2.153 Adaptive Control Fall 2019 Lecture 13: Adaptive PI...
Transcript of 2.153 Adaptive Control Fall 2019 Lecture 13: Adaptive PI...
2.153 Adaptive ControlFall 2019
Lecture 13: Adaptive PI Control
Anuradha Annaswamy
October 22, 2019
( [email protected]) October 22, 2019 1 / 13
Adaptive Control of a Second-order PlantGc(s)
1s(Js+B)
+ e τ
−
Plant: Jω + Bω = τ J > 0
PI Control: Gc(s) = kp +ki
sτ = kpe(t) + ki
∫e(τ)dτ
Adaptive PI Control: τ = kp(t)e(t) + ki(t)∫
e(τ)dτ
PID Control: Gc(s) = kp + kds + kis
τ = kpe(t) + ki∫
e(τ)dτ + kddedt
Adaptive PID Control: τ = kp(t)e(t) + ki(t)∫
e(τ)dτ + kd(t)e(t)
J and B are unknown. Adjust kp(t), ki(t) and kd(t) so that theclosed-loop system is stable and limt→∞ e(t) = 0.
( [email protected]) October 22, 2019 2 / 13
Adaptive Control of a Second-order PlantGc(s)
1s(Js+B)
+ e τ
−
Plant: Jω + Bω = τ J > 0
PI Control: Gc(s) = kp +ki
s
τ = kpe(t) + ki∫
e(τ)dτ
Adaptive PI Control: τ = kp(t)e(t) + ki(t)∫
e(τ)dτ
PID Control: Gc(s) = kp + kds + kis
τ = kpe(t) + ki∫
e(τ)dτ + kddedt
Adaptive PID Control: τ = kp(t)e(t) + ki(t)∫
e(τ)dτ + kd(t)e(t)
J and B are unknown. Adjust kp(t), ki(t) and kd(t) so that theclosed-loop system is stable and limt→∞ e(t) = 0.
( [email protected]) October 22, 2019 2 / 13
Adaptive Control of a Second-order PlantGc(s)
1s(Js+B)
+ e τ
−
Plant: Jω + Bω = τ J > 0
PI Control: Gc(s) = kp +ki
sτ = kpe(t) + ki
∫e(τ)dτ
Adaptive PI Control: τ = kp(t)e(t) + ki(t)∫
e(τ)dτ
PID Control: Gc(s) = kp + kds + kis
τ = kpe(t) + ki∫
e(τ)dτ + kddedt
Adaptive PID Control: τ = kp(t)e(t) + ki(t)∫
e(τ)dτ + kd(t)e(t)
J and B are unknown. Adjust kp(t), ki(t) and kd(t) so that theclosed-loop system is stable and limt→∞ e(t) = 0.
( [email protected]) October 22, 2019 2 / 13
Adaptive Control of a Second-order Plant
AdaptiveController
1s(Js+B)
+ e τ
−
Plant: Jω + Bω = τ J > 0
PI Control: Gc(s) = kp +ki
sτ = kpe(t) + ki
∫e(τ)dτ
Adaptive PI Control: τ = kp(t)e(t) + ki(t)∫
e(τ)dτ
PID Control: Gc(s) = kp + kds + kis
τ = kpe(t) + ki∫
e(τ)dτ + kddedt
Adaptive PID Control: τ = kp(t)e(t) + ki(t)∫
e(τ)dτ + kd(t)e(t)
J and B are unknown. Adjust kp(t), ki(t) and kd(t) so that theclosed-loop system is stable and limt→∞ e(t) = 0.
( [email protected]) October 22, 2019 2 / 13
Adaptive Control of a Second-order Plant
AdaptiveController
1s(Js+B)
+ e τ
−
Plant: Jω + Bω = τ J > 0
PI Control: Gc(s) = kp +ki
sτ = kpe(t) + ki
∫e(τ)dτ
Adaptive PI Control: τ = kp(t)e(t) + ki(t)∫
e(τ)dτ
PID Control: Gc(s) = kp + kds + kis
τ = kpe(t) + ki∫
e(τ)dτ + kddedt
Adaptive PID Control: τ = kp(t)e(t) + ki(t)∫
e(τ)dτ + kd(t)e(t)
J and B are unknown. Adjust kp(t), ki(t) and kd(t) so that theclosed-loop system is stable and limt→∞ e(t) = 0.
( [email protected]) October 22, 2019 2 / 13
Adaptive Control of a Second-order Plant
AdaptiveController
1s(Js+B)
+ e τ
−
Plant: Jω + Bω = τ J > 0
PI Control: Gc(s) = kp +ki
sτ = kpe(t) + ki
∫e(τ)dτ
Adaptive PI Control: τ = kp(t)e(t) + ki(t)∫
e(τ)dτ
PID Control: Gc(s) = kp + kds + kis
τ = kpe(t) + ki∫
e(τ)dτ + kddedt
Adaptive PID Control: τ = kp(t)e(t) + ki(t)∫
e(τ)dτ + kd(t)e(t)
J and B are unknown. Adjust kp(t), ki(t) and kd(t) so that theclosed-loop system is stable and limt→∞ e(t) = 0.
