Institution profile

China Pharmaceutical University

Academic institutionasia · cn
Official website
Research library11linked papers
Opportunities0open roles
Selected work

Representative Papers

Can a Robot Read Braille? - Learning to Adapt Contact via Imitation Learning for Tactile Braille Recognition

Sep 24, 2026

This study addresses the problem of inadequate and often overlooked physical contact quality in robotic Braille reading by proposing an adaptive contact framework. This method introduces active contact adjustment into the Braille reading pipeline for the first time, jointly optimizing contact acceptability and pose correction through imitation learning. Furthermore, it integrates multi-head policy learning with pose-aware tactile fusion to achieve reliable reconstruction. Experimental evaluations on 20 physical Braille boards demonstrate that the proposed system attains a 94.0% tactile quality assessment score and an 88.6% reconstruction accuracy. These results validate the critical role of the active contact mechanism in enhancing recognition robustness for tactile reading tasks.

0 citationsRead paper

CAMP: Cooperative Arm-Hand Motion Planning in Constrained Spaces

Sep 24, 2026

This study addresses the limitations of decomposition-based methods, which often miss feasible solutions, and the inefficiency of high-dimensional search in coordinated planning for robotic arms and dexterous hands within confined spaces. To this end, we propose CAMP, a coordinated planning framework whose core innovation lies in constructing feasible hand fibers to precisely characterize arm-hand coupling. The method combines hierarchical hand search with local arm relaxation to generate initial trajectories, and introduces endpoint-preserving Via-point Movement Primitives (VMPs) to enable compact representation and coarse-to-fine joint optimization. Experimental results demonstrate that CAMP achieves success rates ranging from 84.2% to 98.5% across six simulated tasks, significantly outperforming existing baselines. Furthermore, real-world experiments validate its practical effectiveness on physical hardware.

0 citationsRead paper

RAVEN: Frozen Random Graph Reservoirs with Physics-Informed Interaction Fingerprints for Protein-Ligand Binding Affinity Prediction

Aug 09, 2026

Accurate prediction of protein–ligand binding affinity is hindered by data scarcity, experimental heterogeneity, and conformational dependence. This work proposes a novel paradigm that leverages a multi-head frozen stochastic atom graph encoder to generate diverse structural representations, integrates explicit physicochemical interaction fingerprints, and employs a heterogeneous regressor combining neural networks and tree-based models. To enhance generalization and robustness, the approach avoids end-to-end training and instead adopts a validation-based non-negative fusion strategy. Evaluated on the GEMS-reconstructed PDBbind 2020R1 similarity-isolated split and the CASF-2016 benchmark, the method demonstrates consistently strong and stable predictive performance.

0 citationsRead paper
Recent publications

Latest Papers

Can a Robot Read Braille? - Learning to Adapt Contact via Imitation Learning for Tactile Braille Recognition

Sep 24, 2026

This study addresses the problem of inadequate and often overlooked physical contact quality in robotic Braille reading by proposing an adaptive contact framework. This method introduces active contact adjustment into the Braille reading pipeline for the first time, jointly optimizing contact acceptability and pose correction through imitation learning. Furthermore, it integrates multi-head policy learning with pose-aware tactile fusion to achieve reliable reconstruction. Experimental evaluations on 20 physical Braille boards demonstrate that the proposed system attains a 94.0% tactile quality assessment score and an 88.6% reconstruction accuracy. These results validate the critical role of the active contact mechanism in enhancing recognition robustness for tactile reading tasks.

0 citationsRead paper

CAMP: Cooperative Arm-Hand Motion Planning in Constrained Spaces

Sep 24, 2026

This study addresses the limitations of decomposition-based methods, which often miss feasible solutions, and the inefficiency of high-dimensional search in coordinated planning for robotic arms and dexterous hands within confined spaces. To this end, we propose CAMP, a coordinated planning framework whose core innovation lies in constructing feasible hand fibers to precisely characterize arm-hand coupling. The method combines hierarchical hand search with local arm relaxation to generate initial trajectories, and introduces endpoint-preserving Via-point Movement Primitives (VMPs) to enable compact representation and coarse-to-fine joint optimization. Experimental results demonstrate that CAMP achieves success rates ranging from 84.2% to 98.5% across six simulated tasks, significantly outperforming existing baselines. Furthermore, real-world experiments validate its practical effectiveness on physical hardware.

0 citationsRead paper

RAVEN: Frozen Random Graph Reservoirs with Physics-Informed Interaction Fingerprints for Protein-Ligand Binding Affinity Prediction

Aug 09, 2026

Accurate prediction of protein–ligand binding affinity is hindered by data scarcity, experimental heterogeneity, and conformational dependence. This work proposes a novel paradigm that leverages a multi-head frozen stochastic atom graph encoder to generate diverse structural representations, integrates explicit physicochemical interaction fingerprints, and employs a heterogeneous regressor combining neural networks and tree-based models. To enhance generalization and robustness, the approach avoids end-to-end training and instead adopts a validation-based non-negative fusion strategy. Evaluated on the GEMS-reconstructed PDBbind 2020R1 similarity-isolated split and the CASF-2016 benchmark, the method demonstrates consistently strong and stable predictive performance.

0 citationsRead paper