A new demo for traversability analysis in agricultural robotics: this application allows to load a single points cloud or a dataset, try filtering, frame reference transformation and traversability analysis.

  • Download here the demo and here the source code--- This is a highly experimental software, I am not responsible for any possible damage on your system, use it on your own responsibility!

  • Download here a sample dataset

Using instruction:

  1. Download the demo and install it, I have noted that it does not create a shortcut on desktop or main menu, so locate where the demo is installing and run the demo from there; Default folder is "c:\program files x86\PCL_upd_DEMO\ "

  2. Dowload here a dataset, (or you can use your own if you have !); Dowload and extract the file (there is a code inside, just through it out) and use only the dataset inside the Ext_data folder;

  3. Load the point cloud list in the dataset folder, File -> Open -> Points Cloud list

    • you can use this function to load any txt file containing a list of cloud files, there is the function to load a folder without the list but I have noted some bugs there!!! I'll fix it in a future release

  • Load the images file using the same procedure, it is not mandatory but it provides a good visible feedback;

  • The UPD is INDEPENDENT from the reference frame but if you also plan to see a good traversability analysis which also consider the climbing capability of a robot you should transform the cloud from the robot reference frame to the camera reference frame. So, go to the "Frame" tab and set:

    • Pitch -> 90deg

    • yaw -> 90deg

    • check "apply to all" and push "apply"

  • You can update the point cloud visualization using the "Points Cloud" tab, just double click to one of the cloud (it doesn't mater which one)

  • Then go to UPD tab and run, if everything works you will get a message with the cloud generated and its size

  • Check "visualize traversability analysis" and push "update"

  • Now you should see a colored cloud, you can try to change robot climbing parameters and unevenness threshold to see what happen. Every time you do it you should update, it is not automatic (sorry for this, I'll fix it!!!)

What you see when you change these parameters is that the terrain is analysed using different thresholds, if your robot can climb high rocks you can have a low unevenness threshold, on the contrary a car should drive on a really smooth terrain so it should be quite high. Wheres climbing parameter analyses how much can the robot climb, so it is in degrees and it should be low for a car and high for other vehicles.