# Goal-Driven Deep RL Policy for Robot Navigation
Deep Reinforcement Learning for mobile robot navigation in ROS2 Gazebo simulator. Using Twin Delayed Deep Deterministic Policy Gradient (TD3) neural network, a robot learns to navigate to a random goal point in a simulated environment while avoiding obstacles. Trained in ROS2 Humble & Gazebo simulator with PyTorch.

![Screenshot from 2023-12-09 11-41-59](https://github.com/vishweshvhavle/deep-rl-navigation/assets/52754207/704922f1-d5c7-4665-9c31-e15169521624)

![sim2](https://github.com/vishweshvhavle/deep-rl-navigation/assets/52754207/1d9b8f3d-6d1b-4b01-a573-e380ced20679)

![sim1](https://github.com/vishweshvhavle/deep-rl-navigation/assets/52754207/334d1542-58d8-4cd4-8bb5-d6611ea33d9d)

# How To Run

Install Python 3.10, ROS2 Humble, Gazebo 11 on Ubuntu 22.04
```
git clone git@github.com:vishweshvhavle/deep-rl-navigation.git
cd deep-rl-navigation/DRL_robot_navigation_ros2/
sudo rosdep init
rosdep install --from-paths src --ignore-src -y
colcon build
cd ..
```

For Training - 
```
ros2 launch td3 train_simulation.launch.py
```

For Testing - 
```
ros2 launch td3 test_simulation.launch.py
```

# Academic Integrity
If you are currently enrolled in this course, please refer to IIIT-Delhi's Policy on Academic Integrity before referring to any of the repository contents. This repository contains the work we did as undergrads at IIIT-Delhi in CSE-564 Reinforcement Learning course. We do not encourage plagiarism of any kind.