This page contains answers to some common questions about the belief space planning library based on OMPL that I have developed in my research group based on my own work and that of Dr. Ali-akbar Agha-mohammadi. You can find the library on Github.

Who uses this library?

Please visit our research group EDPLab for more information about our research, you can find links to videos and papers that use this library. This library has also been used by students at MIT for running FIRM on a quadrotor, you can find their code here.

Can you initialize the graph/roadmap with a stored one or save the one generated at runtime?

We have the choice to do both, i.e., we can load a previously computed roadmap or make the  planner generate a new one. When the planner starts up, it looks for a FIRMRoadMap.xml file in the top directory. We use this file to load a pre-computed roadmap. If this file is found, the planner will load it and use it (so you get to re-use an old map). After completing planning, the FIRM planner will update the file if it already exists in case new FIRM nodes/edges are added to the graph. If there is no such file, a new roadmap will be generated and saved to the same file name (FIRMRoadMap.xml). You can then move this xml file to the SavedRoadmaps folder for later use.

What are these c11, c12 values in the FIRMRoadmap.xml file?

C11 etc are the covariance matrix values for the FIRM nodes. When the state space is x,y,theta (2d). The covariance matrix is a 3×3 matrix and these are its entries.

How do you handle collision checking?

We use the FCL collision checker library wrapper provided by OMPL to handle the collision checking and cad objects (mesh). Please check with the FCL  / OMPL developers for your questions about how these work.

What are the landmarks and how can I provide my own map of world features?

First of all, by map here we mean the map of the features / landmarks in the world. There is a set of setup files in the “SetupFiles” folder which contain information about the landmarks, the robot’s motion and observations model etc. We can choose which setup file to use by setting the correct path in the main.cpp file. You can create your own setup by creating a new setup xml file and adding the marker locations, robot motion noise, observation model noise etc. accordingly.

How are the robot and environment mesh loaded?

Please have a look at the file FIRM2DSetup.h, in the “loadParameters” function to see how we load in the robot mesh.

Can I load obstacles while the planner is running, i.e., dynamic obstacles?

Currently, it is only possible to load a mesh in the beginning in the beginning of the simulation. You cannot insert meshes while the planner/simulation is running (dynamic obstacles). This would be a great feature to have but I haven’t implemented it yet :(.

Is it possible to use a Quadrotor or any other robot model?

Yes yes yes! The planner is independent of what robot you have. We define a motion model and observation model class. You can define a new motion model and observation model for your robot (could be a quadrotor). Just like we have UnicycleMotionModel.cpp you can create your own ABCMotionModel derived from MotionModelMethod (cpp and header file). Please look at our code which is commented for your convenience.  For example, since a quadrotor pose has 6 dimensions (x,y,z,roll,pitch,yaw), you would need to create a new belief space (SE3BeliefSpace). Good news is that some students at MIT have already implemented this stuff in their own fork of our library. You can see their page here. Please fork our library if you would like to add more robot models, we would love to add this stuff back to our repository for everybody’s benefit.


One thought on “FAQ: Belief Space Planning With OMPL

  1. Hello Saurav,

    Does Dynamic Feedback Linearization (DFL) control used in these codes? I couldn’t find where it is implemented.

    Thanks for your help

    1. Hi Ali,

      DFL is not implemented in this code currently. If you’re interested in implementing DFL FIRM, let us know we would be happy to take your contribution.


Leave a Reply

Your email address will not be published. Required fields are marked *