An omnidirectional mobile platform, with a 3 omnidirectional wheels layout, with two objectives: to facilitate the development of mobile robot control algorithms and to facilitate the development of mobile robots in general by providing a ready to use movement solution.
It is free and open source, licensed with the MIT/Expat license, with the intent of powering the research of robotics anywhere for anyone.
It’s provided with tutorials and the code is commented and modularized, to facilitate the work with it as much as possible.
It is the result of my final graduation work in Computer Engineering and, as such, it has a lot of scientific research behind it. I could make it available here too, but it's in Portuguese, and i would have to translate about 50 pages to English, so it will not come in the near future.
The idea was to make a physical model, also with all the necessary instructions, but the hardware didn’t arrive in time, so only the simulation part for validation has been developed.
It uses the Gazebo simulator, so you need to know how to use it first. I recommend [1] and [2] and skimming over [3] is good.
It’s also a good idea to at least superficially understand the kinematics. The works of [4] and [5] are very good starting points and are freely available online.
This is an image I created that should contain all necessary geometric constraints to derive the kinematics equations:
Remember that V1, V2 and V3 are also called, V_left, V_back and V_right, respectively. These are the forward kinematics equations relative to the robot's reference frame:
If you need forward kinematics relative to the world's frame, transform the result of the last equations with the following ones:
For the inverse kinematics, you can start with the velocities relative to the robot's frame, or you can convert the velocities relative to the world's frame with the following equations:
Then, you can use the following equations for the inverse kinematics relative to the robot's frame:
Given the purpose of this work, I may improve it in the future with more advanced features like a dynamic model instead of a kinematic model or better movement control, but since this is an open project, please do contact me (or use the available GitHub funcionalities) to improve this work so everyone can have an even better experience, specially if it’s a bug fix or a correction in the text of the tutorials.
The "model" folder contains the model of an omnidirectional wheel (38MM DOUBLE PLASTIC OMNI WHEEL) and the model of a robot made out of an abstract body and these wheels for the sake of validation. The models are Gazebo readable: just drop them in Gazebo's default model folder and you'll be able to add them to any Gazebo world through the GUI.
The "plugin" folder contains the source code of the control plugin I developed, and a demo plugin that makes the robot execute some random movements for testing. It needs to be compiled because the plugin compiled in a system might not work on another.
The "tutorial" folder contain several tutorials that need to be read so this project can be used. Start with the one called "getting started".
The "world" folder contains a Gazebo world that simply contains the default ground plane and the aforementioned abstract robot model in the center of this world.
This project was developed either in Gazebo 7 or 8. Unfortunately, I don't have the system that I used anymore and I don't remember which version was is. Gazebo 9 brought breaking changes in the API and this project has been updated accordingly. A branch called gazebo7 has been created to contain the version of the project that is compatible with Gazebo 7 (or 8). Now, the master branch contains the version of the project that is compatible with the latest version of Gazebo (9 for now). Thanks tuanngo0898 for letting me know!
[1] http://gazebosim.org/tutorials?cat=guided_b&tut=guided_b1
[2] http://gazebosim.org/tutorials?tut=plugins_hello_world&cat=write_plugin
[3] http://gazebosim.org/tutorials?cat=guided_i&tut=guided_i1
[4] J. Borenstein et. al., “Sensors for Dead Reckoning”, in “Where am I?” Sensors and Methods for Mobile Robot Positioning, Ann Arbor: Univ. of Michigan, 1998, ch. 1, sec. 1.3.5, pp. 25–26.
[5] J. Gonçalves et. al., “REAL TIME TRACKING OF AN OMNIDIRECTIONAL ROBOT - An Extended Kalman Filter Approach”, in ICINCO 2008, PROCEEDINGS OF THE FIFTH INTERNATIONAL CONFERENCE ON INFORMATICS IN CONTROL, AUTOMATION AND ROBOTICS, ROBOTICS AND AUTOMATION 2, FUNCHAL, MADEIRA, PORTUGAL, MAY 11-15, 2008, Funchal, Madeira, 2008, p.5–10.
When I developed this entirely for Gazebo, I only had heard about ROS. Then I started working with it and decided to port this project to ROS. It's usable, but keep in mind it's in a very early development stage. Besides, I stopped working with ROS and higher priority stuff showed up.
I still have great plans for it, and it's supposed to supersede the Gazebo version, but it will take quite a while before I can get back to it.
Before using it, you'll need to install some packages. These are more than you need, but it doesn't hurt to install them all. Replace <version> by your ROS version, e.g. indigo, kinetic, lunar, etc.
sudo apt-get install ros-<version>-effort-controllers
sudo apt-get install ros-<version>-joint-state-controller
sudo apt-get install ros-<version>-position-controllers
It's been long enough that I forgot how and why I did some of the things that I did but, to get it moving, launch velocity controller.launch. Then, after the simulation has been started in Gazebo, you have two options:
- Write in a terminal
rostopic pub /open_base/commandthen hitTaba few times to autocomplete the message template and fill the values; - Edit the
scripts/movement*.shfiles to your needs and run them withsource scripts/movement*.sh.
Or to control the robot using keyboard: teleop_keyboard_omni3 (thanks YugAjmera!)
Explanation of Movement.msg:
movement: 0 for Bézier curve; 1 for generic movement (see below); 2 for none; 3 for wheel speed.generic:type: 0 for absolute position; 1 for relative position; 2 for velocity (linear and angular).frame: 0 for hybrid; 1 for mobile; 2 for raw mobile; 3 for world.target: desired position in(x, y, θ)or desired velocity in(vx, vy, vθ).
bezier: coming soon.
See here for more explanation about movements, specially the parts about what I called "hybrid" and "raw". What I called "direct" there I called "generic" here.