( [email protected]) October 22, 2019 2 / 13
Adaptive Control of a Second-order Plant
AdaptiveController
1s(Js+B)
+ e τ
−
Plant: Jω + Bω = τ J > 0
PI Control: Gc(s) = kp +ki
sτ = kpe(t) + ki
∫e(τ)dτ
Adaptive PI Control: τ = kp(t)e(t) + ki(t)∫
e(τ)dτ
PID Control: Gc(s) = kp + kds + kis
τ = kpe(t) + ki∫
e(τ)dτ + kddedt
Adaptive PID Control: τ = kp(t)e(t) + ki(t)∫
e(τ)dτ + kd(t)e(t)
J and B are unknown. Adjust kp(t), ki(t) and kd(t) so that theclosed-loop system is stable and limt→∞ e(t) = 0.
( [email protected]) October 22, 2019 2 / 13
Adaptive Control of a Second-order Plant
AdaptiveController
1s(Js+B)
+ e τ
−
Plant: Jω + Bω = τ J > 0
PI Control: Gc(s) = kp +ki
sτ = kpe(t) + ki
∫e(τ)dτ
Adaptive PI Control: τ = kp(t)e(t) + ki(t)∫
e(τ)dτ
PID Control: Gc(s) = kp + kds + kis
τ = kpe(t) + ki∫
e(τ)dτ + kddedt
Adaptive PID Control: τ = kp(t)e(t) + ki(t)∫
e(τ)dτ + kd(t)e(t)
J and B are unknown. Adjust kp(t), ki(t) and kd(t) so that theclosed-loop system is stable and limt→∞ e(t) = 0.
( [email protected]) October 22, 2019 2 / 13
PI -Control: Algebraic Part
NominalController
1Js+B
r + er τ x
−
Gc(s) = kp +ki
s
Parameterize kp = K > 0, ki = Kλ > 0
Closed-loop transfer function:K(s + λ)
s(Js + B) + K(s + λ)
=K(s + λ)
Js2 + s(B + K) + Kλ
Stable if K > |B|. Design the controller so that x→ xd
( [email protected]) October 22, 2019 3 / 13
PI -Control: Algebraic Part
NominalController
1Js+B
r + er τ x
−
Gc(s) = kp +ki
sParameterize kp = K > 0, ki = Kλ > 0
Closed-loop transfer function:K(s + λ)
s(Js + B) + K(s + λ)
=K(s + λ)
Js2 + s(B + K) + Kλ
Stable if K > |B|. Design the controller so that x→ xd
( [email protected]) October 22, 2019 3 / 13
PI -Control: Algebraic Part
NominalController
1Js+B
r + er τ x
−
Gc(s) = kp +ki
sParameterize kp = K > 0, ki = Kλ > 0
Closed-loop transfer function:K(s + λ)
s(Js + B) + K(s + λ)
=K(s + λ)
Js2 + s(B + K) + Kλ
Stable if K > |B|. Design the controller so that x→ xd
( [email protected]) October 22, 2019 3 / 13
PI -Control: Algebraic Part
NominalController
1Js+B
r + er τ x
−
Gc(s) = kp +ki
sParameterize kp = K > 0, ki = Kλ > 0
Closed-loop transfer function:K(s + λ)
s(Js + B) + K(s + λ)
=K(s + λ)
Js2 + s(B + K) + Kλ
Stable if K > |B|.
