Plot of the controller tracking a sinusoidal reference within its steady-state bounds, comparing nonlinear, emulated and earlier designs over 20 seconds.

Flexible-Joint Robot Control.

Advanced Control
Simulink
State observers
LQR and LQG

Classical, state-space and optimal controllers for a single-link flexible-joint robot, designed and tested in a high-fidelity Simulink model.

Reports

Phase 1: Classical control
PDF, 1.3 MB
Download Open in a new tab (PDF)
Phase 2: State-space control
PDF, 2.2 MB
Download Open in a new tab (PDF)
Phase 3: Optimal control
PDF, 1.9 MB
Download Open in a new tab (PDF)

Step response of the tuned controller tracking the square-wave reference.

Classical.

The brief: swing the arm between 15° and 45° every 3 seconds with at most 2° steady-state error, 5° overshoot and a 2-second settling time, sampling at no more than 50 Hz.

Simulink model of the full-state observer.

Observers.

Pole placement used to design continuous and discrete full-state and reduced-order observers, compared across motor and sensor combinations.

Symmetric root locus plot used to choose the LQR weighting.

Optimal.

An LQR controller designed with the symmetric root locus and Riccati equation, then an LQG controller that handles a sinusoidal input disturbance.

Gallery.