ROBOT MOTION PLANNING & CONTROL

Jiarun Dong

董家润

Focused on robotics algorithm design / robot system development

M.S. student in Mechanical Engineering at Johns Hopkins University, focused on motion planning, optimal control, and safe autonomy in unstructured environments.

Seeking robotics algorithm roles for 2027 recruitingJdong38@jh.edu
Education
Johns Hopkins University · Mechanical Engineering, M.S.
Research focus
Robot Locomotion Planning & Control

SELECTED WORK

Projects

From problem formulation and algorithm design to ROS 2 integration and simulation, each case study documents reproducible engineering decisions and validation results.

CAD model of the TerraGoat 15-DOF active-spine quadrupedCore research project
TerraGoat physical prototype on the laboratory incline

CORE RESEARCH · QUADRUPED ROBOTICS

15-DOFActive SpineROS 2URDFRL

TerraGoat: 15-DOF Quadruped with an Active Spine

A goat-inspired quadruped platform for steep, contact-rich terrain, using an active spine, iterative hoof design, and a ROS 2 control architecture to study coordinated spine–limb climbing.

The mechanical prototype, wiring/hardware integration, ROS 2 modeling foundation, and open-loop tests are in place; learning-based climbing remains future work.

15 DOF

Actuated system

ROS 2

Modular control base

Open-loop

Current validation

PPO

Planned controller

View full case study
NMPC-CBF corridor test in Gazebo
Latest project

2026 · SAFE MOTION CONTROL

ROS 2C++NMPCCBFCasADi

ROS 2 NMPC-CBF Controller for Moving-Obstacle Avoidance

A ROS 2 local controller for dynamic simulated environments, combining predictive NMPC with slack-relaxed discrete control barrier function constraints for online tracking and moving-obstacle avoidance.

Independently developed the controller model, CasADi/IPOPT solver, ROS 2 integration, and Gazebo stress tests.

20 Hz

Closed-loop control

< 25 ms*

Reported solve

0

Test collisions

100%

Solver success

* Performance and success figures are presentation-reported; full-system logs are forthcoming.

View full case study
Bidirectional EST planning for a UR5e inside an automotive modelCore algorithm project
UR5e wall narrow-passage planning scene

MOTION PLANNING · 6-DOF MANIPULATOR

MoveIt 2C++Bi-ESTRRT-ConnectUR5e

EST-Inspired Bidirectional Planner with RRT-Connect

A custom planner for 6-DOF UR5e configuration-space search in narrow passages such as an automotive interior, combining EST-inspired degree-weighted local sampling, Gaussian mirror sampling, and greedy RRT-Connect expansion across two trees.

Implemented the MoveIt 2 PlanningContext plugin, sampling and tree expansion, continuous joint-space collision checking, and bidirectional path reconstruction.

2 Trees

Alternating expansion

0.04 rad

Collision resolution

12k

Iteration cap

80%+*

Reported success

* Success rate is presentation-reported; the source archive contains no batch logs.

View full case study

More projects will be added chronologically