Design the controller so that x→ xd
( [email protected]) October 22, 2019 3 / 13
PI -Control: Algebraic Part
NominalController
1Js+B
r + er τ x
−
Gc(s) = kp +ki
sParameterize kp = K > 0, ki = Kλ > 0
Closed-loop transfer function:K(s + λ)
s(Js + B) + K(s + λ)
=K(s + λ)
Js2 + s(B + K) + Kλ
Stable if K > |B|. Design the controller so that x→ xd
( [email protected]) October 22, 2019 3 / 13
PI Control - Algebraic Part: Tracking
Gc(s) 1Js+B
r + er τ x
−
Wcl(s)
Wcl(s) =Gc(s)
Js + B + Gc(s)
W−1cl (s) = 1 + (Js + B)G−1
cl (s)
r = W−1cl (s)[xd]
= xd +((Js + B)G−1
cl (s))[xd]
= xd + (Js + B)[ωd] = xd + Bωd + Jωd
( [email protected]) October 22, 2019 4 / 13
PI Control - Algebraic Part: Tracking
Gc(s) 1Js+B
r + er τ x
−
Wcl(s)
Wcl(s) =Gc(s)
Js + B + Gc(s)
W−1cl (s) = 1 + (Js + B)G−1
cl (s)
r = W−1cl (s)[xd]
= xd +((Js + B)G−1
cl (s))[xd]
= xd + (Js + B)[ωd] = xd + Bωd + Jωd
( [email protected]) October 22, 2019 4 / 13
PI Control - Algebraic Part: Tracking
W−1cl (s) Gc(s) 1
Js+Bτxd r + er x
−
Wcl(s)
Wcl(s) =Gc(s)
Js + B + Gc(s)
W−1cl (s) = 1 + (Js + B)G−1
cl (s)
r = W−1cl (s)[xd]
= xd +((Js + B)G−1
cl (s))[xd]
= xd + (Js + B)[ωd] = xd + Bωd + Jωd
( [email protected]) October 22, 2019 4 / 13
PI Control - Algebraic Part: Tracking
W−1cl (s) Gc(s) 1
Js+Bτxd r + er x
−
Wcl(s)
Wcl(s) =Gc(s)
Js + B + Gc(s)
W−1cl (s) = 1 + (Js + B)G−1
cl (s)
r = W−1cl (s)[xd]
= xd +((Js + B)G−1
cl (s))[xd]
= xd + (Js + B)[ωd] = xd + Bωd + Jωd
( [email protected]) October 22, 2019 4 / 13
PI Control - Algebraic Part: Tracking
W−1cl (s) Gc(s) 1
Js+Bτxd r + er x
−
Wcl(s)
Wcl(s) =Gc(s)
Js + B + Gc(s)
W−1cl (s) = 1 + (Js + B)G−1
cl (s)
r = W−1cl (s)[xd]
= xd +((Js + B)G−1
cl (s))[xd]
= xd + (Js + B)[ωd]
= xd + Bωd + Jωd
( [email protected]) October 22, 2019 4 / 13
PI Control - Algebraic Part: Tracking
W−1cl (s) Gc(s) 1
Js+Bτxd r + er x
−
Wcl(s)
Wcl(s) =Gc(s)
Js + B + Gc(s)
W−1cl (s) = 1 + (Js + B)G−1
cl (s)
r = W−1cl (s)[xd]
= xd +((Js + B)G−1
cl (s))[xd]
= xd + (Js + B)[ωd] = xd + Bωd + Jωd
( [email protected]) October 22, 2019 4 / 13
PI Control - Algebraic Part: Tracking
Using r = Jωd + Bωd + xd the block diagram can be represented as
J
B Gc(s) 1Js+B
ωd
ωd
xd
+
++ r + er τ x
−
which can then be simplified to
Gc(s)
B
J
1Js+B
Plantxd +
xd
xd
+e + τ x
−
( [email protected]) October 22, 2019 5 / 13
PI Control - Algebraic Part: Tracking
Using r = Jωd + Bωd + xd the block diagram can be represented as
J
B Gc(s) 1Js+B
ωd
ωd
xd
+
++ r + er τ x
−
which can then be simplified to
Gc(s)
B
J
1Js+B
Plantxd +
xd
xd
+e + τ x
−
( [email protected]) October 22, 2019 5 / 13
PI Control - Algebraic Part: Tracking - Revised Design
J: reparametrize Gc(s):Gc(s) = kp +
ki
s= K +
Kλs
=Ks + Kλ
s
