Project
- Quantitative Hardness Assessment with Vision-based Tactile Sensing for Fruit Classification and Grasping (2025)
Accurate estimation of fruit hardness is essential for automated classification and handling systems, particularly in determining fruit variety, assessing ripeness, and ensuring proper harvesting force. This study presents an innovative framework for quantitative hardness assessment utilizing vision-based tactile sensing, tailored explicitly for robotic applications in agriculture. The proposed methodology derives normal force estimation from a vision-based tactile sensor, and, based on the dynamics of this normal force, calculates the hardness. This approach offers a rapid, non-destructive evaluation through single-contact interaction. The integration of this framework into robotic systems enhances the real-time adaptability of grasping forces, thereby reducing the likelihood of fruit damage. Moreover, the general applicability of this approach, through a universal criterion based on average normal force dynamics, ensures its effectiveness across a wide variety of fruit types and sizes. Extensive experimental validation conducted across different fruit types and ripeness-tracking studies demonstrates the efficacy and robustness of the framework, marking a significant advancement in the domain of automated fruit handling.
- Digital Twin Systems for Reconfigurable Soft Robots (2023-2024)
In the rapidly evolving field of soft robotics, advancements in new materials, structural designs, and conceptual frameworks have propelled the rise of soft robot technology, particularly towards a highly versatile modular architecture with vast potential applications across various industries. However, one of the main challenges in this domain is the shape-morphing issue, as existing visualization and simulation tools struggle to adequately represent the complex and continuous deformation behaviors of soft robots. Furthermore, there is a distinct lack of intuitive, user-friendly platforms for visualizing and interactively controlling the shape-shifting capabilities of these robots. In response to these challenges, this paper introduces an innovative Digital Twin (DT) system specifically designed for reconfigurable soft robots, operating within an Augmented Reality (AR) environment. This system facilitates a more natural and accurate depiction of 3D soft deformations while providing an intuitive interface for simulation. We utilize a parameterized curve-driven method to dynamically adapt the DT in the AR space, ensuring smooth transitions between various 3D shape-morphing states. We identify three fundamental shape-morphing patterns—stretching, bending, and twisting—and create advanced visualization tools to precisely demonstrate these morphological changes. To enhance real-time representation of shape-morphing, we employ sensor fusion to detect and depict the soft robot's structural changes as parameterized curves. Our system is fully operational in an AR environment, empowering users to conduct immersive examinations and simulate reconfigurations of real-world soft robotic systems.
- Three-levels Digital Twin Systems for Human-Robot Interaction (2022-2023)
This research focuses on the development of a novel Augmented Reality (AR)-based digital twin system for human-robot interaction in manufacturing settings, with the aim of improving efficiency, accuracy, and adaptability. A three-level approach to digital twin technology is introduced, comprising virtual twin, hybrid twin, and cognitive twin levels, each offering unique functionalities. An intuitive AR-based interface is created, enabling users to interact with the digital twin through natural gestures, thereby streamlining programming and control processes. A comprehensive human-centric user study is conducted to validate the efficacy of the proposed system in minimizing setup time, reducing errors, and enhancing overall productivity.
Design, Optimization, and Manufacturing Systems for Reconfigurable Modular Soft Robots (2021-2023)
This research involves designing soft robots using Computer-Aided Design (CAD) software, such as SolidWorks and Inventor. Soft robot performance is optimized using simulation software, including COMSOL and MATLAB. Additive manufacturing and molding technologies are employed to fabricate soft robots. Additionally, the connection mechanism is designed for the reconfigurable modular soft robots.Data-Driven Shape Optimization Design for Auxetics Using Isogeometric Analysis (2019-2021)
This project implements a back-propagation neural network (BPNN)-based design framework for petal-shaped auxetics using isogeometric analysis, and proposes a deep neural networks (DNN) framework for tetra-chiral auxetics shape optimization design. The highly nonlinear relation between the input geometry variables and the effective material properties is fitted by a data-driven method (i.e., BPNN, DNN), facilitated by the NURBS-based parametric modeling scheme with a small number of design variables. This enables an easy analytical sensitivity analysis, demonstrating high accuracy and efficiency. Optimal auxetic structures are produced using 3D printing and experimentally tested for their properties. The implementation is based on the collaboration between Matlab and TensorFlow.
Multiple Accelerated Methods for Mesh-based and Isogeometric Topology Optimization (2018-2019)
This research proposes multilevel mesh, effective iterative methods, and local-update strategies to accelerate computing efficiency in topology optimization. For mesh-based topology optimization, the method projects density from coarse to fine meshes to accelerate convergence, adopts an initial-value-based preconditioned conjugate-gradient (PCG) method to solve the equations of Finite Element Analysis (FEA), and decreases the number of updated meshes according to their density. This results in a 35%-80% reduction in computational time compared to the classical TOP88 code. For isogeometric topology optimization, where control points are the design variables, the method applies *h*-refinement to subdivide the mesh, adopts a Multigrid conjugate gradient method (MGCG) to solve the equilibrium equation, and reduces control points according to their density. This successfully reduces 37%-93% of computational time compared to unaccelerated cases. The multiple accelerated methods perform better in large-scale cases and have general applicability for topology optimizations based on either meshes or control points.
Graded-density Lattice Structure Optimization Design Based on Topology Optimization (2017-2018)
This research proposes a multiscale topology optimization method based on the homogenization method, which generates graded-density lattice structures according to actual loads to achieve optimal performance. The method combines MATLAB with ANSYS, where optimization is performed in MATLAB and FEA computing and modeling are conducted in ANSYS. The lattice structure is rebuilt using Rhinoceros. Compared with the beam-model-based lattice optimization method from commercial software HyperWorks, the proposed method achieves better performance in mass reduction and stress distribution. The feasibility of manufacturing the lattice structure is demonstrated through 3D printing.