Imitation Learning for generalizable long-horizon visuomotor control (2024, SDSURF)
I am developing a behavior cloning method to learn robotic manipulations from unstructured demonstrations. The approach leverages a Generalizable Neural Field (GNF) to reconstruct deep 3D voxel representations, enriched with generalizable semantics using Stable Diffusion. Initially, we explored an actor predictor based on the Chain of Thought Control architecture; however, its success rate was unsatisfactory. Consequently, we transitioned to a scale autoregressive predictor inspired by Visual Autoregressive Modeling (VAR). The VAR transformer generates a 3D action space, from which Q-functions are derived to produce executable actions. The model is undergoing continuous upgrades to achieve superior performance compared to its predecessors (target at CoRL 2025)

Task offloading for networked UAVs using deep reinforcement learning (2024, SDSURF)
Networked Airborne Computing (NAC) addresses UAVs' limited onboard computing capacity by enabling resource sharing among multiple UAVs. In a random mobility scenario, and without prior system knowledge, we optimize task allocation through three Deep Reinforcement Learning algorithms: TD3, PPO, and DDPG. Simulations demonstrate that our methods significantly accelerate task execution compared to existing approaches. (journal paper in preparation)

Hand joint localization and Labeling with SpatialConfiguration-Net (2024, Pitt)
This work implements the convolutional SpatialConfiguration-Net (SCN) for anatomical landmark labeling by dividing the task into two sub-problems: one for locally accurate predictions and another for resolving ambiguities using spatial configurations. SCN combines heatmap predictions from both components, enabling efficient end-to-end training. Evaluations on open-source hand radiographs show that SCN achieves high performance even with limited training data.

In vivo subject-specific estimation of cervical intervertebral disc degeneration (2023–2024, Pitt)
A computational method is proposed to estimate the in vivo material degradation of cervical intervertebral discs. The kinematics of the cervical spinal column is modeled using a nonlinear finite element method, incorporating the latest nonlinear material properties from experiments. In vivo intervertebral motion data, acquired from biplane radiography for a given subject, were used to validate the spine mechanical model and inversely calibrate the deterioration of discs.
In Vivo Subject-Specific Estimation of Cervical Spine Discs Material Properties
Dongge Jia, Soumaya Ouhsousou, Clarissa M. LeVasseur, Jeremy Shaw, William Anderst, John C. Brigham
8th International Conference on Computational and Mathematical Biomedical Engineering 2024 (conference presentation)

An efficient static solver for the lattice discrete particle model (2022–2023, Pitt)
This study presents the first static solver for LDPM, introducing several fundamental innovations to overcome convergence issues. Additionally, graph coloring of the Jacobian operator, automatic differentiation, and advanced matrix manipulations are utilized to improve computational efficiency by up to 1000 times.
An Efficient Static Solver for the Lattice Discrete Particle Model
Dongge Jia, John C. Brigham, Alessandro Fascetti
Computer-Aided Civil and Infrastructure Engineering 2024 (journal paper, 5-year IF:10.8, 97th percentile in “Computational Theory and Mathematics”)
An Efficient Static Solver for the Lattice Discrete Particle Model
Dongge Jia, John C. Brigham, Alessandro Fascetti
Engineering Mechanics Institute Conference and Probabilistic Mechanics & Reliability Conference 2024 (conference presentation)

Lattice discrete particle modeling for coupled mechanical and mass transport simulation of reinforced concrete (2022–2023, Pitt)
This study introduces three new elements to the LDPM: mass transport elements through cracks, steel elements, and steel-to-concrete bond elements. These enhancements enable the model to simulate chloride transport in damaged reinforced concrete. The integrated model uses a fully dynamic solution strategy: reinforced concrete is solved with an explicit method and central difference scheme, while mass transport models use an implicit method with the Crank-Nicolson scheme.
A Novel Dual Lattice Discrete Particle Model for Multiphysics Simulation of Coupled Mechanical and Transport Behavior in Concrete Members Subjected to Long-Term Loading
Dongge Jia, Yingbo Zhu, John C. Brigham, Alessandro Fascetti
16th World Congress on Computational Mechanics and 4th Pan American Congress on Computational Mechanics 2024 (conference presentation)
Coupled Lattice Discrete Particle Model for the simulation of water and chloride transport in cracked concrete members
Yingbo Zhu (co-first author), Dongge Jia (co-first author, responsible for model and solution development), John C. Brigham, Alessandro Fascetti
Computer-Aided Civil and Infrastructure Engineering 2024 (journal paper, 5-year IF:10.8, 97th percentile in “Computational Theory and Mathematics”)