Change to Gc(s) =(K + Jλ)s + Kλ
s
Closed-loop transfer function:(K + Jλ)s + Kλ)
s(Js + B) + (K + Jλ)s + Kλ)
Move B from feedforward - to feedback
Closed-loop transfer function:(K + Jλ)s + Kλ
Js2 + (K + Jλ)s + Kλ
( [email protected]) October 22, 2019 6 / 13
PI Control - Algebraic Part: Tracking - Revised Design
J: reparametrize Gc(s):Gc(s) = kp +
ki
s= K +
Kλs
=Ks + Kλ
s
Change to Gc(s) =(K + Jλ)s + Kλ
s
Closed-loop transfer function:(K + Jλ)s + Kλ)
s(Js + B) + (K + Jλ)s + Kλ)
Move B from feedforward - to feedback
Closed-loop transfer function:(K + Jλ)s + Kλ
Js2 + (K + Jλ)s + Kλ
( [email protected]) October 22, 2019 6 / 13
PI Control - Algebraic Part: Tracking - Revised Design
J: reparametrize Gc(s):Gc(s) = kp +
ki
s= K +
Kλs
=Ks + Kλ
s
Change to Gc(s) =(K + Jλ)s + Kλ
s
Closed-loop transfer function:(K + Jλ)s + Kλ)
s(Js + B) + (K + Jλ)s + Kλ)
Move B from feedforward - to feedback
Closed-loop transfer function:(K + Jλ)s + Kλ
Js2 + (K + Jλ)s + Kλ
( [email protected]) October 22, 2019 6 / 13
PI Control - Algebraic Part: Tracking - Revised Design
J: reparametrize Gc(s):Gc(s) = kp +
ki
s= K +
Kλs
=Ks + Kλ
s
Change to Gc(s) =(K + Jλ)s + Kλ
s
Closed-loop transfer function:(K + Jλ)s + Kλ)
s(Js + B) + (K + Jλ)s + Kλ)
Move B from feedforward - to feedback
Closed-loop transfer function:(K + Jλ)s + Kλ
Js2 + (K + Jλ)s + Kλ
( [email protected]) October 22, 2019 6 / 13
PI Control - Algebraic Part: Tracking - Revised Design
Gc(s)
B
J
1Js+B
Plantxd +
xd
xd
+e + τ x
−
Gc(s) =(K + Jλ)s + Kλ
s
( [email protected]) October 22, 2019 7 / 13
PI Control - Algebraic Part: Tracking - Revised Design
J
Gc(s) 1Js+B
B
xd +
xd
+e + τ x
+−
Gc(s) =(K + Jλ)s + Kλ
s
( [email protected]) October 22, 2019 7 / 13
PI Control - Algebraic Part: Tracking - Revised Design
(K+Jλ)s+Kλs
1Js+B
B
r + er + τ x
+−
Wcl(s) :(K + Jλ)s + Kλ
Js2 + (K + Jλ)s + Kλ
ωn =
√KλJ
ζ =K + Jλ
2J·√
JKλ
=K + Jλ2√
JKλ
Less sensitive to uncertainties in JLess sensitive to uncertainties in B
( [email protected]) October 22, 2019 8 / 13
PI Control - Algebraic Part: Tracking - Revised Design
(K+Jλ)s+Kλs
1Js+B
B
r + er + τ x
+−
Wcl(s) :(K + Jλ)s + Kλ
Js2 + (K + Jλ)s + Kλ
ωn =
√KλJ
ζ =K + Jλ
2J·√
JKλ
=K + Jλ2√
JKλ
Less sensitive to uncertainties in JLess sensitive to uncertainties in B
( [email protected]) October 22, 2019 8 / 13
PI Control - Algebraic Part: Tracking - Revised Design
(K+Jλ)s+Kλs
1Js+B
B
r + er + τ x
+−
Wcl(s) :(K + Jλ)s + Kλ
Js2 + (K + Jλ)s + Kλ
ωn =
√KλJ
ζ =K + Jλ
2J·√
JKλ
=K + Jλ2√
JKλ
Less sensitive to uncertainties in JLess sensitive to uncertainties in B
( [email protected]) October 22, 2019 8 / 13
PI Control - Algebraic Part: Tracking - CompleteDesign
J
Gc(s)1
Js+B
B
xd +
xd
+e + τ x
+−
