https://github.com/zsdonghao/Imitation-Learning-Dagger-Torcs

A Simple Example for Imitation Learning with Dataset Aggregation (DAGGER) on Torcs Env
https://github.com/zsdonghao/Imitation-Learning-Dagger-Torcs

dagger gym-torcs imitation-learning tensorflow tensorflow-tutorials tensorlayer torcs torcs-env

Last synced: about 1 month ago
JSON representation

A Simple Example for Imitation Learning with Dataset Aggregation (DAGGER) on Torcs Env

• Host: GitHub
• URL: https://github.com/zsdonghao/Imitation-Learning-Dagger-Torcs
• Owner: zsdonghao
• Created: 2017-08-10T17:33:43.000Z (over 7 years ago)
• Default Branch: master
• Last Pushed: 2017-08-10T19:42:20.000Z (over 7 years ago)
• Last Synced: 2024-10-11T08:41:41.507Z (2 months ago)
• Topics: dagger, gym-torcs, imitation-learning, tensorflow, tensorflow-tutorials, tensorlayer, torcs, torcs-env
• Language: Python
• Homepage: https://github.com/zsdonghao/tensorlayer
• Size: 13.7 KB
• Stars: 71
• Watchers: 4
• Forks: 28
• Open Issues: 2
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

• awesome-tensorflow - DAGGER - For Playing [Gym Torcs](https://github.com/ugo-nama-kun/gym_torcs) (Models/Projects)
• awesome-tensorflow - DAGGER - For Playing [Gym Torcs](https://github.com/ugo-nama-kun/gym_torcs) (Models/Projects)
• fucking-awesome-tensorflow - DAGGER - For Playing <b><code>&nbsp;&nbsp;&nbsp;407⭐</code></b> <b><code>&nbsp;&nbsp;&nbsp;163🍴</code></b> [Gym Torcs](https://github.com/ugo-nama-kun/gym_torcs)) (Models/Projects)

README

        
## Imitation Learning with Dataset Aggregation (DAGGER) on Torcs Env

This repository implements a simple algorithm for imitation learning: [DAGGER](https://www.cs.cmu.edu/~sross1/publications/Ross-AIStats11-NoRegret.pdf).

In this example, the agent only learns to control the steer [-1, 1], the speed is computed

automatically in `gym_torcs.TorcsEnv`.

## Requirements

1. Ubuntu (I only test on this)

2. Python 3

3. TensorLayer and TensorFlow 

4. [Gym-Torcs](https://github.com/ugo-nama-kun/gym_torcs)

## Setting Up

It is a little bit boring to set up the environment, but any incorrect configurations will lead to FAILURE.

After installing [Gym-Torcs](https://github.com/ugo-nama-kun/gym_torcs), please follow the instructions to confirm everything work well:

- Open a terminal:

  - Run `sudo torcs -vision` to start a game

  - `Race --> Practice --> Configure Race`: set the driver to `scr_server 1` instead of `player`

  - Open Torcs server by selecting `Race --> Practice --> New Race`:

This should result that Torcs keeps a blue screen with several text information.

- Open another terminal:

  - Run `python snakeoil3_gym.py` on another terminal, it will shows how the fake AI control the car.

  - Press F2 to see the driver view.

- Set image size to 64x64x3:

  - The model is trained on 64x64 RGB observation.

  - Run `sudo torcs -vision` to start a game

  - `Options --> Display --> select 64x64 --> Apply`

## Usage

Make sure everything above work well and then run:

- `python dagger.py`

It will start a Torcs server at the beginning of every episode, and terminate the server when the car crashs or the speed is too low.

Note that, the self-contained `gym_torcs.py` is modified from [Gym-Torcs](https://github.com/ugo-nama-kun/gym_torcs), you can try different settings (like default speed, terminated speed) by modifying it.

## Results 

After Episode 1, the car crashes after 315 steps. 

![](http://i.imgur.com/YfqFXQZ.gif)

After Episode 3, the car does not crash anymore !!!

![](http://i.imgur.com/doz8U0z.gif)

The number of steps and episodes might vary depending on the parameters initialization.

ENJOY !