Skip to content

Latest commit

 

History

History
29 lines (22 loc) · 1.21 KB

README.md

File metadata and controls

29 lines (22 loc) · 1.21 KB
Raw

Obstacle avoidance agent

This project is developed in a custom Gym environment and with Pygame to help to visualize the agent and the environment. The goal of this agent is to avoid some n obstacles with helping himself with sensors all around the agent, the obstacles positions are random generated also as the agent position.

How the obstacles are random generated

The obstacles random positions has some criteria to respect:

  1. Does not have to spawn on a position already taken
  2. Each obstacles needs to be at least 200 pixels from eachother
    • calculated from the center of each obstacles
  3. Does not have to collide with any other obstacles

Prerequisites

To install the requirements for the project run

pip install -r requirements.txt

Training

In train.py file, set the HYPER PARAMETERS as your choice then train the agent by running

python train.py

File of the model is saved in this location trained_agent/models/PPO_MODELS

Contributing

Feel free to contribute on get the code works better or some new features to implement. Any contributions you make are greatly appreaciated.