ZJ_Zheng Jiatong · 郑佳桐
Portrait of Zheng Jiatong
Zheng Jiatong 郑佳桐

Robotics & Embodied Intelligence Engineer

+ ROBOTICS / EMBEDDED SYSTEMS / EMBODIED AI

Engineering
intelligence.
Into motion.

I build robotic systems from embedded control to embodied intelligence. My recent work focuses on UMI trajectory validation, vision-language navigation, and 3D mapping on Jetson AGX Thor.

Explore my work GitHub
01 / ROBOT CONTROL STACKARCHITECTURE
From robot software to physical motionA conceptual control architecture. ROS on Jetson sends motion commands to an STM32 controller. The controller drives actuators, while sensor feedback closes the control loop. ROBOTICS SOFTWAREEMBEDDED CONTROLPHYSICAL SYSTEMROS / JetsonSTM32 / PIDActuators motion commandsmotor drivesensor feedback COMPUTECONTROLMOTIONCLOSED-LOOP SYSTEM↺
From high-level intent to low-level control.SYSTEM VIEW
HARDWARE × SOFTWARE × INTELLIGENCESCROLL TO EXPLORE ↓
C / C++ROS / ROS 2STM32 & FPGAMotion controlPyTorch & LoRA

01 / SELECTED ENGINEERING

Built for the physical world.

From embedded control loops
to intelligent robotic systems.

UMI / 2026INITIAL VALIDATION

UMI trajectory measurement

Building a stereo–inertial tracking pipeline and validating UMI motion against robot-arm reference trajectories.

UMIORB-SLAM3ROS 2Hand–eye calibration
Engineering details

Integrated stereo cameras and IMU data, worked on calibration and visual–inertial tracking, and completed initial trajectory experiments with robot-held UMI. Recent work uses the Elephant Pro630 to repeat controlled motions and compare recorded trajectories.

Current focus: improve synchronization, coordinate alignment, and repeatability. Initial experiments are complete; accuracy is still being improved.

VLN / 2026INTEGRATION & TESTING

Language-guided robot navigation

Bringing vision-language navigation to a mobile robot powered by Jetson AGX Thor and the AgileX Ranger mini 3.0.

Jetson AGX ThorInternVLAROS 2 JazzyCAN
Engineering details

Integrated the Thor compute platform with the mobile chassis through CAN and ROS 2, including odometry and robot-state topics. Work also covers perception integration and InternVLA model runtime testing.

Current focus: connect language and visual observations to repeatable navigation experiments, with planning and control integration.

LINGBOT-MAP / 2026DEPLOYMENT IN PROGRESS

3D mapping for navigation

Deploying LingBot-Map on Jetson AGX Thor and exploring streaming 3D reconstruction for robotic scene understanding.

LingBot-Map3D reconstructionJetson AGX Thor
Engineering details

Working on environment setup and model deployment. The next evaluation steps are official examples, UMI image sequences, and measurements of speed, memory use, and long-sequence stability on Thor.

Exploration: assess how reconstructed geometry can support mapping and navigation. Navigation integration remains ongoing work.

LingBot-Map · upstream project ↗
PROJECT / 01ROBOTICS

Robot motion.
From board to behavior.

A robot control stack connecting Jetson compute with STM32 controllers, sensor acquisition, and actuator control.

JetsonSTM32Motion
Embedded C/C++ROSMotor control
Engineering details

Developing low-level control and communication between the upper computer and embedded controllers. Work spans sensor interfaces, actuator integration, circuit debugging, and functional verification.

Shanghai MOSI Intelligent Technology · since Mar 2026

PROJECT / 02AUTONOMOUS SYSTEMS

Intelligent cars.
Real-time decisions.

An MCU–FPGA architecture for image processing, encoder acquisition, and closed-loop control on a competition vehicle.

2×National first prize
Intelligent Car Competition · 2021 & 2022
Infineon MCUFPGA / VerilogPID
Engineering details

Integrated real-time sensing and motor control, tuned high-speed tracking and steering, and developed multilayer PCBs across power, drive, communication, and sensor modules.

2021 · National 10th place
2022 · National 6th place

PROJECT / 03ROBOTICS SOFTWARE

Mobile robot navigation

A ROS-based mobile robot platform integrating path planning, chassis control, and sensor visualization.

ROSPath planningSensor integration
PROJECT / 04SENSING

Non-contact 3D measurement

An FPGA-based measurement project connecting digital hardware design with physical-world sensing.

FPGAVerilog3D measurement

THE WAY I WORK

Understand the system.
Build across the stack.
Validate on real hardware.

My work sits at the intersection of robotics, embedded engineering, and embodied intelligence. I work across kinematics, control algorithms, sensor integration, and software–hardware debugging.

From schematics and firmware to robot software and model fine-tuning, I enjoy turning an idea into a system that works beyond the screen.

Illustrated portrait of Zheng Jiatong
Zheng Jiatong / 郑佳桐

Robotics, with a human side.

02 / EXPERIENCE

Across hardware and intelligence.

2021 — 2023

Harbin Engineering University

Robotics Embedded & Control Engineer · Autonomous Vehicle Intelligent Car Platform

Undergraduate mentor: Dexin Xu

MCU–FPGA control, real-time sensing, PCB design, and competitive autonomous driving.

04

03 / RESEARCH

Intelligence with an engineering purpose.

Language models
for industrial control.

Building and evaluating code generation systems with IC-100, semantic alignment, and LoRA fine-tuning.

PLC / Structured TextCodeBLEUpass@kLoRA
02 / INFORMATION AND CONTROL · FIRST AUTHOR

A Code Generation Method for Industrial Control Tasks Based on Alignment-Guided Fine-Tuning

Fine-tuning · industrial automation

04 / EDUCATION

Engineering foundations.

TECHNICAL TOOLKIT

Control & robotics
Kinematics, PID, motion control, sensor fusion, ROS / ROS 2
Embedded & hardware
C/C++, STM32, Infineon TC26x/TC36x, FPGA, Verilog, Altium
Intelligence & tools
Python, PyTorch, LoRA, MATLAB, Linux, Git

05 / GET IN TOUCH

Let's build
something that moves.

GitHub ↗Website repository ↗