Computing Lyapunov Functions II

Follow live at https://slides.com/d/ug2wzQI/live

(or later at https://slides.com/russtedrake/spring24-lec09)

MIT 6.8210: Underactuated Robotics

Spring 2024, Lecture 9

prog = MathematicalProgram()
x = prog.NewIndeterminates(2, "x")
f = [-x[0] - 2 * x[1]**2, -x[1] - x[0] * x[1] - 2 * x[1]**3]

V = x[0]**2 + 2 * x[1]**2
Vdot = V.Jacobian(x).dot(f)

prog.AddSosConstraint(-Vdot)

result = Solve(prog)
assert result.is_success()

Verifying a Lyapunov candidate

prog = MathematicalProgram()
x = prog.NewIndeterminates(2, "x")
f = [-x[0] - 2 * x[1]**2, -x[1] - x[0] * x[1] - 2 * x[1]**3]

V = prog.NewSosPolynomial(Variables(x), 2)[0].ToExpression()

prog.AddLinearConstraint(V.Substitute({x[0]: 0, x[1]: 0}) == 0)
prog.AddLinearConstraint(V.Substitute({x[0]: 1, x[1]: 0}) == 1)

Vdot = V.Jacobian(x).dot(f)

prog.AddSosConstraint(-Vdot)

result = Solve(prog)
assert result.is_success()

Searching for a Lyapunov candidate

Variations and extensions

  • Implicit dynamics
  • Piecewise-continuous dynamics
    • e.g. closed-loop w/ input limits
    • contact mechanics

Lecture 9: Computing Lyapunov Functions II

By russtedrake

Lecture 9: Computing Lyapunov Functions II

MIT Underactuated Robotics Spring 2024 http://underactuated.csail.mit.edu

  • 729