xijunke / multi_scale_search

Multi-Scale Object Search

This repository contains the implementation of our work currently under review for ICRA 2021:

• Luc Holzherr, Julian Förster, Michel Breyer, Juan Nieto, Roland Siegwart and Jen Jen Chung, “Efficient Multi-scale POMDPs for Robotic Object Search and Delivery”

Installation

The following instructions have been tested with Python 3. First, clone the repo.

git clone --recursive --branch devel [email protected]:ethz-asl/multi_scale_search.git

Next, we recommend to create a new virtual environment.

python3 -m venv --system-site-packages .venv
source .venv/bin/activate
pip install -r requirements.txt

The package can either be built using catkin,

catkin build multi_scale_search
source path/to/catkin_ws/devel/setup.zsh

or alternatively installed into the virtual environment using

pip install -e .

Finally, compile SARSOP.

cd third_party/sarsop/src
make

Running the Experiments

The following command will open a GUI that allows you to select and run different scenarios with different agents.

python scripts/run_experiment.py --gui

For example, selecting scenario 01, then clicking set test scenario followed by Start Simulation will produce the following solution: