Path-Finder-Drone

What does this do?

A Unity3d simulation of a drone model using a ML Agents to map a given terrian and identify possible paths to any two given points. This project makes use of ML-Agents to trian AI models using reinforced learning algorithms.

Here is a working demo video of this project

---

Software prerequisites:

• Unity Hub (Download it with this link)
• Unity3d 2019.3 or above. (currently using 2019.4.10f)
• Visual Studio installed (normally comes on installing Unity)
• All the required packages are installed just by opening with opening the project with appropriate unity version.

---

How to run the game:

• Clone this repository
• Open the cloned repository with UnityHub by clicking add in unity hub and opening the clone repository folder.
• Open the Assets -> Autonomous Drone -> Scenes then open the Training for Movement then open the Training for Movement scene by double clicking on it.
• play/Edit the game if you want and to run the click on play button on middle top of the unity editor.
• To build the cloud point information from the lidar sensor just turn on the saveToFile boolean in EnvironmentTraining gameobject in the scene hierarchy and on playing scene in editor should create the cloud point file just after the game directory inside the cloudpoints\cloudpoints.txt.
• To visualize the cloud data stored in the cloudpoints\cloudpoints.txt just open the Assets -> Autonomous Drone -> Scenes then open the Training for Movement then open the ExternalVisualization scene then play the scene in editor.
• That should show the cloudpoints in game view and you can roam around the scene using w\a\s\d keys and left mouse button to free lock the scene.
• To build the program simple go to file -> build settings -> build and run (make sure the build is on pc-standalone or webGL (note webgl is not supported by ML-Agents))
• Select the folder to build the program and wait for the build process to complete.
• Now play the game in your local computer.