To use repository,
git clone https://github.com/jacobmoroni/diff_flat_ws.git &&
cd diff_flat_ws &&
git submodule update --init --recursive &&
catkin_make &&
source devel/setup.bash
To run simulation
for LQR and feed forward control:
roslaunch differential_flatness lqr_ff.launch
For PID and feed forward control:
roslaunch differential_flatness pid_ff.launch
For feed forward control only, no feedback control:
roslaunch differential_flatness ff_only.launch
For PID only, no feed forward term:
roslaunch differential_flatness lqr_ff.launch control_type:=o
To only show states plotter not gazebo for any of the above commands, just add visualize:=false to the end.
mostly derived from this
paper
Background
State Space
Block Diagram for differential flatness control
The following is the explanations of the blocks in the block diagram:
TRAJECTORY:
DIFFERENTIALLY FLAT:
LQR:
F and F inverse:
Assumptions of Differential Flatness and remedies to minimize issues from them.
Modified Block Diagram for PID control
traj.py generates the trajectory and outputs derivatives of the trajectory to be used by controllers
LQR.py Solves for K gain by solving algebraic riccati equation, then computes control on error state and adds in feed forward control
inverse.py Performs f inverse mapping to convert control from accelerations and yawrate to angle thrust and yawrate
attitude_control.py Adds the thrust compensation to keep discussed in assumptions section
states_plotter.py plots desired and actual states during flight
df_controller.cpp PID controller used to get multirotor into position before trajectroy begins, then passes control from inverse directly through to multirotor
controller_old.cpp PID controller used for PID feed forward used in pid_ff.launch tuned better for differential flatness control
multirotor.yaml parameter file for multirotor