Gc(s) =(K + Jλ)s + Kλ
s,
Wcl(s) =(K + Jλ)s + Kλ
Js2 + (K + Jλ)s + Kλ
r = W−1cl (s)[xd]
= xd +((Js)G−1
cl (s))[xd]
= xd + Jωd
τ = Bx + Jxd + Gc(s)[e]
= Bx + Jxd + (K + Jλ)e + Kλ∫
e(τ)dτ
= J (xd + λe) + Bx + K(
e + λ
∫e(τ)dτ
)= θ∗Tφ(t)
( [email protected]) October 22, 2019 9 / 13
PI Control - Algebraic Part: Tracking - CompleteDesign
J
Gc(s)1
Js+B
B
xd +
xd
+e + τ x
+−
Gc(s) =(K + Jλ)s + Kλ
s,Wcl(s) =
(K + Jλ)s + KλJs2 + (K + Jλ)s + Kλ
r = W−1cl (s)[xd]
= xd +((Js)G−1
cl (s))[xd]
= xd + Jωd
τ = Bx + Jxd + Gc(s)[e]
= Bx + Jxd + (K + Jλ)e + Kλ∫
e(τ)dτ
= J (xd + λe) + Bx + K(
e + λ
∫e(τ)dτ
)= θ∗Tφ(t)
( [email protected]) October 22, 2019 9 / 13
PI Control - Algebraic Part: Tracking - CompleteDesign
J
Gc(s)1
Js+B
B
xd +
xd
+e + τ x
+−
Gc(s) =(K + Jλ)s + Kλ
s,Wcl(s) =
(K + Jλ)s + KλJs2 + (K + Jλ)s + Kλ
r = W−1cl (s)[xd]
= xd +((Js)G−1
cl (s))[xd]
= xd + Jωd
τ = Bx + Jxd + Gc(s)[e]
= Bx + Jxd + (K + Jλ)e + Kλ∫
e(τ)dτ
= J (xd + λe) + Bx + K(
e + λ
∫e(τ)dτ
)= θ∗Tφ(t)
( [email protected]) October 22, 2019 9 / 13
PI Control - Algebraic Part: Tracking - CompleteDesign
J
Gc(s)1
Js+B
B
xd +
xd
+e + τ x
+−
Gc(s) =(K + Jλ)s + Kλ
s,Wcl(s) =
(K + Jλ)s + KλJs2 + (K + Jλ)s + Kλ
r = W−1cl (s)[xd]
= xd +((Js)G−1
cl (s))[xd]
= xd + Jωd
τ = Bx + Jxd + Gc(s)[e]
= Bx + Jxd + (K + Jλ)e + Kλ∫
e(τ)dτ
= J (xd + λe) + Bx + K(
e + λ
∫e(τ)dτ
)= θ∗Tφ(t)
( [email protected]) October 22, 2019 9 / 13
PI Control - Algebraic Part: Tracking - CompleteDesign
J
Gc(s)1
Js+B
B
xd +
xd
+e + τ x
+−
Gc(s) =(K + Jλ)s + Kλ
s,Wcl(s) =
(K + Jλ)s + KλJs2 + (K + Jλ)s + Kλ
r = W−1cl (s)[xd]
= xd +((Js)G−1
cl (s))[xd]
= xd + Jωd
τ = Bx + Jxd + Gc(s)[e]
= Bx + Jxd + (K + Jλ)e + Kλ∫
e(τ)dτ
= J (xd + λe) + Bx + K(
e + λ
∫e(τ)dτ
)= θ∗Tφ(t)
( [email protected]) October 22, 2019 9 / 13
PI Control - Algebraic Part: Tracking - CompleteDesign
J
Gc(s)1
Js+B
B
xd +
xd
+e + τ x
+−
Gc(s) =(K + Jλ)s + Kλ
s,Wcl(s) =
(K + Jλ)s + KλJs2 + (K + Jλ)s + Kλ
r = W−1cl (s)[xd]
= xd +((Js)G−1
cl (s))[xd]
= xd + Jωd
τ = Bx + Jxd + Gc(s)[e]
= Bx + Jxd + (K + Jλ)e + Kλ∫
e(τ)dτ
= J (xd + λe) + Bx + K(
e + λ
∫e(τ)dτ
)= θ∗Tφ(t)
( [email protected]) October 22, 2019 9 / 13
PI Control - Algebraic Part: Tracking - CompleteDesign
J
Gc(s)1
Js+B
B
xd +
xd
+e + τ x
+−
Gc(s) =(K + Jλ)s + Kλ
s,Wcl(s) =
(K + Jλ)s + KλJs2 + (K + Jλ)s + Kλ
r = W−1cl (s)[xd]
= xd +((Js)G−1
cl (s))[xd]
= xd + Jωd
τ = Bx + Jxd + Gc(s)[e]
= Bx + Jxd + (K + Jλ)e + Kλ∫
e(τ)dτ
= J (xd + λe) + Bx + K(
e + λ
∫e(τ)dτ
)= θ∗Tφ(t)
( [email protected]) October 22, 2019 9 / 13
PI Control - Algebraic Part: Tracking - CompleteDesign
J
Gc(s)1
Js+B
B
xd +
xd
+e + τ x
+−
Gc(s) =(K + Jλ)s + Kλ
s,Wcl(s) =
(K + Jλ)s + KλJs2 + (K + Jλ)s + Kλ
r = W−1cl (s)[xd]
= xd +((Js)G−1
