roborregos / home

Service robot development for RoboCup @Home OPL

RoBorregos RoboCup @HOME 2024

Current development of RoBorregos, for the RoboCup @HOME competition, in the Open League Platform (OPL). The goal is to built a service robot capable of accomplishing tasks in domestic applications. The software areas being developed are: Human-Robot Interaction and Cooperation, Object Manipulation, Computer Vision, and Navigation and Mapping.

This repository features the task_manager package, in charge of managing Behavior Integration with all the areas. For more information on the project, check our documentation.

Setup

Each development area has its own repository, added as a submodule here. To only clone this repo run:

git clone https://github.com/Roborregos/home

And to update the submodules content execute:

git submodule update --init --recursive

All the development environments use Docker, follow the instructions below or the README inside the docker folder for insights.

Software Architecture

Docker Development

The project uses Docker for easier development within ROS and CUDA/Jetson compatibility. Both this main engine repository and each area's contain a docker folder with dockerfiles and a Makefile for easier image and container creation and modification.

Requirements

• Docker Engine
• Post-installation steps for Linux If using GPU:
• NVIDIA Driver
• NVIDIA Container Toolkit)

Container Creation

To build an image, run:

# For CPU
make main.build
# For GPU
make main.build.cuda
# For Jetson L4T: 35.4.1
make main.build.jetson

To create a container, run the following commands. The ws folder is mounted by default, and additional folders can be added with the volumes argument, with both absolute and relative paths allowed:

# For CPU
make main.create volumes="another_folder1/,~/another_folder2"
# For GPU
make main.create.cuda volumes="another_folder1/,~/another_folder2"
# For Jetson L4T: 35.4.1
make main.create.jetson volumes="another_folder1/,~/another_folder2"

To enter the container, run:

make main.up
make main.shell

You can stop and remove the container with:

make main.down
make main.remove

Additional commands can be added within the Makefile and the scripts inside the docker/scripts folder can help for easier integration and sharing. These include a build script to run the dockerfile and create a new image and a run script to create containers from it. Any additional dependency or system/environment configuration should be added to these scripts.

Task manager

The frida_task_manager is the central package in charge of processing the command information and directing the tasks with its respective area. For code readability, each area development its inside a Python module area_tasks.py, and they are called by task_manager_server.py, the script were the ROS node is created.

At the top of this mentioned file, there are four constants to enable or disable the tasks, you can change any of these to False is for the current test there is no need of an area:

NAV_ENABLED = True 
MANIPULATION_ENABLED = True
CONVERSATION_ENABLED = True 
VISION_ENABLED = False

For executing the task_manager, follow the above steps of Docker setup. Inside the bash terminal, setup the ROS network and execute the node:

export ROS_MASTER_URI=http://192.168.31.23:11311 # IP of the ROS Master (if master is in another machine)
export ROS_IP=192.168.31.97 # IP of own machine
rosrun frida_task_manager task_manager_server.py

Team Members

Name	Github	Role
Adán Flores	@afr2903	Integration, HRI & Manipulation
Emiliano Flores	@EmilianoHFlores	Integration, Manipulation & Computer Vision
Iván Romero	@IvanRomero03	Integration, HRI & Computer Vision
Alejandra Coeto	@Ale-Coeto	Computer Vision & Manipulation
Oscar Arreola	@Oscar-gg	HRI & Web
Alexis Chapa	@Chapa-1810	Manipulation & Navigation
Marina Villanueva	@mariinaVillanueva	HRI
David Vázquez	@Deivideich	Electronics, Navigation & Manipulation
Diego Hernández	@Diego-HC	Navigation
Franciso Salas	@Francisco-SP3	HRI, Navigation
Leonardo Sánchez	@LeoLFSH	Mechanics
Alex Guerrero	@alex-guerreroc	Mechanics

home's Issues

Easier container integration

A Makefile or further documentation to setup the development environment could enhace the CI Workflow. Also, the submodules were initialized but they are targeting a temporal and not visible branch. It may not be a problem but it's worth taking a look at it.

Task Manager Serve Breakfast - First Iteration

Development within areas for serve breakfast has reached a stable state where points can be obtained from the task. Using the task manager template and pipeline being developed in the GPSR task, develop a first iteration of a task manager that can pick and place breakfast objects from one table to another and pour milk and cereal to a bowl.

Create utiity scripts

It's very common to have data problems whenever there's integration between different areas. This happens because different algorithms are created to solve the same problem or to retrieve the same data.

Create utility folder and scripts for the multiple areas to use if necessary and have the same data by using the same algorithms.

Main engine migration

Base on the legacy repo Main engine, and adapt a template and documentation to build all main engine's on top of that template

Add speech devices in dockerfiles

• Add audio devices to docker.
• Script to create pulse audio socket

Task Manager Carry my Luggage - First Iteration

New carry my luggage required the robot to pick a bag pointed at by an operator and then follow them until asked to return. The robot is currently capable of most of these tasks but the task requires integration of navigation, vision and manipulation not yet developed.

Compose execution

Define the setup for the external device computing, for running the Receptionist task first and then transfer the implementations for all the tasks and devices.

Task manager GPSR first iteration

Build the Main Engine for the GPSR task

Task Manager Storing Groceries - First Iteration

The storing groceries task implies taking objects from a table with precise location unknown and place them on shelves according to the already stored objects. A task manager following the structure being developed for GPSR has to be developed to accomplish this task, successfully carrying objects from the table to the cabinet, categorizing already stored objects and storing them in the assigned shelf.

Init submodules and repo structure

Intialize the configuration of this main repo, along with the repositories of each area as a submodule.

Task manager Receptionist

For the task of Receptionist of Stage 1

Speech feedback on executing task

Related to this HRI issue.

Call the speak feature asynchonically while executing other tasks.

Integration messages

Due to the architecture of multiple workspaces, it was decided every module would develop its own integration messages package with name frida_area_interfaces, and share it with the task management repo.

The custom messages or actions should only include ROS standard messages (std_msgs, geometry_msgs). This to avoid installing unnecessary dependencies in the main repo, and maintaining simplicity.

ops: Timeout action server

Add timeouts and more useful debugs to the task manager execution

GPSR second iteration

Integrate developments made in the other task managers and enhance the current capabilities