L. K.
L. K.
The controller logs show an average loop time of ~1020µS for every impedance control thread,which is slightly above the limit for the desired 1kHz frequency.
Log usage of every thread / CPU used by the controller to determine load headroom.
Explain how to build and develop and run the code using the Docker container.
Create a `NLP` directory which contains both `nlp.so` and an auto-generated C++ header file containing the MPC parameters, such as prediction horizon length (`N`), time sample interval (`dt`) etc. Then...
Try JIT compilation for the NLP and increase optimization level to O3 as seen [here](https://github.com/casadi/casadi/blob/master/docs/examples/python/nlp_codegen.py). Also use `subprocess` instead of `os.system`.
[Plotly](https://plotly.com/python/) looks like a great solution for interactive plots, play around with it and decide if moving from matplotlib is worth it. It is important to check export capabilities before...
Implement something like a Pose3D or Orientation3D class that contains functions for accessing the quaternions as Euler Angles for readability and the quaternions themselves. Then use those class instances where...
See how fast the leg model can be integrated in C++ to determine whether or not it's feasible to use this "simulation" for MPC force limits and such.
The iteration times currently logged by the controller might be incorrect / artificially limited by the timed loop on the simulation-side plugin. Move this to controller-side to be able to...
Reset lift off position on unexpected contact change. For now it will just use the old trajectory which is not correct.