mees / AIS-Alexa-Robot

Alexa skill to control and interact with a robot via ROS

AIS Alexa Robot Skill

This repository contains an Alexa skill to control and interact with a robot via ROS topics. Concretely, the current skill implementation was used for picking-and-placing tabletops objects from natural language. More information at the project page.

Reference

If you find the code helpful please consider citing our work

@inproceedings{mees21iser,
  author = {Oier Mees and Wolfram Burgard},
  title = {Composing Pick-and-Place Tasks By Grounding Language},
  booktitle = {Proceedings of the International Symposium on Experimental Robotics (ISER)},
  year = 2021,
  address = {La Valletta, Malta}
}

@inproceedings{mees20icra_placements,
author = {Oier Mees and Alp Emek and Johan Vertens and Wolfram Burgard},
title = {Learning Object Placements For Relational Instructions by Hallucinating Scene Representations},
booktitle = {Proceedings of the IEEE International Conference on Robotics and Automation  (ICRA)},
year = 2020,
address = {Paris, France}
}

Usage

First, launch rosbridge_server:

roslaunch rosbridge_server rosbridge_websocket.launch

In a separate shell, make your local port visible for the VM

ssh -R 9090:localhost:9090 aisalexa /opt/aisalexa/forward_port.sh

In case your local server does not run on port 9090, replace the first port number only.

There will be three ros topics: /alexa/command, alexa/debug and alexa/reply. See test_response.py for examplary handling of commands.

Start server

Start all containers

• cd /opt/aisalexa/nginx
• sudo ./start.sh

Show container logs

• cd /opt/aisalexa/nginx
• sudo docker-compose logs aisalexa (leave out the service name to get the logs of all services)

Setup

see Setup

For configuration, change the environment variables in nginx/.env.

Update configuration by using git pull. When running start.sh, updated images (i.e., the updated skill) are retrieved automatically. Remember that changes to the skill must be pushed to aisgit for the updated docker container to be built

License

For academic usage, the code is released under the GPLv3 license. For any commercial purpose, please contact the authors.

Authorship

This code was developed by Henrich Kohlkhorst and Oier Mees at the University of Freiburg, Germany.