Path Planning and Obstacle Avoidance for Kuka Robotic Arm using Artificial Potential Field

Introduction

This project implements obstacle avoidance for a Kuka robotic arm using the Artificial Potential Field (APF) method. The simulation is conducted in PyBullet, and point cloud data is processed using Open3D.

Features

• Point Cloud Processing: Generates and processes 3D point clouds of the environment.
• Artificial Potential Field (APF): Computes attractive and repulsive forces for navigation.
• Path Planning: Uses gradient descent for trajectory generation.
• Kuka Arm Simulation: Executes motion in PyBullet while avoiding obstacles.
• Visualization: Uses Open3D for point cloud and trajectory visualization.

---

Project Structure

• getPLY.py: Captures depth images in PyBullet and converts them into a PLY point cloud.
• plySEE.py: Filters and visualizes the generated point cloud.
• apf2.py: Implements APF-based path planning, including:
  • Point cloud loading and processing
  • Attractive and repulsive potential field computation
  • Path generation using gradient descent
  • 3D visualization of the environment and trajectory
• APF.py: Integrates APF planning with the Kuka robot simulation, executing path following in PyBullet.

---

Dependencies

Ensure the following Python packages are installed:

pip install numpy open3d matplotlib pybullet opencv-python

---

Usage

1. Generate Point Cloud Data

Run getPLY.py to capture depth images and convert them into a PLY point cloud:

python getPLY.py

This will create a point_cloud.ply file in the image/ directory.

2. Filter and Visualize the Point Cloud

To remove noisy points and visualize the processed cloud:

python plySEE.py

This generates filtered_point_cloud.ply.

3. Compute APF-Based Path

Run apf2.py to generate an APF trajectory and visualize the result:

python apf2.py

4. Simulate in PyBullet

Finally, execute APF.py to move the Kuka robotic arm while avoiding obstacles:

python APF.py

This runs the simulation and controls the robotic arm.

---

How It Works

1. Point Cloud Processing:
   • PyBullet generates depth images and converts them into a PLY point cloud.
   • Open3D filters noise and downsamples the cloud.
2. APF-Based Path Planning:
   • Computes attractive force (pulling the robot towards the goal).
   • Computes repulsive force (pushing the robot away from obstacles).
   • Uses gradient descent to iteratively generate a safe path.
3. Kuka Arm Motion Execution:
   • Uses inverse kinematics (IK) to convert path points into joint angles.
   • Controls the robotic arm using PyBullet.

---

Future Improvements

• Dynamic obstacle avoidance using real-time updates.
• Reinforcement Learning (RL) integration for adaptive planning.
• Multi-robot coordination to handle complex scenarios.

---

License

This project is licensed under the MIT License.