cl (s))[xd]
= xd + Jωd
τ = Bx + Jxd + Gc(s)[e]
= Bx + Jxd + (K + Jλ)e + Kλ∫
e(τ)dτ
= J (xd + λe) + Bx + K(
e + λ
∫e(τ)dτ
)= θ∗Tφ(t)
( [email protected]) October 22, 2019 9 / 13
PI Control - Algebraic Part: Tracking - CompleteDesign
J
K
B
1Js+B
xd
xd
e1
e2++
+
τ x
τ = Bx + Jxd + Gc(s)[e]
= Je1(t) + Bx(t) + Ke2(t) = θ∗Tφ(t)
e1 = (xd + λe) , e2 =
(e + λ
∫e(τ)dτ
)
φ = [ e1 x e2 ]> , θ∗ = [ J B K ]>
Adaptive PI control:
τ = J(t)e1 + B(t)x + Ke2
( [email protected]) October 22, 2019 10 / 13
PI Control - Algebraic Part: Tracking - CompleteDesign
J
K
B
1Js+B
xd
xd
e1
e2++
+
τ x
τ = Bx + Jxd + Gc(s)[e]
= Je1(t) + Bx(t) + Ke2(t) = θ∗Tφ(t)
e1 = (xd + λe) , e2 =
(e + λ
∫e(τ)dτ
)
φ = [ e1 x e2 ]> , θ∗ = [ J B K ]>
Adaptive PI control:
τ = J(t)e1 + B(t)x + Ke2
( [email protected]) October 22, 2019 10 / 13
PI Control - Algebraic Part: Tracking - CompleteDesign
J
K
B
1Js+B
xd
xd
e1
e2++
+
τ x
τ = Bx + Jxd + Gc(s)[e]
= Je1(t) + Bx(t) + Ke2(t) = θ∗Tφ(t)
e1 = (xd + λe) ,
e2 =
(e + λ
∫e(τ)dτ
)
φ = [ e1 x e2 ]> , θ∗ = [ J B K ]>
Adaptive PI control:
τ = J(t)e1 + B(t)x + Ke2
( [email protected]) October 22, 2019 10 / 13
PI Control - Algebraic Part: Tracking - CompleteDesign
J
K
B
1Js+B
xd
xd
e1
e2++
+
τ x
τ = Bx + Jxd + Gc(s)[e]
= Je1(t) + Bx(t) + Ke2(t) = θ∗Tφ(t)
e1 = (xd + λe) , e2 =
(e + λ
∫e(τ)dτ
)
φ = [ e1 x e2 ]> , θ∗ = [ J B K ]>
Adaptive PI control:
τ = J(t)e1 + B(t)x + Ke2
( [email protected]) October 22, 2019 10 / 13
PI Control - Algebraic Part: Tracking - CompleteDesign
J
K
B
1Js+B
xd
xd
e1
e2++
+
τ x
τ = Bx + Jxd + Gc(s)[e]
= Je1(t) + Bx(t) + Ke2(t) = θ∗Tφ(t)
e1 = (xd + λe) , e2 =
(e + λ
∫e(τ)dτ
)
φ = [ e1 x e2 ]> ,
θ∗ = [ J B K ]>
Adaptive PI control:
τ = J(t)e1 + B(t)x + Ke2
( [email protected]) October 22, 2019 10 / 13
PI Control - Algebraic Part: Tracking - CompleteDesign
J
K
B
1Js+B
xd
xd
e1
e2++
+
τ x
τ = Bx + Jxd + Gc(s)[e]
= Je1(t) + Bx(t) + Ke2(t) = θ∗Tφ(t)
e1 = (xd + λe) , e2 =
(e + λ
∫e(τ)dτ
)
φ = [ e1 x e2 ]> , θ∗ = [ J B K ]>
Adaptive PI control:
τ = J(t)e1 + B(t)x + Ke2
( [email protected]) October 22, 2019 10 / 13
PI Control - Algebraic Part: Tracking - CompleteDesign
J
K
B
1Js+B
xd
xd
e1
e2++
+
τ x
τ = Bx + Jxd + Gc(s)[e]
= Je1(t) + Bx(t) + Ke2(t) = θ∗Tφ(t)
e1 = (xd + λe) , e2 =
(e + λ
∫e(τ)dτ
)
φ = [ e1 x e2 ]> , θ∗ = [ J B K ]>
Adaptive PI control:
τ = J(t)e1 + B(t)x + Ke2
( [email protected]) October 22, 2019 10 / 13
Adaptive PI Control
τ = J(t)e1 + B(t)x + Ke2
Plant+controller: x =1J(−Bx + τ)
=1J
(−Bx + J(t)e1 + B(t)x + Ke2
)e2 =
(e + λ
∫e(τ)dτ
)e2 = e + λe = xd − x + λe
= xd −1J
(Je1 + Bx + Ke2
)+ λe
=
(1− J
J
)e1 −
1J
(Bx + Ke2
)= −K
Je2 +
1J
(−Je1 − Bx
)− Error Model 3
( [email protected]) October 22, 2019 11 / 13
