Optimal Interactive Learning on the Job via Facility Location Planning

Welcome to our repository implementing COIL, as presented in:

Vats*, S., Zhao*, M., Callaghan, P., Jia, M., Likhachev, M., Kroemer, O., & Konidaris, G.D. (2025). Optimal Interactive Learning on the Job via Facility Location Planning. In Robotics: Science and Systems (RSS).

Website

Installation

We have tested our code on Ubuntu 20.04.

Requirements

• miniconda

Installation Steps

1. Clone this repository and change directory into it:

   git clone git@github.com:shivamvats/coil.git 
   cd coil

2. Create a conda environment and activate it:

   conda env create -f environment.yaml
   conda activate coil-env

3. Install the local packages robosuite and mimicgen.

   pip install -e deps/robosuite
   pip install -e deps/mimicgen 

4. Install the adaptive_teaming (COIL) package:

   pip install -e .

---

Planning with COIL

conda activate coil-env

The simplest way to run COIL is via the run_interaction_planner.py script:

python scripts/run_interaction_planner.py env=pick_place planner=fc_pref_planner task_seq.num_tasks=10 render=True

This script will simulate human-robot collaboration in our mujoco and robosuite based pick_place environment using the specified interaction planner. Results will be displaye on the terminal and automatically logged in a sub-directory in the outputs directory.

Parameters can be modified using the command line. For example, use render=False to disable rendering. All parameters and options are specified in cfg/run_interaction_planner.yaml. Please look at make_planner and make_env functions in scripts/run_interaction_planner.py for the list of supported planners and environment.

---

Citation

If you use our work or code in your research, please cite our paper:

@inproceedings{vats2025optimal,
  title={Optimal Interactive Learning on the Job via Facility Location Planning},
  author={Vats, Shivam and Zhao, Michelle and Callaghan, Patrick and Jia, Mingxi and Likhachev, Maxim and Kroemer, Oliver and Konidaris, George},
  booktitle={Robotics: Science and Systems (RSS)},
  year={2025},
}