Xu Wang

Dr. Xu Wang

orcid

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
12 June 2023 in Nano Energy
Verified
Journal Article
Closed access
YMYuchen Miao
Yuchen Miao
Jun WuJun Wu
Jun Wu
XQXiaorong Qi
Xiaorong Qi
LYLiu Yang
Liu Yang
XWXu Wang
Xu Wang
ZHZiyang Hu
Ziyang Hu

4 September 2024 in IEEE Transactions on Neural Networks and Learning Systems
Verified
Journal Article
Closed access
YLYanfang Liu
Yanfang Liu
XWXu Wang
Xu Wang
YSYituo Song
Yituo Song
BWBo Wang
Bo Wang
DDDesong Du
Desong Du
Embedding the Hamiltonian formalisms into neural networks (NNs) enhances the reliability and precision of data-driven models, in which substantial research has been conducted. However, these approaches require the system to be represented in canonical coordinates, i.e., observed states should be generalized position-momentum pairs, which are typically unknown. This poses limitations when the method is applied to real-world data. Existing methods tackle this challenge through coordinate transformation or designing complex NNs to learn the symplectic phase flow of the state evolution. However, these approaches lack generality and are often difficult to train. This article proposes a versatile framework called general Hamiltonian NN (GHNN), which achieves coordinates free and handles sophisticated constraints automatically with concise form. GHNN employs two NNs, namely, an HNet to predict the Hamiltonian quantity and a JNet to predict the interconnection matrix. The gradients of the Hamiltonian quantity with respect to the input coordinates are calculated using automatic differentiation and are then multiplied by the interconnection matrix to obtain state differentials. Subsequently, ordinary differential equations (ODEs) are solved by numerical integration to provide state predictions. The accuracy and versatility of the GHNN are demonstrated through several challenging tasks, including the nonlinear simple and double pendulum, coupled pendulum, and real 3-D crane dynamic system.

19 January 2026 in Drones
Verified
Journal Article
Open access
PDF file
XWXu Wang
Xu Wang
YLYanfang Liu
Yanfang Liu
DDDesong Du
Desong Du
HXHuarui Xv
Huarui Xv
NQNaiming Qi
Naiming Qi
Establishing precise and reliable quadrotor dynamics model is crucial for safe and stable tracking control in obstacle environments. However, obtaining such models is challenging, as it requires precise inertia identification and accounting for complex aerodynamic effects, which handcrafted models struggle to do. To address this, this paper proposes a safety-critical control framework built on a Hamiltonian neural differential model (HDM). The HDM formulates the quadrotor dynamics under a Hamiltonian structure over the SE(3) manifold, with explicitly optimizable inertia parameters and a neural network-approximated control input matrix. This yields a neural ordinary differential equation (ODE) that is solved numerically for state prediction, while all parameters are trained jointly from data via gradient descent. Unlike black-box models, the HDM incorporates physical priors—such as SE(3) constraints and energy conservation—ensuring a physically plausible and interpretable dynamics representation. Furthermore, the HDM is reformulated into a control-affine form, enabling controller synthesis via control Lyapunov functions (CLFs) for stability and exponential control barrier functions (ECBFs) for rigorous safety guarantees. Simulations validate the framework’s effectiveness in achieving safe and stable tracking control.

20 May 2026 in Aerospace
Verified
Journal Article
Open access
PDF file
NQNaiming Qi
Naiming Qi
LHLong He
Long He
RZRui Zhou
Rui Zhou
KLKaiyuan Liu
Kaiyuan Liu
XWXu Wang
Xu Wang
LYLi Yao
Li Yao
To meet the high-precision and automated requirements for the insertion assembly between large-scale components and non-cooperative outer shells, an impedance-controlled large-component insertion assembly technology, namely compliant insertion assembly technology, is proposed. This paper explains the working principle of the technology from a theoretical perspective, elaborates on two key technical aspects—pose control and force-following control based on a parallel mechanism—and conducts horizontal insertion assembly simulation for components. The simulation results demonstrate that force-following control via the parallel mechanism can reduce the axial pose accuracy error between the component and the shell by more than 85%, meeting the pose accuracy requirements for insertion assembly. It is also verified that force-following control can adjust the pose of the shell in real time based on the coaxiality between the component and the shell, satisfying the minor deformation requirements during the insertion assembly process.

15 May 2026 in Aerospace
Verified
Journal Article
Open access
PDF file
LHLong He
Long He
PCPeilun Cai
Peilun Cai
RZRui Zhou
Rui Zhou
XWXu Wang
Xu Wang
LYLi Yao
Li Yao
NQNaiming Qi
Naiming Qi
The bonding between cabin sections and exterior shells represents a critical manufacturing operation in shell assembly, directly determining the reliability and structural performance of the assembled structure. However, traditional manual scraping suffers from low efficiency, poor consistency, and heavy reliance on manual operation, while conventional teach-and-repeat robotic automation fails to adapt to significant manufacturing tolerances and complex surface curvatures common in large-scale shell components. To address these challenges, this paper proposes a vision-guided robotic scraping method that generates adaptive trajectories on irregular cabin sections. The method achieves full pipeline integration and is particularly suited for production lines where various models share similar macro-geometries but possess subtle geometric variations. A system integrating a laser profile sensor is developed to perceive surface geometry and local normal vectors. By establishing a unified coordinate transformation chain and a scan–mesh–spline workflow, the sensed geometric information is directly mapped to the robot end-effector pose. A trajectory generation algorithm based on point cloud meshing and B-spline interpolation is employed to construct continuous, smooth scraping paths that accommodate geometric deviations without relying on complex fixtures. Unlike RGB-D correction-based methods that require pre-programmed initial trajectories, or CAD-driven offline programming that cannot adapt to manufacturing deviations, the proposed approach directly generates conformal scraping paths from measured geometry. Experimental results on a typical cabin section demonstrate that the generated trajectories accurately follow the surface normals, achieving a low standard deviation of 36 μm in adhesive layer thickness, indicating excellent thickness consistency and uniformity. Furthermore, the automated process reduced the total operation time to approximately 40 min, improving production efficiency by more than two times compared to manual operations, thereby validating the robustness and suitability of the method for high-precision batch manufacturing.

1 July 2026 in IEEE Robotics and Automation Letters
Verified
Journal Article
Closed access
YSYituo Song
Yituo Song
YLYanfang Liu
Yanfang Liu
JSJianhang Sun
Jianhang Sun
MLMuhang Liu
Muhang Liu
DKDebo Kong
Debo Kong
LCLei Chen
Lei Chen