Skip to content

[ICRA'23] Learning Continuous Control Policies for Information-Theoretic Active Perception

Notifications You must be signed in to change notification settings

JaySparrow/RL-for-active-mapping

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

64 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Landmark-based mapping

Thank you for being interested in our work! This repository is a PyTorch implementation for paper Learning Continuous Control Policies for Information-Theoretic Active Perception in ICRA 2023. Authors: Pengzhi Yang, Yuhan Liu, Shumon Koga, Arash Asgharivaskasi, Nikolay Atanasov. If you are using the code for research work, please cite:

@inproceedings{yang2023icra,
  title={Learning Continuous Control Policies for Information-Theoretic Active Perception},
  author={Yang, Pengzhi and Liu, Yuhan and Koga, Shumon and Asgharivaskasi, Arash and Atanasov, Nikolay},
  booktitle={IEEE international conference on robotics and automation (ICRA)},
  year={2023}
}

Brief Intro

This paper proposes a method for learning continuous control policies for active landmark localization and exploration using an information-theoretic cost. We consider a mobile robot detecting landmarks within a limited sensing range, and tackle the problem of learning a control policy that maximizes the mutual information between the landmark states and the sensor observations. Here is a pipeline for our work:

Instruction

  1. First clone this repo and install the dependencies running the following commands in the terminal:
git clone https://github.com/JaySparrow/RL-for-active-mapping.git
git clone https://github.com/pengzhi1998/Landmark-based-mapping-Unity.git
conda create -n landmark_mapping python==3.7 -y
conda activate landmark_mapping
cd ./RL-for-active-mapping
pip install -r requirements.txt
cd ../
git clone --branch release_18 https://github.com/Unity-Technologies/ml-agents.git
cd ./ml-agents
pip install -e ./ml-agents-envs
pip install gym-unity==0.27.0
  1. cd to the ./RL-for-active-mapping/toy_active_mapping_ppo directory, then train and test the model:
python ./agent.py 
python ./agent_test.py  

There are multiple arguments for you to tune with for different scenarios. You could also test the model in Unity environment:

python ./agent_unity_test.py 
  1. To play with Unity env, build it with the repo you have cloned:
    (1) Create a directory ./toy_active_mappking_ppo/Unity_envs in your RL-for-active-mapping repo.
    (2) Build the environment with the cloned Landmark-based-mapping-Unity repo to Unity_envs just created.

About

[ICRA'23] Learning Continuous Control Policies for Information-Theoretic Active Perception

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages