance in real tasks. An origami-based pneumatic chamber is used as the robot backbone to achieve a high extension ratio, and the hybrid variable stiffness principle combining antagonistic actuation and layer jamming further enhances the robot's overall structural stiffness. The robot performs precise movements by controlling tendons distributed in an external origami. An iterative training strategy based on long short-term memory is employed to model the inverse kinematics of the robot in consideration of the inherent hysteresis of origami robots. Step size features are introduced to improve model accuracy with limited data. The capability for variable stiffness enables the migration of the training model on the basis of a single segment, the accuracy of which is close to the submillimeter level. Decoupling control of each segment for a rear-driven multi-segment origami continuum robot is also achieved. Experiment results reveal that the trajectory tracking errors for single-segment and multi-segment of the robot are 1.58, and 2.81 mm, respectively, with relative errors of 0.71% and 0.63% over the robot length, demonstrating the good performance of the proposed design and control method. Orientation data can also be added to the dataset to achieve orientation control of the robot, with an angle error of 0.75°. The robot shows its stability and safety in a series of real tasks, including writing, LED tracking, board cleaning, and pick and place. This study demonstrates a potential solution for continuum robots performing precise tasks without any sensory feedback in confined spaces and specialized environments where electronic sensors fail.
- 5 June 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Automation Science and Engineering
- Vol. 23 (15455955) , 11228-11244
Decoupling Control of a Multi-Segment Hybrid-Actuated Soft Origami Continuum Robot Through Variable Stiffness
Soft origami continuum robots have attracted considerable attention because of their flexibility and adaptability, but they face challenges in control accuracy. This study presents a novel design and control scheme for large-deformation hybrid-actuated soft origami continuum robots to improve their motion performance in real tasks. An origami-based pneumatic chamber is used as the robot backbone to achieve a high extension ratio, and the hybrid variable stiffness principle combining antagonistic actuation and layer jamming further enhances the robot's overall structural stiffness. The robot performs precise movements by controlling tendons distributed in an external origami. An iterative training strategy based on long short-term memory is employed to model the inverse kinematics of the robot in consideration of the inherent hysteresis of origami robots. Step size features are introduced to improve model accuracy with limited data. The capability for variable stiffness enables the migration of the training model on the basis of a single segment, the accuracy of which is close to the submillimeter level. Decoupling control of each segment for a rear-driven multi-segment origami continuum robot is also achieved. Experiment results reveal that the trajectory tracking errors for single-segment and multi-segment of the robot are 1.58, and 2.81 mm, respectively, with relative errors of 0.71% and 0.63% over the robot length, demonstrating the good performance of the proposed design and control method. Orientation data can also be added to the dataset to achieve orientation control of the robot, with an angle error of 0.75°. The robot shows its stability and safety in a series of real tasks, including writing, LED tracking, board cleaning, and pick and place. This study demonstrates a potential solution for continuum robots performing precise tasks without any sensory feedback in confined spaces and specialized environments where electronic sensors fail.
- 2 June 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Medical Robotics and Bionics
- Vol. PP (3) , 1
Design and Evaluation of a Compact, MR-Compatible Origami Continuum Robot with Variable Length
Continuum robots have recently gained attention for their great value in medical applications. This paper presents a novel design approach for MR-compatible medical continuum robots with the ability to variable length and controllable motion at compact size and light weight. The robot employs a small-scale origami continuum structure with an outside diameter of 15 mm, is driven by three tendon-shape memory alloy (SMA) actuators, and senses the motion of actuators by a thin linear potentiometer (Thinpot). The robot system is lightweight at 76.7 g and compact, with the proximal sensing and actuation module measuring only 34 mm in diameter and 90 mm in length. It is first demonstrated that the Thinpot is insensitive to magnetic disturbance and MR-compatible and can realize accurate feedback control of the large stroke of SMA actuators, even in a real 1.5T MR environment. Further, the bending angle and tip position of the robot can be controlled based on an inverse kinematic model. The robots equipped with different tools will show versatility in biomedical applications, such as endoscopy and needle insertion.
- 2 June 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Medical Robotics and Bionics
- Vol. PP (3) , 1
Design and Evaluation of a Compact, MR-Compatible Origami Continuum Robot with Variable Length
Continuum robots have recently gained attention for their great value in medical applications. This paper presents a novel design approach for MR-compatible medical continuum robots with the ability to variable length and controllable motion at compact size and light weight. The robot employs a small-scale origami continuum structure with an outside diameter of 15 mm, is driven by three tendon-shape memory alloy (SMA) actuators, and senses the motion of actuators by a thin linear potentiometer (Thinpot). The robot system is lightweight at 76.7 g and compact, with the proximal sensing and actuation module measuring only 34 mm in diameter and 90 mm in length. It is first demonstrated that the Thinpot is insensitive to magnetic disturbance and MR-compatible and can realize accurate feedback control of the large stroke of SMA actuators, even in a real 1.5T MR environment. Further, the bending angle and tip position of the robot can be controlled based on an inverse kinematic model. The robots equipped with different tools will show versatility in biomedical applications, such as endoscopy and needle insertion.
