ROS packages for TIME4HRI: Learning TIME constrained sequences for natural Human-Robot Interaction

Introduction

This repository contains the source code and supporting documentation for the TIME4HRI project, developed under the euROBIN cascade funding. TIME4HRI (Learning TIME Constrained Sequences for Natural Human-Robot Interaction) focuses on the development of a Dynamic Neural Field (DNF) architecture for improving human-robot collaboration in dynamic, time-sensitive environments, such as assembly work in industrial settings. The project aims to enhance the ability of robots to adapt their behavior based on human interaction, ensuring smooth collaboration and error detection during the task execution.

The EuroCore repository itself does not directly contain the DNF implementation but provides links to the individual repositories where the core components and system configurations of the DNF architecture are maintained. It acts as a central access point for the various software packages and codebases related to the TIME4HRI project.

Purpose and Scope

This page serves as the central hub for:

• Linking to all relevant repositories associated with the TIME4HRI project.

• Facilitating the integration of the Dynamic Neural Field (DNF) architecture with different robotic platforms.

The repository’s goal is to support the euROBIN network by centralizing resources that help researchers and developers work with the TIME4HRI DNF architecture across different robotic platforms.

TBA: Demonstration videos.

Architecture and Communication

While this repository does not directly include the implementation details of the DNF architecture, it provides links to the repositories where the DNF is implemented. The TIME4HRI project is designed to work with robotic platforms via communication protocols like ROS1 and ROS2, enabling real-time adaptation to human interactions.

The communication between the DNF system and robotic platforms involves:

• Inputs: External data streams (e.g., camera, microphone)

• Outputs: Robot control commands (e.g., motion, grasping, speech)

For detailed information about node interactions, configuration, and execution, please refer to the specific GitHub repositories linked below.

Documentation and Setup Instructions

This repository provides access to the overall setup and configuration guides, which include:

• Links to detailed documentation for setting up and configuring the DNF architecture on different robotic platforms.

• Information on system requirements and dependencies.

Detailed setup and usage instructions for each repository can be found in the respective README files of the linked projects.

Related Publications

The Dynamic Neural Field (DNF) architecture was first implemented in MATLAB in a previous publication. For further details on the development and application of the architecture, please refer to the following paper:

Wojtak, W., Ferreira, F., Louro, L., Bicho, E. and Erlhagen, W., 2023. Adaptive timing in a dynamic field architecture for natural human–robot interactions. Cognitive Systems Research, 82, p.101148.  https://www.sciencedirect.com/science/article/pii/S1389041723000761

TBA: link to the resulting article.

Links to GitHub Repositories

• Python DNF Implementation: https://github.com/w-wojtak/dnf_architecture_python.git 

• ROS1 DNF Implementation: https://github.com/w-wojtak/tiago_dnf.git

• ROS2 DNF Implementation: https://github.com/w-wojtak/time4hri_dnf.git

• Speech module: https://github.com/w-wojtak/speech-recognition-for-ros.git

Contact

Weronika Wojtak (weronika.wojtak@ccg.pt)

CCG/ZGDV Institute

Acknowledgments

This work is supported by Cascade Funding form euROBIN, the European ROBotics and AI Network (Grant agreement 101070596), funded by the European Commission.