TE3001B Control — xArm Lite 6 Robotic Control Labs

ROS 2 Humble workspace for the TE3001B Fundamentacion de Robotica laboratory series. This repository contains two independent ROS 2 packages that implement and compare different control strategies on the UFACTORY xArm Lite 6 six-degree-of-freedom manipulator, using MoveIt Servo as the hardware interface.

Course: TE3001B.101 — Fundamentacion de Robotica (Gpo 101) Institution: Tecnologico de Monterrey Team: Team Funcionaba Ayer

• David Alejandro Soni Cuevas (A01571777)
• Abraham de Jesus Maldonado Mata (A00838581)
• David Gilberto Lomeli Leal (A01571193)
• Irvin David Ornelas Garcia (A00839065)

Professor: Nezih Nieto Gutierrez

---

Table of Contents

1. Repository Overview
2. Hardware
3. Prerequisites and Dependencies
4. Repository Structure
5. Build Instructions
6. Packages
   • xarm_ctc_challenge — Challenge 4.1
   • xarm_perturbations — Lab 4.2
7. Common Launch Sequence
8. Data and Results
9. Documentation
10. Known Issues and Workarounds

---

1. Repository Overview

This repository implements and evaluates two control paradigms on the xArm Lite 6:

Package	Lab / Challenge	Control Strategy	Task
xarm_ctc_challenge	Challenge 4.1	CTC (model-based) vs Cartesian PD (model-free)	PCB component placement — 2x2 grid, quintic spline trajectory
xarm_perturbations	Lab 4.2	Cartesian PD/PID	Lissajous / inspection trajectory under baseline, sinusoidal, and Gaussian perturbations

Both packages interface with the robot through MoveIt Servo's velocity command interface, publishing TwistStamped or JointJog messages at 100 Hz.

---

2. Hardware

• Robot: UFACTORY xArm Lite 6 (6-DOF collaborative manipulator)
• Connection: Ethernet, static IP 192.168.1.154
• Control PC: Ubuntu 22.04, ROS 2 Humble
• Robot mode before running: Mode 1, State 0 (velocity servo mode)

Verify connectivity before any experiment:

ping 192.168.1.154

---

3. Prerequisites and Dependencies

System:

• Ubuntu 22.04 LTS
• ROS 2 Humble Hawksbill (full desktop install)
• Python 3.10+

Python packages (inside your environment):

pip install numpy pandas matplotlib scipy

ROS 2 packages (sourced from workspace):

• moveit_ros_planning_interface
• moveit_servo
• tf2_ros, tf2_geometry_msgs
• control_msgs, sensor_msgs, geometry_msgs
• xarm_ros2 (included as a git submodule at ros2_ws/src/xarm_ros2/)

---

4. Repository Structure

TE3001B_Control/
|-- README.md                        This file
|-- LICENSE
|-- .gitmodules                      Submodule: xarm_ros2
|
|-- docs/
|   |-- Robotics_Control_Challenge_4.1/
|   |   |-- Instructions/
|   |       |-- Robotics Control Challenge 4.1-1.pdf
|   |       |-- CTCvsPID.ipynb
|   |       `-- TE300XB-5.pdf
|   |-- Robotics_Control_Lab_4.1/
|   |   `-- code_documentation/      Auto-generated per-module docs
|   `-- Robotics_Control_Lab_4.2/
|       |-- code_documentation/      Auto-generated per-module docs
|       `-- Report/
|           `-- Robotics Control Lab 4.2.pdf
|
|-- data/                            Trial results (CSV, PNG, JSON)
|   |-- trial_ctc_nopert_*.csv
|   |-- trial_ctc_pert_*.csv
|   |-- trial_pdpid_nopert_*.csv
|   |-- trial_pdpid_pert_*.csv
|   |-- *_metadata.json
|   |-- joints_*.png
|   |-- taskspace_*.png
|   |-- phase_*.png
|   |-- comparison_*.png
|   |-- summary_table.png
|   `-- metrics_summary.csv
|
`-- ros2_ws/                         ROS 2 Humble colcon workspace
    |-- README.md                    Workspace build notes
    `-- src/
        |-- xarm_ctc_challenge/      Challenge 4.1 — see package README
        |-- xarm_perturbations/      Lab 4.2 — see package README
        |-- xarm_ros2/               xArm driver (git submodule)
        `-- gazebo_mujoco_bridge/    Simulation bridge (skip in build)

---

5. Build Instructions

Initialize submodule

If cloning for the first time, initialize the xarm_ros2 submodule:

git submodule update --init --recursive

Build all non-simulation packages

cd ros2_ws
colcon build --packages-skip gazebo_mujoco_bridge mbot_demo

Conda environment workaround

If the build fails with ModuleNotFoundError: No module named 'catkin_pkg', conda is intercepting CMake's Python lookup. Fix by installing catkin_pkg in the active conda environment:

pip install catkin_pkg

Alternatively, deactivate conda before building and reactivate afterwards:

conda deactivate
colcon build --packages-skip gazebo_mujoco_bridge mbot_demo
conda activate <your_env>

