RPKG(ROS Package Knowledge Graph)

implementation of the paper Recommending ROS Packages for Robot Task by Knowledge Graph presented in arxiv.

Method Overview

High-level Abstraction of ROS Packages

RPKG Construction

RPKG-based ROS Package Search

Evaluation

Installation

Here are some instructions to help you get RPKG deployed in your machine. These instructions assume that you already have a working installation of Neo4j. RPKG has been tested with Linux Ubuntu 16.04, Linux Ubuntu 18.04 and MacOS 13.3. These setups should provide you with most of the basic dependencies of RPKG, namely Python 3.7+ and Neo4j 4.4+.

1. Install Neo4j.
2. Download the codes into local machine:

git clone https://github.com/iwangshuo/RPKG.git

3. Install other dependencies:

cd RPKG/
pip install -r requirements.txt

Tutorials

1. Extracting ROS Package Information

python package_information/crawl_robot.py

this step will generate csv files into data/ folder.

2. Extracting Multi-dimensional Features

extract features on code files, function and characteristics.

python feature_extraction/feature_extractor.py

extract features on robot and sensor.

python feature_extraction/hardware_feature_extractor.py

this step will generate csv files into data/output/ folder.

3. Representing ROS Package Knowledge

further train Bert. In practice, we choose bert-base-uncased model for further training.

python knowledge_representation/further_train_bert.py

4. Creating Database and Importing Data

create a new database in Neo4j and import data from data/folder.

python neo4j/run_neo4j_import_data.py

5. RPKG-based ROS Package Search

search ROS packages with user queries

python package_retrieval/run_search.py

6. Evaluation

6.1 ChatGPT

prompt ChatGPT to recommend ROS packages.

cd evaluation/chatgpt
python ./RunChatGPT_evaluation_search.py --code_dir ../user_queries_chatgpt.csv --log_dir ../log/test_package_search_0_10.log --output_dir search_result --start 0 --end 1

analyze the accuracy of recommended ROS packages and time cost.

python analyze_results_chatgpt.py
python chatgpt_log_parser.py

6.2 GitHub

use Github to search ROS packages manually.

6.3 Google

use Google to search ROS packages manually.

6.4 ROS Index

use ROS Index to search ROS packages semi-automatically. For quickly acquire and analyze the search results, we deploy ROS Index in our local machine with GitHub repository index.ros.org

cd evaluation/rosindex
python analyze_results_rosindex.py

6.5 RPKG-based Search Method

• analyze search accuracy

cd evaluation/rpkg
python NDCG.py --rank_input_dir ./all_results.csv --acc_output_dir ./accuracy/acc.csv

• ablation experiments

python package_retrieval/run_search.py
cd evaluation/rpkg
python NDCG.py --rank_input_dir ./ablation_experiments/data/xxx_feature_necessity.csv --acc_output_dir ./ablation_experiments/data/xxx_acc.csv

• solidity validation

cd evaluation/rpkg/solidity_validation
python ./solid_validation.py --rank_input_dir ../all_results.csv --acc_output_dir ./solid_result.csv

Citation

If you find our work useful in your research, please consider citing our paper:

@misc{wang2023ros,
      title={ROS package search for robot software development: a knowledge graph-based approach}, 
      author={Shuo Wang and Xinjun Mao and Shuo Yang and Menghan Wu and Zhang Zhang},
      year={2023},
      eprint={2312.14781},
      archivePrefix={arXiv},
      primaryClass={cs.SE}
}

Questions

If you have any questions, please contact "iwangshuo@qq.com".