neural_Lyapunov

This repo constructs the Lyapunov candidate based on neural network.

The controller is specified below:

    kD = diag([5,10]);
    de = dq - dr;
    ddq = -kD * de + ddr;

Step 1. Show the tracking error.

Plot the tracking error, e(t).

Step 2. Visualize the reference velocity and the actual velocity.

Plot (dr1,dr2) and (dq1,dq2) diagrams.

velocity_animation.mp4

Step 3. Find the Lyapunov Candidate using Neural Network.

The found Lyapunov candidate has been tested for the unseen intial condition, leading to a good result with negative constraints as expected.