High-temperature-induced deterioration of fibre-reinforced polymer (FRP)-to-concrete bonded joints and the resultant reduction in the load-bearing capacity of FRP-strengthened bridge decks (2020–2021, SJTU)
Based on my theoretical analytical solution and high-temperature FRP-concrete interface shear test data, a precise constitutive model for the bond-slip behavior at high temperatures was developed. This model accurately captures the reduction in local bond strength, increased peeling deformation, and slower softening of bond stress due to high temperatures. A finite element model was then used to analyze the effects of fires and hot asphalt laying on the load-bearing capacity of FRP-strengthened concrete bridge decks.
Mechanical Behavior, Constitutive Model and Application of the FRP-to-Concrete Interface Under Coupled Effects of High Temperature and Loading
Dongge Jia
Master's Thesis, SJTU 2022

Analytical solution for the full-range behavior of FRP-to-concrete bonded joints under combined loading and temperature variations (2019–2021, SJTU)
This analytical study predicts the full-range deformation behavior of a mechanically loaded FRP-to-concrete bonded joint with arbitrary bond lengths under temperature variations. The closed-form solution isolates interfacial thermal stress effects from temperature-induced material property changes, accurately calibrating bond-slip characteristics and interfacial fracture energy.
Full-Range Behavior of FRP-to-Concrete Bonded Joints Subjected to Combined Effects of Loading and Temperature Variation
Dongge Jia, Wanyang Gao, Dexin Duan, Jian Yang, Jianguo Dai
Engineering Fracture Mechanics 2021 (journal paper, 5-year IF:4.8, 90th percentile in “Mechanical Engineering”)

An elastic visco-plastic nonlinear consolidation model of soft clay under cyclic loading (2017–2018, HUST)
Long-term consolidation and settlement analysis of soft clay under cyclic loading is critical for settlement prediction of infrastructures such as highways and railways. Currently available constitutive models of soft clay do not take into account of both creep settlement and the dynamic effect of train loading. This study establishes an elastic visco-plastic nonlinear consolidation model for cyclic loading conditions with nonlinear behavior of permeability coefficient k_v and volume compressibility coefficient m_v.
One Dimensional Elastic Visco-Plastic Nonlinear Consolidation Model of Soft Clay Under Cyclic Loading
Dongsheng Xu (advisor), Dongge Jia, Yangguang Zheng
Chinese Preprint 2018

Development of long-term settlement calculation software for soft clay foundations considering different creep effects (2017–2018, HUST)
To support the operation and maintenance of the Hangzhou Metro, a MATLAB-based software named Calcusettlement was developed. This software calculates settlement in multi-layered mixed soft clay foundations using consolidation models from Zhu and Yin (2005) and the U.S. Navy (1985). It also includes a module to account for creep effects and allows users to customize dynamic loading conditions.
MATLAB-Based Software: Long-Term Settlement Calculation Software for Soft Clay Foundations Considering Different Creep Effects
Dongsheng Xu (advisor), Dongge Jia
China Copyright Administration, No. 04768603, 2019. (Software copyright)

I am a passionate enthusiast of Chinese opera, with a particular fondness for Henan Opera, Quju Opera, and Yue Diao. In addition to opera, I enjoy popular music and have participated in numerous singing competitions. I appreciate all genres of music, whether ethnic, classical, popular, or experimental.