Adaptive PI Control
τ = J(t)e1 + B(t)x + Ke2
Plant+controller: x =1J(−Bx + τ)
=1J
(−Bx + J(t)e1 + B(t)x + Ke2
)
e2 =
(e + λ
∫e(τ)dτ
)e2 = e + λe = xd − x + λe
= xd −1J
(Je1 + Bx + Ke2
)+ λe
=
(1− J
J
)e1 −
1J
(Bx + Ke2
)= −K
Je2 +
1J
(−Je1 − Bx
)− Error Model 3
( [email protected]) October 22, 2019 11 / 13
Adaptive PI Control
τ = J(t)e1 + B(t)x + Ke2
Plant+controller: x =1J(−Bx + τ)
=1J
(−Bx + J(t)e1 + B(t)x + Ke2
)e2 =
(e + λ
∫e(τ)dτ
)
e2 = e + λe = xd − x + λe
= xd −1J
(Je1 + Bx + Ke2
)+ λe
=
(1− J
J
)e1 −
1J
(Bx + Ke2
)= −K
Je2 +
1J
(−Je1 − Bx
)− Error Model 3
( [email protected]) October 22, 2019 11 / 13
Adaptive PI Control
τ = J(t)e1 + B(t)x + Ke2
Plant+controller: x =1J(−Bx + τ)
=1J
(−Bx + J(t)e1 + B(t)x + Ke2
)e2 =
(e + λ
∫e(τ)dτ
)e2 = e + λe = xd − x + λe
= xd −1J
(Je1 + Bx + Ke2
)+ λe
=
(1− J
J
)e1 −
1J
(Bx + Ke2
)= −K
Je2 +
1J
(−Je1 − Bx
)− Error Model 3
( [email protected]) October 22, 2019 11 / 13
Adaptive PI Control
τ = J(t)e1 + B(t)x + Ke2
Plant+controller: x =1J(−Bx + τ)
=1J
(−Bx + J(t)e1 + B(t)x + Ke2
)e2 =
(e + λ
∫e(τ)dτ
)e2 = e + λe = xd − x + λe
= xd −1J
(Je1 + Bx + Ke2
)+ λe
=
(1− J
J
)e1 −
1J
(Bx + Ke2
)= −K
Je2 +
1J
(−Je1 − Bx
)− Error Model 3
( [email protected]) October 22, 2019 11 / 13
Adaptive PI Control
τ = J(t)e1 + B(t)x + Ke2
Plant+controller: x =1J(−Bx + τ)
=1J
(−Bx + J(t)e1 + B(t)x + Ke2
)e2 =
(e + λ
∫e(τ)dτ
)e2 = e + λe = xd − x + λe
= xd −1J
(Je1 + Bx + Ke2
)+ λe
=
(1− J
J
)e1 −
1J
(Bx + Ke2
)
= −KJ
e2 +1J
(−Je1 − Bx
)− Error Model 3
( [email protected]) October 22, 2019 11 / 13
Adaptive PI Control
τ = J(t)e1 + B(t)x + Ke2
Plant+controller: x =1J(−Bx + τ)
=1J
(−Bx + J(t)e1 + B(t)x + Ke2
)e2 =
(e + λ
∫e(τ)dτ
)e2 = e + λe = xd − x + λe
= xd −1J
(Je1 + Bx + Ke2
)+ λe
=
(1− J
J
)e1 −
1J
(Bx + Ke2
)= −K
Je2 +
1J
(−Je1 − Bx
)
− Error Model 3
( [email protected]) October 22, 2019 11 / 13
Adaptive PI Control
τ = J(t)e1 + B(t)x + Ke2
Plant+controller: x =1J(−Bx + τ)
=1J
(−Bx + J(t)e1 + B(t)x + Ke2
)e2 =
(e + λ
∫e(τ)dτ
)e2 = e + λe = xd − x + λe
= xd −1J
(Je1 + Bx + Ke2
)+ λe
=
(1− J
J
)e1 −
1J
(Bx + Ke2
)= −K
Je2 +
1J
(−Je1 − Bx
)− Error Model 3
( [email protected]) October 22, 2019 11 / 13
Adaptive PI Control - Stability
J
K
B
1Js+B
xd
xd
e1
e2++
+
τ x
τ = J(t)e1 + B(t)x + Ke2
Error Equation: e2 = −KJ
e2 +1J
(−Je1 − Bx
)Adaptive Law: ˙J = γ1e2e1,
˙B = γ1e2x
Lyapunov function: V =12
(e2
2 +1J
(J2
γ1+
B2
γ2
))V = −K
Je2
2
( [email protected]) October 22, 2019 12 / 13
Adaptive PI Control - Stability
J
K
B
1Js+B
xd
xd
e1
e2++
+
τ x
τ = J(t)e1 + B(t)x + Ke2
Error Equation: e2 = −KJ
e2 +1J
(−Je1 − Bx
)
Adaptive Law: ˙J = γ1e2e1,˙B = γ1e2x
Lyapunov function: V =12
(e2
2 +1J
(J2
γ1+
B2
γ2
))V = −K
Je2
2
