Higher Order Sliding Mode Control
Professor Aamer Iqbal Bhatti
Department of Electrical Engineering
Capital University of Science and Technology, Islamabad
April 26, 2019
c Dr. A. I. Bhatti, CASPR SMC Workshop April 26, 2019 1 / 14
Outline
1 Revisiting first order sliding mode
2 Higher order sliding mode
c Dr. A. I. Bhatti, CASPR SMC Workshop April 26, 2019 2 / 14
Revisiting first order sliding mode
Revisiting first order sliding mode
Advantages
Insensibility against perturbations (matching perturbations)
The choice of surface s(x, t) = 0 allow to choose a priori the closed-loop
dynamics
Disadvantages
Chattering phenomenon
s(x, t) = 0 must have relative degree equal to 1 w.r.t. u, the control input
c Dr. A. I. Bhatti, CASPR SMC Workshop April 26, 2019 3 / 14
Higher order sliding mode
Higher order sliding mode I
Objective
In higher order sliding mode (HOSM) the system trajectories are restricted to
evolve onto the sliding surface
Sr = {x ∈ Rn : s = ṡ = . . . = s (r −1) = 0}
(a) FOSM (r = 1) (b) HOSM (r > 1)
c Dr. A. I. Bhatti, CASPR SMC Workshop April 26, 2019 4 / 14
Higher order sliding mode
Higher order sliding mode II
Introduced by A. Levant (Ph. D. supervisor Emel’yanov) in 1987
Ideal
Sr = {x ∈ Rn : s = ṡ = . . . = s (r −1) = 0}
Real
|s| = O(Tsr ),
|ṡ| = O(Ts(r −1) ),
... = ...,
(r −1)
|s | = O(Ts ),
where Ts is sampling time
c Dr. A. I. Bhatti, CASPR SMC Workshop April 26, 2019 5 / 14
Higher order sliding mode
Higher order sliding mode III
Advantages
Robustness w.r.t. bounded matching perturbation,
Reduce the sliding dynamics up to at most (n − r ) (in-fact if counting the
added integrators exactly (n − r )).
Finite Time convergence to Sr,
Chattering reduction (sometimes see relative degree of s),
Higher convergence accuracy.
c Dr. A. I. Bhatti, CASPR SMC Workshop April 26, 2019 6 / 14
Higher order sliding mode
Second order sliding modes
Trajectories are steered in the vicinity of sliding manifold s = ṡ = 0 in finite
time
The second order sliding mode (SOSM) algorithms have been successfully
implemented on many practical systems
The most popular SOSM algorithms are
Real twisting
Super twisting
c Dr. A. I. Bhatti, CASPR SMC Workshop April 26, 2019 7 / 14
Higher order sliding mode
Real twisting algorithm I
Real twisting is historically a first
second-order sliding mode
controller
It is used to avoid chattering and
provides good robustness properties
This algorithm requires the
measurement of ṡ for the
implementation
State trajectories are twisted about
s = ṡ = 0 and converged in finite
time
This algorithm can be applied to
both relative degree 1 and 2 Figure 2: Convergence of Real Twisting
systems Algorithm
c Dr. A. I. Bhatti, CASPR SMC Workshop April 26, 2019 8 / 14
Higher order sliding mode
Real twisting algorithm II
Relative degree 1 system
The dynamic control input has the following form
(
−u, |u| > 1
u̇ = ,
−r1 (s) − r2 (ṡ), |u| ≤ 1
where r1 and r2 are controller gains.
If ṡ is not available then ṡ ≈ ∆s, ∆s = s(ti ) − s(ti − ti−1 )
c Dr. A. I. Bhatti, CASPR SMC Workshop April 26, 2019 9 / 14
Higher order sliding mode
Real twisting algorithm III
Relative degree 2 system
The controller takes the following form
(
−Km (s), |u| > 1
u̇ = ,
−KM (s), |u| ≤ 1
where Km and KM are the controller gains
c Dr. A. I. Bhatti, CASPR SMC Workshop April 26, 2019 10 / 14
Higher order sliding mode
Super twisting algorithm I
It is the only algorithm that steers
both s and ṡ to 0 manifold in finite
time
The error dynamics during sliding
mode also converges in finite time
(FOSM only yields asymptotic error
convergence)
It is not sensitive to sampling time
Ts
It does not require the derivative of
the sliding variable for
implementation Figure 3: Convergence of Super Twisting
Algorithm
c Dr. A. I. Bhatti, CASPR SMC Workshop April 26, 2019 11 / 14
Higher order sliding mode
Super twisting algorithm II
Sliding variable dynamics
Consider the second order system
ṡ1 = s2 ,
ṡ2 = a(t, x, u) + b(t, x, u)u̇.
After some transients
|a(t, x, u)| ≤ A > 0,
0 < Γm ≤ b(t, x, u) ≤ Γm .
c Dr. A. I. Bhatti, CASPR SMC Workshop April 26, 2019 12 / 14
Higher order sliding mode
Super twisting algorithm III
Control law
The total control input is given by
u = u1 + u2 ,
(
−u, |u| > 1
u̇1 = ,
−W (s), |u| ≤ 1
(
−λ|s0 |ρ (s), |s| > s0
u2 = ,
−λ|s|ρ (s), |s| ≤ s0
A 4AΓM (W + A) 1
where W > , λ2 ≥ and 0 < ρ ≤ .
Γm Γ3m (W − A) 2
If dynamics of the sliding variable is affine in control then s0 = ∞ and the
controller reduces to
Z t
u = −λ|s|ρ (s) − W (s)dt
0
c Dr. A. I. Bhatti, CASPR SMC Workshop April 26, 2019 13 / 14
Higher order sliding mode
Thank You
c Dr. A. I. Bhatti, CASPR SMC Workshop April 26, 2019 14 / 14