🧠 EEG Controlled Robotic Arm

A robotics system that enables brain-signal based control of a robotic arm using EEG signals, machine learning, and computer vision.

The system processes EEG signals, predicts user intent, and executes robotic actions such as object detection, alignment, and grasping.

---

🚀 Project Overview

This project integrates EEG signal processing, machine learning, computer vision, and robotic control to create a brain-computer interface (BCI) for controlling a robotic arm.

Pipeline:

EEG Signals → Machine Learning Prediction → Robot Command → Object Detection → Object Grasping

---

⚙️ System Architecture

EEG Signals
↓
Signal Processing (Bandpass Filtering + Feature Extraction)
↓
Machine Learning Model (Random Forest + CSP)
↓
Command Prediction
↓
ROS Communication
↓
Robot Control (Doosan API)
↓
Intel RealSense Camera (Object Detection + Depth Estimation)
↓
Robot Alignment and Pick

---

🧰 Technologies Used

• Python
• ROS2
• Intel RealSense D435i
• Doosan Robot API (DSR_ROBOT)
• Scikit-learn
• OpenCV
• NumPy / SciPy
• Matplotlib / Seaborn

---

🧠 EEG Processing

Electrode positions used:

FP1 FP2
C3 C4
P3 P4
O1 O2

Processing pipeline:

1. Bandpass filtering
2. Feature extraction using Common Spatial Patterns (CSP)
3. Feature normalization
4. Classification using Random Forest

---

👁️ Computer Vision

The system uses Intel RealSense D435i for object detection and depth estimation.

Features:

• RGB + Depth sensing
• Real-world coordinate estimation
• Camera-to-robot coordinate transformation
• Real-time object tracking

---

🤖 Robot Control

Robot is controlled using the Doosan Robotics Python API.

Motion commands:

• movej() -- Joint space motion
• movel() -- Linear motion

Robot tasks:

• Align with detected object
• Descend to pick position
• Close gripper
• Lift object

---

📂 Project Structure

EEG-Robot-Control │ ├── eeg_processing/ ├── robot_control/ ├── vision/ ├── ros_nodes/ ├── models/ └── README.md

---

🔄 Robot Pick Pipeline

1. Detect cube using camera
2. Align robot with cube in X and Y
3. Descend using depth value
4. Close gripper
5. Lift object

---

📊 Machine Learning Model

Model: Random Forest Classifier

Feature extraction: Common Spatial Pattern (CSP)

Evaluation:

• Leave-One-Group-Out Cross Validation
• Confusion Matrix
• Classification Report

---

🛠️ Installation

git clone https://github.com/FourCoreProcessor/eeg-robot-control.git
cd eeg-robot-control
pip install -r requirements.txt

---

▶️ Running the Project

Start ROS:

ros2 launch robot_control robot_launch.py

Run EEG prediction, Cube Detection, Robot Control:

python eeg_robot.py

---

📸 Hardware Used

• OpenBCI Ultracortex EEG Headset for EEG aquisition
• Intel RealSense D435i depth camera
• Doosan A0509 robot
• DH-Robotics PGE100 gripper

---