Jing Dong

Ph.D. student in Computer Science, Georgia Institute of Technology

Email, Google Scholar, Github, Linkedin

I'm now a 5th year Ph.D. student in computer science, working with professor Frank Dellaert and professor Byron Boots at Georgia Institute of Technology. Prior to joining Georgia Tech, I got my bachelor degree in engineering mechanics and aerospace engineering from Tsinghua University, Beijing, China.

My current research interests cover various topics in robotics and computer vision, include but not limit to Simultaneous Localization and Mapping (SLAM), 3D reconstruction, visual feature learning and matching, and real-time motion planning.

I'm looking for full-time research scientist or engineer positions start at mid 2018 in US, with special interests in SLAM, mapping and motion planning. Please refer to my CV.

News



Research Projects


4D Agriculture: Spatio-Temporal 3D Reconstruction for Precision Agriculture


Automatic crop mornitoring by low-cost UAVs and UGVs is an important task in precision agriculture. In this project we collected a multi-years dataset (from 2013) at Tifton, GA with low-cost ground vehicles and drones, and proposed a full-pipeline to perform large scale spatio-temporal 3D reconstruction, called 4D reconstruction (3D + time) over the field. After 4D reconstruction, we further process the 4D results to get crop information like height, growth rate, etc.


Publication:

[1] J. Dong, J. Burnham, B. Boots, G. C. Rains, F. Dellaert, 4D Crop Monitoring: Spatio-Temporal Reconstruction for Agriculture, in ICRA 2017. [PDF]

[2] K. Ahlin et al., Robotics for Spatially and Temporally Unstructured Agricultural Environments. Book chapter in Robotics and Mechatronics for Agriculture, 2017.

[3] L. Carlone, J. Dong, S. Fenu, G. C. Rains, F. Dellaert, Towards 4D Crop Analysis in Precision Agriculture: Estimating Plant Height and Crown Radius over Time via Expectation-Maximization, in ICRA Workshop on Robotics in Agriculture, 2015. [PDF]


Real-time and Online Trajectory Optimization as a Probabilistic Inference Framework


In this project we proposed a tajectory optimization framework using sparse Gaussian process (GP), formulating motion planning or state estimation problems as probabilistic inference on factor graphs, and efficiently solve the problems by non-linear least squares or Bayes trees. Evaluation shows our C++ implementation of proposed algorithm reaches real-time performance on online state estimation and motion planning tasks.

We have released the open-source C++/Matlab code at GPMP2 for motion planning tasks, and GP-SLAM for state estimation/SLAM tasks.


Publication: (* equal contribution)

[1] J. Dong, M. Mukadam, B. Boots, F. Dellaert, Sparse Gaussian Processes on Matrix Lie Groups: A Unified Framework for Optimizing Continuous-Time Trajectories, accepted to ICRA 2018. [PDF coming soon]

[2] M. Mukadam*, J. Dong*, Y. Yan, F. Dellaert, B. Boots, Continuous-Time Gaussian Process Motion Planning via Probabilistic Inference, ArXiv preprint 1707.07383, 2017, conditionally accepted in IJRR. [PDF]

[3] M. Mukadam, J. Dong, F. Dellaert, B. Boots, Simultaneous Trajectory Estimation and Planning via Probabilistic Inference, in RSS 2017. [PDF]

[4] J. Dong, B. Boots, F. Dellaert, Sparse Gaussian Processes for Continuous-Time Trajectory Estimation on Matrix Lie Groups, ArXiv preprint 1705.06020, 2017. [PDF]

[5] J. Dong, M. Mukadam, F. Dellaert, B. Boots, Motion Planning as Probabilistic Inference using Gaussian Processes and Factor Graphs, in RSS 2016. [PDF]


Real-time Distributed and Cooperative Multi-robot SLAM


We demonstrated a distributed, online, and real-time cooperative SLAM between multiple robots operating throughout an unknown environment with unknown initial deployment. We present a Expectation Maximization (EM) based approach to efficiently online identify inlier multi-robot loop closures by incorporating robot pose uncertainty, which significantly improves the accuracy over long-term navigation. Evaluation was performed on CMU quadrotor swarm.


Publication:

[1] J. Dong, E. Nelson, V. Indelman, N. Michael, F. Dellaert, Distributed Real-time Cooperative Localization and Mapping using an Uncertainty-Aware Expectation Maximization Approach, in ICRA 2015. [PDF]

[2] V. Indelman, E. Nelson, J. Dong, N. Michael, F. Dellaert, Incremental Distributed Inference from Arbitrary Poses and Unknown Data Association: Using Collaborating Robots to Establish a Common Reference, in IEEE Control Systems Magazine, 2016. [PDF]


Other Interesting Stuff


Tutorial of GTSAM

GTSAM is a library of C++/Matlab/Python classes that implement smoothing and mapping (SAM) in robotics and vision. I participate development and maintenance of GTSAM, and use GTSAM in most of my projects. A short tutorial slides and example code for GTSAM version 4.0 is available.

Slides and example code are available, webinar lecture video is available in Mandarin Chinese.

Tutorial of Gaussian Processes in Robotics

Continuous-time trajectory representations are powerful tools to overcome many difficulties in discrete-time SLAM, like motion distortion, asynchronous data, etc. In this tutorial, we discuss Gaussian processes (GPs) as continuous-time trajectory representations in SLAM and motion planning, and give several applications of GPs in robotics.

Slides and example code are available.

Research on Graphene 2D Nanomaterials

Before I start my robotics journey, I spent a fantastic year (2011-2012) at Nanoscale Physics and Devices Lab, Institute of Physics, Chinese Academy of Sciences, with Prof. Hongjun Gao, studying properties of 2D nanomaterials graphene, and exploring its application in surface material, field-emission display and catalyst. This is a really fun experience, let me have a dream to build a nano-robot in the (very far distant) future!

Although I published a few papers on this, I generally don't list them in my CV to let people (in robotics area) confused.

[1] J. Dong et al. Control of superhydrophilic and superhydrophobic graphene interface, Scientific reports, Vol. 3, 2013. [PDF]

[2] S. Yang et al. One-pot synthesis of graphene-supported monodisperse Pd nanoparticles as catalyst for formic acid electro-oxidation, Scientific reports, Vol. 4, 2014. [PDF]

[3] L. Jiang et al. Controlled synthesis of large‐scale, uniform, vertically standing graphene for high‐performance field emitters, Advanced Materials, Vol. 25.2, p. 250-255, 2013. [PDF]