( [email protected]) October 22, 2019 12 / 13
Adaptive PI Control - Stability
J
K
B
1Js+B
xd
xd
e1
e2++
+
τ x
τ = J(t)e1 + B(t)x + Ke2
Error Equation: e2 = −KJ
e2 +1J
(−Je1 − Bx
)Adaptive Law: ˙J = γ1e2e1,
˙B = γ1e2x
Lyapunov function: V =12
(e2
2 +1J
(J2
γ1+
B2
γ2
))V = −K
Je2
2
( [email protected]) October 22, 2019 12 / 13
Adaptive PI Control - Stability
J
K
B
1Js+B
xd
xd
e1
e2++
+
τ x
τ = J(t)e1 + B(t)x + Ke2
Error Equation: e2 = −KJ
e2 +1J
(−Je1 − Bx
)Adaptive Law: ˙J = γ1e2e1,
˙B = γ1e2x
Lyapunov function: V =12
(e2
2 +1J
(J2
γ1+
B2
γ2
))V = −K
Je2
2
( [email protected]) October 22, 2019 12 / 13
Adaptive PI Control - Stability
J
K
B
1Js+B
xd
xd
e1
e2++
+
τ x
τ = J(t)e1 + B(t)x + Ke2
Error Equation: e2 = −KJ
e2 +1J
(−Je1 − Bx
)Adaptive Law: ˙J = γ1e2e1,
˙B = γ1e2x
Lyapunov function: V =12
(e2
2 +1J
(J2
γ1+
B2
γ2
))V = −K
Je2
2( [email protected]) October 22, 2019 12 / 13
Adaptive PI Control - Stability and Asymptotic Tracking
J
K
B
1Js+B
xd
xd
e1
e2++
+
τ x
e2, J, B ∈ L∞
=⇒ e2 ∈ L∞e2 ∈ L∞ =⇒ e ∈ L∞e ∈ L∞ =⇒ e1 ∈ L∞Therefore e2 ∈ L∞; e2 ∈ L2
=⇒ limt→∞ e2(t) = 0 =⇒ limt→∞ e(t) = 0
( [email protected]) October 22, 2019 13 / 13
Adaptive PI Control - Stability and Asymptotic Tracking
J
K
B
1Js+B
xd
xd
e1
e2++
+
τ x
e2, J, B ∈ L∞ =⇒ e2 ∈ L∞
e2 ∈ L∞ =⇒ e ∈ L∞e ∈ L∞ =⇒ e1 ∈ L∞Therefore e2 ∈ L∞; e2 ∈ L2
=⇒ limt→∞ e2(t) = 0 =⇒ limt→∞ e(t) = 0
( [email protected]) October 22, 2019 13 / 13
Adaptive PI Control - Stability and Asymptotic Tracking
J
K
B
1Js+B
xd
xd
e1
e2++
+
τ x
e2, J, B ∈ L∞ =⇒ e2 ∈ L∞e2 ∈ L∞ =⇒ e ∈ L∞
e ∈ L∞ =⇒ e1 ∈ L∞Therefore e2 ∈ L∞; e2 ∈ L2
=⇒ limt→∞ e2(t) = 0 =⇒ limt→∞ e(t) = 0
( [email protected]) October 22, 2019 13 / 13
Adaptive PI Control - Stability and Asymptotic Tracking
J
K
B
1Js+B
xd
xd
e1
e2++
+
τ x
e2, J, B ∈ L∞ =⇒ e2 ∈ L∞e2 ∈ L∞ =⇒ e ∈ L∞e ∈ L∞ =⇒ e1 ∈ L∞
Therefore e2 ∈ L∞; e2 ∈ L2
=⇒ limt→∞ e2(t) = 0 =⇒ limt→∞ e(t) = 0
( [email protected]) October 22, 2019 13 / 13
Adaptive PI Control - Stability and Asymptotic Tracking
J
K
B
1Js+B
xd
xd
e1
e2++
+
τ x
e2, J, B ∈ L∞ =⇒ e2 ∈ L∞e2 ∈ L∞ =⇒ e ∈ L∞e ∈ L∞ =⇒ e1 ∈ L∞Therefore e2 ∈ L∞; e2 ∈ L2
=⇒ limt→∞ e2(t) = 0 =⇒ limt→∞ e(t) = 0
( [email protected]) October 22, 2019 13 / 13
Adaptive PI Control - Stability and Asymptotic Tracking
J
K
B
1Js+B
xd
xd
e1
e2++
+
τ x
e2, J, B ∈ L∞ =⇒ e2 ∈ L∞e2 ∈ L∞ =⇒ e ∈ L∞e ∈ L∞ =⇒ e1 ∈ L∞Therefore e2 ∈ L∞; e2 ∈ L2
=⇒ limt→∞ e2(t) = 0
=⇒ limt→∞ e(t) = 0
( [email protected]) October 22, 2019 13 / 13
Adaptive PI Control - Stability and Asymptotic Tracking
J
K
B
1Js+B
xd
xd
e1
e2++
+
τ x
e2, J, B ∈ L∞ =⇒ e2 ∈ L∞e2 ∈ L∞ =⇒ e ∈ L∞e ∈ L∞ =⇒ e1 ∈ L∞Therefore e2 ∈ L∞; e2 ∈ L2
=⇒ limt→∞ e2(t) = 0 =⇒ limt→∞ e(t) = 0
( [email protected]) October 22, 2019 13 / 13