Lecture 1.
4
Tuning of PID and High Gain Controllers
Prof. Shyam Kamal
IIT (BHU) Varanasi
Purpose of the Discussion and Outcomes
Purpose of the Discussion
▶ Understand the tuning of PD and PID controllers
using Newton’s framework.
▶ Analyze the limitations of output feedback for
second-order systems. Explore the role of PD
controllers in achieving desired system performance.
▶ Investigate the concept of high-gain feedback and its
limitations.
Outcomes:
▶ Understanding of the tuning process for PD and PID
controllers using Newton’s framework.
▶ Understanding the connection between high-gain feedback
and sliding mode control.
Controller based on Information of Present and Future
Stabilization with PD-Controller when d ̸= 0 is constant
PD-Controller
▶ u := −kp v − kD v̇
▶ Closed loop system v̇ = −kp v − kD v̇ + d
Simplification:
kp d(t)
v̇ = − v+
1 + kD 1 + kD
▶ Observation: disturbance is scaled down.
▶ Analysis: same as proportional action.
Second order system with P-controller
System of Equations
▶ Given the system: ẋ1 = x2 , ẋ2 = u, u = −kx1 .
√
▶ Transformation: kx1 := r cos(θ), x2 := r sin(θ).
Step 1: Compute r˙
q
▶ Define r : r= kx12 + x22 .
▶ Differentiate: r˙ = 1
2 · 2kx
√1 ẋ1 +2x
2
2 ẋ2
2
.
kx1 +x2
▶ Substitute ẋ1 = x2 and ẋ2 = −kx1 :
r˙ = kx√1 x2 −kx
2
1 x2
2
= 0.
kx1 +x2
▶ Conclusion: r is constant over time.
Computation of θ̇ and Final Results
Step 2: Compute θ̇
▶ Transformation: tan(θ) = √x2 .
kx1
√ √
▶ Differentiate tan(θ): sec2 (θ)θ̇ = k ẋ2 x1 − kx2 ẋ1
kx12
.
▶ Substitute ẋ1 = x2 and ẋ2 = −kx1 :
√ √
2 k(−kx1 )x1 − kx2 x2
sec (θ)θ̇ = .
kx12
√ √
− k(kx12 +x22 )
▶ Simplify: θ̇ = = − k.
kx12 +x22
√
▶ r˙ = 0, r is constant over time. θ̇ = − k. The system
exhibits simple harmonic motion with constant angular
velocity.
Second-Order System (Continue)
Solution and Challenge
▶ Solution: PD controller. Challenge: How can a
noisy signal be differentiated effectively?
▶ PD Controller:
ẋ1 (t) = x2 (t), ẋ2 (t) = −k1 x1 (t) − k2 x2 (t).
▶ ẋ(t) = Ax(t); x(t) = [x1 (t), x2 (t)]⊤ .
▶ Now select k1 and k2 such that eig(A) is strictly
negative.
(
ẋ1 = x2 ,
Simulate: 1 1
ẋ2 = −k1 |x1 | 3 sign(x1 ) − k2 |x2 | 2 sign(x2 )
Controller based on Information of Present, Past and Future
Stabilization with PID-Controller when d ̸= 0 is constant
PID-Controller
▶ u := −kp v − kI 0t v (τ )dτ − kD v̇
R
Rt
▶ Closed loop system v̇ = −kp v − kI 0 v (τ )dτ − kD v̇ + d
Simplification:
Z t
kp kI d(t)
v̇ = − v− v (τ )dτ +
1 + kD 1 + kD 0 1 + kD
▶ Observation: disturbance is scaled down.
▶ Analysis: same as PI action.
Second-Order System
Simulate: Nonsmooth PID
(
ẋ1 (t) = x2 (t),
ẋ2 (t) = u(t) + d(t).
Z t
1/3
u = −k1 ⌊ϕ⌉ − k3 ⌊ϕ⌉0 dτ, where ⌊y ⌉p = |y |p sgn(y )
0
3/2
ϕ = x1 + k2 ⌊x2 ⌉
Simulate:
(
ẋ1 = x2 ,
1 1 Rt
ẋ2 = −k1 |x1 | 3 sign(x1 ) − k2 |x2 | 2 sign(x2 ) − k3 0 sign(x1 )dτ + d(t)
Controller Based on Information of Present with High Gain
High Gain Feedback
▶ Recall: v̇ = u + d.
▶ Control Law: u := − limτ →0 τ1 kp v ; kp > 0.
▶ Closed-Loop System: v̇ = − limτ →0 τ1 kp v + d.
Closed-Loop System Analysis
▶ As τ → 0:
lim (τ v̇ = −kp v + τ d) .
τ →0
▶ Result:
v ≈ 0.
Observations:
Discussion
▶ Able to compensate any kind of disturbances.
▶ Order of the system is reduced by one.
▶ Control has infinite magnitude, therefore
practically not feasible.
▶ Visualization:
1.5
1
v(t)
0.5
-0.5
0 2 4 6 8 10
Time (s)
High-Gain Feedback Control Law
Given System
▶ System Equations: ẋ1 = x2 , ẋ2 = u + d(t, x1 , x2 ).
▶ Control Law:
u := − limτ →0 τ1 (kp x1 + kd x2 ), kp , kd > 0.
Understanding the Control Input
▶ Interpretation: u = − τ1 (kp x1 + kd x2 ), where τ is
very small.
▶ Closed-Loop System:
ẋ1 = x2 , ẋ2 = − τ1 (kp x1 + kd x2 ) + d(t, x1 , x2 ).
Time-Scale Separation and Stability
Time-Scale Separation
▶ Fast Dynamics:
z = kp x1 + kd x2 , ż = kp ẋ1 − kτd z + kd d(t, x1 , x2 ).
▶ Quasi-Steady State Condition: kp x1 + kd x2 ≈ 0.
▶ Approximation: x2 ≈ − kkp x1 .
d
Stability Analysis
▶ First-Order System: ẋ1 ≈ − kkp x1 .
d
kp
▶ Solution: x1 (t) = x1 (0)e − kd t .
▶ Conclusion: Both x1 and x2 decay exponentially to
zero.
Interpretation
Interpretation of Behavior
▶ Extremely Fast Convergence: The term τ1 enforces
near-instantaneous decay of kp x1 + kd x2 .
▶ Quasi-Sliding Mode Behavior: The system is
driven to the manifold kp x1 + kd x2 = 0 very quickly.
Summary
▶ Fast Exponential Stability: The system exhibits
fast exponential stability.
▶ High-Precision Tracking: For practical
implementation, choosing a small τ achieves
high-precision tracking.
What is Next?
Key Insights
▶ High-gain feedback provides robustness but suffers
from issues like unbounded control magnitude and
sensitivity to noise.
▶ There is a need for controllers that retain the benefits
of high-gain feedback while ensuring finite control
magnitude.
▶ Such controllers can improve practical
implementation by avoiding actuator saturation and
reducing energy consumption.
Conclusions
Key Takeaways
▶ PD and PID Tuning: Newton’s framework provides
a systematic approach for tuning PD/PID controllers.
▶ Limitations of Output Feedback: For second-order
systems, output feedback alone may not suffice due
to its inability to capture full state information.
▶ Connection to Sliding Mode Control: High-gain
feedback shares similarities with sliding mode control,
particularly in achieving robustness against
disturbances.
Future Directions:
▶ Study the impact of actuator limitations on high-gain
feedback systems.
Thanks