# ICM-PPO-implementation

**Repository Path**: chenglijie1015/ICM-PPO-implementation

## Basic Information

- **Project Name**: ICM-PPO-implementation
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: MIT
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2024-03-22
- **Last Updated**: 2024-03-23

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# ICM-PPO-implementation

Experiment with ICM and PPO bunch for environment with sparse reward signal.

## Description

The experiment tests the contribution of [intrinsic reward](https://blogs.unity3d.com/ru/2018/06/26/solving-sparse-reward-tasks-with-curiosity/) to the agent's ability to solve the [sparse-reward environment](https://github.com/Unity-Technologies/ml-agents/blob/master/docs/Learning-Environment-Examples.md#pyramids) from [Unity ML-Agents Toolkit](https://github.com/Unity-Technologies/ml-agents/).

### Results

#### Tensorboard logs for extrinsic and intrinsic rewards
![Mean extrinsic reward](media/Extr_reward.PNG)
![Mean intrinsic reward](media/Intr_reward.PNG)

#### Running examples
![1](media/1_l.gif)

![2](media/2_l.gif)

![3](media/3_l.gif)

## Built With

* [Unity ML-Agents Toolkit](https://github.com/Unity-Technologies/ml-agents/).