nmpc_ddp
|
Non-linear model predictive control (NMPC) with differential dynamic drogramming (DDP)
https://github.com/isri-aist/NMPC/assets/6636600/02f64c91-88aa-42d8-abfd-f8062d7406e9
See here.
See the following for a detailed algorithm.
The source code implementation is based on the following.
Make sure that it is built with --catkin-make-args tests
option.
Controlling on CoM-ZMP dynamics with time-variant CoM height. System is linear and time-variant.
Controlling vertical motion with time-variant number of contacts (including floating phase). System is linear and time-variant. The dimension of the control input changes (there are even time steps with an empty control input). The upper and lower limits are imposed on the control input.
Controlling cart-pole (also known as inverted pendulum). System is non-linear. This is an example of a realistic setup where the control and simulation have different time periods.
You can interactively add disturbances and change the target position via the GUI. See the video.