motoko ice dragon line follower

source structure

• 3d_print
• hardware
• simulations
• firmware

LQR controller

motor velocity control

• inner loop controls brushless motors
• LQR + Kalman are used for velocity control
• runs 4kHz

position control

• outer loop controls robot state - distance and orientation
• LQR controller is used
• state is : (distance, orientation, velocity, angular rate)
• runs 250Hz

LQR optimal control

• basic idea is solving discrete riccati algebraic equation
• to remove distrubance, integral action term is included

motors velocity control

step response 1st order identification

• video of identification proces : https://www.youtube.com/watch?v=B-Wuzev6Ze0

• first the max RPM is estimated, using open loop and FOC controll

• during measuring max RPM the encoder noise variance is also measured (necessary for kalman filter)

• next step is calculating motor time constant (considering 1st order model)

• c++ source identification.cpp

• LQR synthesis lqg_discrete.ipynb

• step response and disturbance rejection

• controller for brushless motors in c++ (1D LQR + kalman) : lqg_single.cpp

• whole motor driver : motor_control.cpp

position controll

• kalman based identification
• LQR synthesis
• controller identification + synthetis
• step response and disturbance rejection

photos

• photos

• brushless motors
• LQG - linear quadratic regulator + Kalman filter
• LSTM for line shape prediction
• multiple local optimal LQR controllers selecting

software architecture