A Lifelong Navigation Approach to Mobile Robot Navigation

This is the source code for this paper (RAL/ICRA 2021)

Enter the following to see the arguments.

python main.py --help

Make Data

In particular, type

python main.py --mode=make-data

to generate training data. Data files should be listed in ./data folder in the form task{i}-lidar.csv, task{i}-cmd.csv.

Learning

Type

python main.py --mode=lifelong-learn

to learn the model. The learned model will be saved by default to ./models/trained_agent.pt

Deployment

Type

python main.py --mode=predict-example

to see a particular example of loading the trained agent and predict the motion command from a lidar beam.

Citation

If you find our paper interesing or the repo useful, please consider cite this paper

@article{liu2021lifelong,
  title={A lifelong learning approach to mobile robot navigation},
  author={Liu, Bo and Xiao, Xuesu and Stone, Peter},
  journal={IEEE Robotics and Automation Letters},
  volume={6},
  number={2},
  pages={1090--1096},
  year={2021},
  publisher={IEEE}
}