- 6 May 2026
- Published by Springer Nature in Multimedia Systems
- Vol. 32 (4) , 267
A unified lightweight enhancement–segmentation framework for one-shot recognition in complex outdoor lighting
- 6 May 2026
- Published by Springer Nature in Multimedia Systems
- Vol. 32 (4) , 267
A unified lightweight enhancement–segmentation framework for one-shot recognition in complex outdoor lighting
- Open Access
- 16 January 2026
- Published by Institution of Engineering and Technology (IET) in High Voltage
- Vol. 11 (4) , 1102-1116
YOLO‐GDCNN: Real‐Time Operating Point Detection for Live Working Robots in the Power Industry
In the power industry maintenance, the capability of live working robots to detect and operate with power components in real time is paramount. This paper proposes a cascaded detection framework for real‐time detection of live working operation points, named YOLO‐GDCNN. The framework consists of two parts. The first part is the proposed Lightweight YOLOv5 (Li‐YOLOv5), which is composed of Stem, ShuffleBlock and SimSPPF, enhancing the network's generalisation ability while reducing the computational load. The second part proposes a Generative Dense Convolutional Neural Network (GDCNN), which improves the detection accuracy of the lightweight network through dense modules. Through the series of the above modules, the framework successfully combines detection accuracy and real‐time performance. Finally, we design experiments in simulated laboratory environments and real‐world live working scenarios. The experimental results demonstrate that the proposed YOLO‐GDCNN is adept at real‐time and accurate detection of live working object types and their operating point, demonstrating its suitability for live working robots.
- 16 January 2026
- Published by Institution of Engineering and Technology (IET) in High Voltage
- Vol. 11 (4) , 1102-1116
Open AccessYOLO‐GDCNN: Real‐Time Operating Point Detection for Live Working Robots in the Power Industry
In the power industry maintenance, the capability of live working robots to detect and operate with power components in real time is paramount. This paper proposes a cascaded detection framework for real‐time detection of live working operation points, named YOLO‐GDCNN. The framework consists of two parts. The first part is the proposed Lightweight YOLOv5 (Li‐YOLOv5), which is composed of Stem, ShuffleBlock and SimSPPF, enhancing the network's generalisation ability while reducing the computational load. The second part proposes a Generative Dense Convolutional Neural Network (GDCNN), which improves the detection accuracy of the lightweight network through dense modules. Through the series of the above modules, the framework successfully combines detection accuracy and real‐time performance. Finally, we design experiments in simulated laboratory environments and real‐world live working scenarios. The experimental results demonstrate that the proposed YOLO‐GDCNN is adept at real‐time and accurate detection of live working object types and their operating point, demonstrating its suitability for live working robots.
- 2 January 2026
- Published by Springer Nature
- p. 39-47
Synergistic Pushing and Grasping in Cluttered Environments via Reinforcement Learning
- 2 January 2026
- Published by Springer Nature
- p. 39-47
Synergistic Pushing and Grasping in Cluttered Environments via Reinforcement Learning
- 19 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 10 (12) , 13344-13351
Customize-Your-Joy Hand: A User-Oriented, Cost-Effective 22-DOF Platform for Future Human-Robot Community
Humanoid dexterous hands have significant potential in prosthetics, service robotics, and high-performance manipulation. However, existing designs often struggle to balance the challenging requirements of lightweight design, high biomimicry, personalized customization, and low cost. To address these challenges, we present the CYJ Hand (Customize-Your-Joy Hand), an innovative 22-DOF humanoid dexterous hand. Featuring a highly biomimetic structure, the CYJ system weighs only 750 grams(forearm included). Its modular design supports user-oriented customization while simplifying assembly, maintenance, and functional expansion. Inspired by Da Vinci's mechanics, the CYJ Hand integrates a novel, controllable tendon mechanism that allows for reconfigurable tendon routing and actuation system to meet diverse needs. Constructed with 3D printing and affordable commercial materials, the hardware cost for the CYJ Hand structure (excluding actuators) is under $60. Experimental results demonstrate that the CYJ Hand achieves a 100% success rate in both the Kapandji Test and GRASP taxonomy, and further exhibits dynamic grasping, in-hand manipulation, sub-millimeter motion repeatability (∼0.7 mm), and reliable load-bearing performance, validating its exceptional dexterity and biomimetic design. With its comprehensive advantages and innovations, the CYJ Hand provides a versatile platform for the future applications and research in personalized prosthetics and dexterous robotic manipulation, bridging the gap between high dexterity and accessibility in humanoid robotics. Related files and methods are open-sourced at GitHub repository.
