Pumas Navigation2

Service robots are intended to help humans in non-industrial environments such as houses or offices. To accomplish their goal, service robots must have several skills such as object recognition and manipulation, face detection and recognition, speech recognition and synthesis, task planning and, one of the most important, navigation in dynamic environments. This repository describes a fully implemented motion-planning system that comprehends from motion and path planning algorithms to spatial representation and behavior-based active navigation.

This paper can be consulted online for free at this link; this system has won the Smoothest, Safest Navigation Award in 2022, 2023, 2024, and 2025 at the international robot competition "RoboCup" in the "at Home" league.

Please, if you use this material, don't forget to add the following reference:

@article{negrete:2018,
author 		= {Marco Negrete and Jesus Savage and Luis Contreras},
title 		= {{A Motion-Planning System for a Domestic Service Robot}},
journal		= {{SPIIRAS Proceedings}},
volume		= {60},
number		= {5},
pages		= {5--38},
year		= {2018}
}

The ROS2 version details are described in:

@article{contreras:2025,
author 		= {Luis Contreras and Marco Negrete and Tomoaki Yoshiaki and Yuichiro Hirose and Hiroyuki Okada},
title 		= {{A Motion Planning Framework for Multi-Robot Navigation in ROS 2}},
journal		= {{RSJ 2025}},
year		= {2025}
}

Setup

1. Create an env of your choice.

e.g.

First, create a workspace:

mkdir -p ~/pumas_nav2_ws/src

Then, clone this repository into the src folder:

cd ~/pumas_nav2_ws/src
git clone https://github.com/artenshi/pumas_nav2.git

2. Install missing packages:

e.g.

sudo apt update
sudo apt install ros-humble-control-msgs libpcl-dev ros-humble-pcl-conversions ros-humble-pcl-ros

3. Build the project:

e.g

cd ~/pumas_nav2_ws
colcon build

Robot Navigation

Structure

Topics

TODO

Services

TODO

Initialization

Prerequisites

To use SLAM and navigation, you need to start a mapping system.

e.g. SLAM Toolbox

Also, a localisation system is required for the correct TF2 frame transformations.

e.g. Nav2 AMCL

or EMCL

Launch

ros2 launch navigation_start navigation.launch.xml

or

ros2 launch navigation_start navigation_namespace.launch.xml namespace:=my_robot_name

Notes

Relevant parameters can be found in the navigation launch files in ~/pumas_nav2_ws/src/pumas_nav2/navigation_start/launch. In specific:

Update your map names

  <arg name="static_map_file"  default="/path/to/your/map/maps/map.yaml/map.yaml"/>
  <arg name="prohibition_map_file"  default="/path/to/your/map/prohibition_maps/map.yaml"/>

For robot localisation, we use the original slam map from the map_server map_saver in maps/maps/. Additionally, for path planning, we edit the previous map to add prohibited and or closed areas (e.g. considering a laser-scan-based mapping, we add the unmapped table area, or, to travel between two main doors, to avoid surrounding it from the outside, we add closed edges; we do not do this with the localisation map to avoid misslocalisation errors between the real sensor reading and an edited map.)

Enable/Disable potential fields

<arg name="use_pot_fields" default="True"/>

For dynamic obstacle avoidance, we use rejective potential fields that can be enabled or disabled with this parameter.

Additionally, in the package:

<node name="potential_fields"  
      pkg="potential_fields"  
      exec="potential_fields_node"
      namespace="$(var ns_prefix)"
      output="screen"
      respawn="true">
  <param name="use_namespace" value="$(var use_namespace)"/>

modify the following parameters accordingly:

  <param name="laser_pot_fields_d0" value="0.50"/>
  <param name="laser_pot_fields_k_rej" value="0.40"/>

  <param name="cloud_pot_fields_d0" value="0.80"/>
  <param name="cloud_pot_fields_k_rej" value="0.20"/>

these parameters indicate the obstacle's distance and rejective force during navigation (too low means that the robot won't avoid obstacles, and too high means that the robot will avoid obstacles too far from it -- these two extreme cases might cause undesired behaviours).

Moreover, in the package:

<node name="map_augmenter" 
      pkg="map_augmenter" 
      exec="map_augmenter_node"
      namespace="$(var ns_prefix)"
      output="screen"
      respawn="true">
  <param name="use_namespace" value="$(var use_namespace)"/>

modify the parameters:

<param name="decay_factor" value="20"/>    
<param name="inflation_radius" value="0.22"/>
<param name="cost_radius" value="0.25"/>

that best suits your navigation requirements, where decay factor refers to the time that the temporal obstacles remain in the map (in time cicles), the inflation radious is the additional size around those obstacles (in meters), and the cost radius is the relevance of the objects for path planning.

Future Work

Similar to our ROS1 version:

• Person follower
• Python wrapper

New features:

• MAPF (Multi-agent Pathfinding)

Authors

• Luis Contreras - QibiTech Inc. / TID Professional University
• Marco Negrete - BioRobotics UNAM
• Ryohei Kobayashi - Kyushu Institute of Technology