Our team at Erle Robotics is glad to announce that the simulation for the Erle-Rover has been launched and open sourced.
For those that don't know it yet, Gazebo is a powerful real time physics simulator that allows the developers to test their algorithms in a secure and agile way. It runs along ROS (Robot Operating System) and that’s probably what makes it perfect, plus it offers a variety of sensors (such as lidars, cameras, range finders) in form of plugins that can be added to any robot model. We expect the APM community to benefit from this work and simulate their autonomous rovers using these tools.
The plugin has been developed over AurelienRoy’s plugin which, in a nutshell, is an interface between Gazebo and APM:Rover (ardupilot). Erle Robotics added to this plugin support for their Rover.
To show the power of this plugin, here are some behaviours we developed and tested:
Obstacle avoidance : a simple but powerful enough algorithm to avoid obstacles. This behaviour was tested first under Gazebo and then brought to real life scenario. It uses a 270º HOKUYO lidar sensor, but all the details and the source code can be found here.
Line following : line follower algorithm using the camera. Again, all details as well as the source code can be found here.
SLAM : SLAM application using hector_mapping node, a powerful SLAM approach that can be used only with a lidar sensor. All the details, including source code are available here. This also was translated to a real scenario.
Additional information and resources:
Comments
Hi Gary,
Great to hear that! From your comments, I'd say that you're good to go.
If you're interested to follow up on this I'd recommend jumping into our forum where our engineers provide support and announce the latest changes in Simulation and ROS support.
Regards,
Hi Victor,
I have pretty much decided to use your Erle platform for my new Traxis XMAXX that is currently in shipment to me.
My intention is to interface it to my Hokuyo URG4LX and use the ROS firmware referenced on the Erle site.
I am hoping to bring up an operational navigation platform to which I can add some hardware and software elements of my own.
Should make the whole process much easier.
I wondered if you might have any suggestions regarding this and if it might be OK to contact you in the future with questions - I suspect there will be a few.
Best Regards,
Gary
Hi Victor,
I have a Hokuyo and may give that a try.
Just a suggestion, the Lightware guys in South Africa have this lovely little scanner that really outperforms the Hokuyo and at only $900.00 it is a huge bargain.
http://www.lightware.co.za/shop/en/products/45-sf40c-100-m.html
In my case I might try to bring it up on the Hokuyo then shift to the Lightware one.
Even though I have a Kinect, I am more interested in trying the newer Kinect 1.
It does quite a bit of internal processing on it's own, but I am planning on using an external processor anyway.
I am looking at both RPi3 and Nvidia TK1 or TX1.
I have a TK1, kind of a beast to program in Cg though.
In any case, really nice Gazebo simulation.
Best Regards,
Gary
Hi Gary,
The sensor mounted is a Hokuyo LIDAR. We've used it to translate the simulation in real scenarios and worked nicely. Erle-Brain 2 had enough computation capability to run the ROS, the autopilot and process the samples coming from the sensor to react accordingly. SICK's TiM5xx scanner family would also do a nice job if you happen to have one around.
Besides 2D LIDARs you could also mount a Kinect in you rover and adapt the SLAM algorithm to use this sensor but you'd probably need a dedicated computer to process all the information coming from the sensor. That'll be a nice project to share :)!
Regards,
Really great job Erle,
Tell me is that Kinect sensor integrated and what capabilities does it provide?
Also is it original Kinect or later TOF Kinect made for Xbox 1?
Definitely time to start paying attention, they shipped my new Traxis XMAXX today and I am planning on using it as exactly this sort of platform.
Thank you for bringing us this.
Best Regards,
Gary
Looks great. I guess algorithms such as obstacle avoidance are simple... once you find the right algorithm.. which is of course the tricky part:)
I really would lve to try this out, when I have some time ! Too many good things like this going on these days.. Can't keep up :)
thank you so much to share your achievement.