Source the workspace

Run this in every terminal that needs to use the ROS 2 packages:

source ros2_ws/install/setup.bash

---

6. Packages

xarm_ctc_challenge — Challenge 4.1

Full documentation: ros2_ws/src/xarm_ctc_challenge/README.md

Implements and compares two joint-space controllers for a PCB component-placement task:

• CTC (Computed Torque Control): Model-based feedforward (M, C, G, F) with PD error correction and a robust sliding-mode term. Nominal link dynamics are derived from xArm Lite 6 URDF parameters.
• Cartesian PD: Model-free proportional-derivative control on TF2 end-effector position error. Simple, stable, and used as the reference baseline.

Both controllers use an online weighted resolved-rate IK solver (DLS pseudoinverse with null-space posture control) to convert Cartesian waypoints to joint references.

Four required trials:

1. CTC, no perturbation
2. CTC, Gaussian perturbation (sigma = 0.5 m/s on x-axis)
3. PD, no perturbation
4. PD, Gaussian perturbation

xarm_perturbations — Lab 4.2

Full documentation: ros2_ws/src/xarm_perturbations/README.md

Implements Cartesian PD/PID control on Lissajous and inspection trajectories. Evaluates controller robustness under three perturbation conditions:

• Baseline (no perturbation)
• Sinusoidal (8 Hz deterministic disturbance)
• Gaussian white noise (sigma = 0.01 m/s)

Includes a perturbation injector node, a grid-search auto-tuner for Kp/Kd gains, and offline analysis scripts for RMSE computation and plot generation.

---

7. Common Launch Sequence

Every experiment requires MoveIt Servo to be running first. Open terminals in order:

Terminal 1 — MoveIt Servo (keep running for all experiments):

ros2 launch xarm_moveit_servo lite6_moveit_servo_realmove.launch.py \
  robot_ip:=192.168.1.154

Wait for the servo server to print Ready to accept commands. Then proceed with the package-specific launch commands documented in each package's README.

Emergency stop (any terminal):

ros2 topic pub /challenge_stop std_msgs/Bool "data: true" --once

---

8. Data and Results

Experimental data from Challenge 4.1 is stored in data/ and archived in data.zip.

Summary of results (Challenge 4.1):

Trial	RMSE EE	Max EE Error	Waypoint Success Rate
CTC, no perturbation	29.4 mm	—	—
CTC, with perturbation	~35.3 mm	—	—
PD, no perturbation	29.2 mm	—	—
PD, with perturbation	~38.9 mm	—	—

CTC shows approximately 20% error increase under perturbation versus approximately 33% for Cartesian PD, demonstrating the robustness advantage of the model-based approach.

Phase portrait analysis shows asymptotically stable spiral convergence in joint error space, with estimated damping ratio of approximately 0.89 for both controllers under nominal conditions.

---

9. Documentation

Additional documentation is available under docs/:

Path	Contents
docs/Robotics_Control_Challenge_4.1/Instructions/	Challenge specification PDF and reference notebook
docs/Robotics_Control_Lab_4.1/code_documentation/	Auto-generated per-module documentation for Lab 4.1
docs/Robotics_Control_Lab_4.2/code_documentation/	Auto-generated per-module documentation for Lab 4.2
docs/Robotics_Control_Lab_4.2/Report/	Final lab report PDF

---

10. Known Issues and Workarounds

catkin_pkg not found during colcon build: Conda environments intercept CMake's Python executable, causing catkin_pkg import failures. Fix: pip install catkin_pkg inside the active conda environment.

Joint state updates at ~10 Hz instead of 100 Hz: The xArm hardware driver publishes /joint_states at approximately 10 Hz. The CTC controller node uses a first-order state predictor to extrapolate joint positions between updates, maintaining smooth 100 Hz control output.

MoveIt Servo velocity scaling (~22% effective): MoveIt Servo internally publishes joint commands at 2 ms intervals, while the control loop runs at 10 ms. Effective Cartesian velocity is approximately 22% of the commanded value. IK task-space gains (K_TASK = 20) and velocity limits are set to compensate for this scaling.

Wrist singularity (joint 5 near 0 rad): The position Jacobian becomes near-singular when joint 5 approaches zero. The controller scales Cartesian velocity commands by clip(|q5| / 0.20, 0.05, 1.0) to prevent command blow-up, reaching minimum 5% velocity at the singular configuration.

Analysis script Python environment: The analysis script requires numpy/pandas/matplotlib. If the ROS 2 Python environment has version conflicts, run with miniconda Python directly:

PYTHONPATH=ros2_ws/src/xarm_ctc_challenge \
  ~/miniconda3/bin/python3 \
  ros2_ws/src/xarm_ctc_challenge/xarm_ctc_challenge/analysis.py \
  data/