Full state estimator for robots.
Pinocchio [https://stack-of-tasks.github.io/pinocchio/download.html]
Python installs, should be available in pip3
- casadi
- casadi-kin-dyn
To get the casadi bindings natively (i.e. skipping the casadi-kin-dyn library), we need to compile Pinocchio from source, from the branch pinocchio3-preview
https://stack-of-tasks.github.io/pinocchio/download.html
You'll also need to grab the FindCASADI.cmake file from Casadi, put it in the cmake folder, include it in the CMakeLists.txt, and rename it to Findcasadi.cmake.
We first set up ROS and get the robot model loaded,
source urdf/devel/setup.bash
roscore &
roslaunch ur_description load_ur16e.launch
For the optimization of params, an offline trajectory is needed, which is stored in a .pkl file of the same name as the bag it's generated from.
python3 main.py --new_traj --bag [bagname]
Then we can optimize the parameters.
python3 main.py --opt_param --bag [bagname]
The optimized parameters are set as rosparams, which will be loaded when an EKF is carried out.
You can then do an offline EKF pass by python3 main.py --new_traj --bag [bagname], optionally giving --est_geom or est_stiff to estimate these parameters online. If you'd like ot run the filter online, leave out the --new_traj