Sensing-target-Obstacle-detection-and-Obstacle-Avoidance-in-a-3D-shoot-and-target-game

KO | EN

gitlite — search

#hacktoberfest #react #llm #machine-learning #rust #linux #nodejs #windows #awesome #cli #kubernetes #vue #security #open-source

Sensing-target-Obstacle-detection-and-Obstacle-Avoidance-in-a-3D-shoot-and-target-game

(★ 9)

The goal is to train an agent that can detect and avoid obstacles inside a polytope. The agent will learn the line of the shot of the ball to reach the goal in optimal steps. Utilizing Tensorflow with Reinforcement learning in Unity along with mlagents package to train the game.

Sensing-target-Obstacle-detection-and-Obstacle-Avoidance-in-a-3D-shoot-and-target-game Latest Version Download

Download Latest Version (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# THESIS: Trajectory Prediction Through Deep Reinforcement Learning in Unity

"The thesis aimed to develop a 3D game. The case of the project was a trajectory prediction shooting game in which the player had to search for the goal(star) inside a polytope and shoot the ball to the goal. There were a few obstacles inside the polytope, and they spawned randomly. The player must continuously hit the goal as many times as possible. If the player misses the goal, then the game is over. This thesis investigates how artificial intelligence techniques can be used to predict the trajectory of the ball."

The game contains two main game scenes: one is the Graphic User Interface (GUI) scene, and the other is the game scene. The first part describes the practicalities of the game. The player could play, control the physics of the ball, get help with the Prediction line, and start or end this game on the GUI scene. 
This thesis presents, to our knowledge, the second part of introducing the vast field of artificial intelligence. Topics introduced include Reinforcement Learning techniques, agent simulation, and Imitation Learning.

We believe our approach demonstrates an artificial agent that can learn a conditioned trajectory to model human motion inside a polytope using imitation learning.

#### 1: Trained an agent to learn a conditioned trajectory inside a polytope to reach the goal using the nature_cnn visual encoder

#### 2: Developed a GUI of the game playable by humans and the AI using Generative Adversarial Imitation Learning

Sensing-target-Obstacle-detection-and-Obstacle-Avoidance-in-a-3D-shoot-and-target-game

Sensing-target-Obstacle-detection-and-Obstacle-Avoidance-in-a-3D-shoot-and-target-game Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

Sensing-target-Obstacle-detection-and-Obstacle-Avoidance-in-a-3D-shoot-and-target-game

Sensing-target-Obstacle-detection-and-Obstacle-Avoidance-in-a-3D-shoot-and-target-game Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

Download README