- 19 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 10 (12) , 13344-13351
Customize-Your-Joy Hand: A User-Oriented, Cost-Effective 22-DOF Platform for Future Human-Robot Community
Humanoid dexterous hands have significant potential in prosthetics, service robotics, and high-performance manipulation. However, existing designs often struggle to balance the challenging requirements of lightweight design, high biomimicry, personalized customization, and low cost. To address these challenges, we present the CYJ Hand (Customize-Your-Joy Hand), an innovative 22-DOF humanoid dexterous hand. Featuring a highly biomimetic structure, the CYJ system weighs only 750 grams(forearm included). Its modular design supports user-oriented customization while simplifying assembly, maintenance, and functional expansion. Inspired by Da Vinci's mechanics, the CYJ Hand integrates a novel, controllable tendon mechanism that allows for reconfigurable tendon routing and actuation system to meet diverse needs. Constructed with 3D printing and affordable commercial materials, the hardware cost for the CYJ Hand structure (excluding actuators) is under $60. Experimental results demonstrate that the CYJ Hand achieves a 100% success rate in both the Kapandji Test and GRASP taxonomy, and further exhibits dynamic grasping, in-hand manipulation, sub-millimeter motion repeatability (∼0.7 mm), and reliable load-bearing performance, validating its exceptional dexterity and biomimetic design. With its comprehensive advantages and innovations, the CYJ Hand provides a versatile platform for the future applications and research in personalized prosthetics and dexterous robotic manipulation, bridging the gap between high dexterity and accessibility in humanoid robotics. Related files and methods are open-sourced at GitHub repository.
- 1 September 2025
- Published by IOP Publishing in Smart Materials and Structures
- Vol. 34 (9) , 095030
A standardized SMA modular actuator enabled by ultrasonic welding technique and an integrated control system
Shape memory alloys (SMAs) are widely used in soft robotics owing to their compact structure, high power-to-weight ratio, and reversible deformation. However, most existing SMA actuators still rely on manual assembly and lack standardization, which undermines reliability and scalability. Although resistance-feedback and drive–sense approaches have been reported in the literature, many solutions are not systematically integrated into standardized actuator modules and often involve trade-offs in sensing precision, sampling rate, or manufacturability, thereby hindering high-precision control in complex environments. To bridge these gaps, we present a universal SMA actuator module–termed the enhanced smart modular structure (eSMS)–based on ultrasonic welding (USW) technology and an integrated drive–sense-control system. Using the USW technique, the module achieves high fastening force (12.58 N/wire), consistent wire length and tension, and lightweight joints, resulting in a 183% improvement in cycle life. The integrated circuit operates at low voltage (7.4 V) and high current (up to 8.4 A), and enables high-precision resistance self-sensing (error < 0.28 % ) at a driving frequency of 100 Hz, markedly improving real-time control. Experimental results show that, under vision-based feedback control, the eSMS module exhibits robust bidirectional motion, high positioning accuracy (error < 3.64 ∘ ), and rapid response. The proposed approach enhances the reliability and precision of SMA systems, shortens development cycles, and reduces manufacturing errors, offering a practical pathway toward standardized SMA actuator production with promising applications in biomimetic robotics, flexible gripping systems, and wearable devices.
- 1 September 2025
- Published by IOP Publishing in Smart Materials and Structures
- Vol. 34 (9) , 095030
A standardized SMA modular actuator enabled by ultrasonic welding technique and an integrated control system
Shape memory alloys (SMAs) are widely used in soft robotics owing to their compact structure, high power-to-weight ratio, and reversible deformation. However, most existing SMA actuators still rely on manual assembly and lack standardization, which undermines reliability and scalability. Although resistance-feedback and drive–sense approaches have been reported in the literature, many solutions are not systematically integrated into standardized actuator modules and often involve trade-offs in sensing precision, sampling rate, or manufacturability, thereby hindering high-precision control in complex environments. To bridge these gaps, we present a universal SMA actuator module–termed the enhanced smart modular structure (eSMS)–based on ultrasonic welding (USW) technology and an integrated drive–sense-control system. Using the USW technique, the module achieves high fastening force (12.58 N/wire), consistent wire length and tension, and lightweight joints, resulting in a 183% improvement in cycle life. The integrated circuit operates at low voltage (7.4 V) and high current (up to 8.4 A), and enables high-precision resistance self-sensing (error < 0.28 % ) at a driving frequency of 100 Hz, markedly improving real-time control. Experimental results show that, under vision-based feedback control, the eSMS module exhibits robust bidirectional motion, high positioning accuracy (error < 3.64 ∘ ), and rapid response. The proposed approach enhances the reliability and precision of SMA systems, shortens development cycles, and reduces manufacturing errors, offering a practical pathway toward standardized SMA actuator production with promising applications in biomimetic robotics, flexible gripping systems, and wearable devices.
- 21 August 2025
- Published by Springer Nature in Multimedia Systems
- Vol. 31 (5) , 1-15
Low-light image enhancement based on adaptive enhancement matrix
- 21 August 2025
- Published by Springer Nature in Multimedia Systems
- Vol. 31 (5) , 1-15
Low-light image enhancement based on adaptive enhancement matrix