Guidance and Control of Unmanned Autonomous Vehicles: Shashi Ranjan Kumar

Guidance and Control of Unmanned Autonomous Vehicles
Shashi Ranjan Kumar
Assistant Professor
Department of Aerospace Engineering
Indian Institute of Technology Bombay
Powai, Mumbai- 400076
https://www.aero.iitb.ac.in/˜shashi
srk@aero.iitb.ac.in
UAV Autopilot: Saturation Constraints and Performance
• Successive loop closure: Performance limitation of the system due to the

inner-most loop
• What are other reasons for performance degradation?

⇒ Saturation constraints, such as bounded angular deflections
⇒ Physical bounds on ailerons restrict roll rate
⇒ Bound on total lateral acceleration due to aircraft stall at higher angle of

attack
• Goal of autopilot design: To design bandwidth of the inner loop to be as

large as possible, without violating the saturation constraints
• It allows outer loops to ensure sufficient bandwidth separation of the

successive loops
Shashi Ranjan Kumar (IITB) Guidance and Control of Unmanned Autonomous Vehicles Autopilot Design Lecture 10 1 / 25
• Consider second order system

b0
P (s) =
s2 + a1 s + a0
• Closed-loop transfer function
y(s) b 0 kp
= 2
y c (s) s + (a1 + b0 kd )s + (a0 + b0 kp )
• Closed-loop poles: depends on gains
• Actuator effort: u = kp e − kd ẏ
• What terms affect the input demand or actuator effort?
• For ẏ ≈ 0, size of error e and kp decide about input u; u ≈ kp e
• For stable system, largest control effort occurs immediately after step input
• Determination of proportional gain
umax
umax = kp emax =⇒ kp =
emax
• Can we achieve any bandwidth at our choice? No

• How to compute the achievable bandwidth of a closed-loop system?
• Canonical form of second order system, with no zeros
y(s) ωn2
= ,
y c (s) s2 + 2ζωn s + ωn2
where ζ is damping ratio and ωn is natural frequency.
• Under-damped system
p
0 ≤ ζ < 1; λ = −ζωn ± jωn 1 − ζ2
• Natural frequency of closed-loop system under saturation constraints

r
p umax
ωn = a0 + b0 kp = a0 + b0
emax
• Upper limit on the bandwidth to avoid possible saturation on input
UAV Autopilot: Lateral directional Autopilot
• How to design lateral autopilot?
• Roll rate damping: kdϕ

• Roll attitude: kpϕ and kiϕ
• Course angle: kpχ and kiχ
• How to select these gains?
• Successive selection of gains beginning with the inner loop and working
outward.
UAV Lateral Autopilot: Roll Attitude Loop Design
• Linear design model:

aϕ2 1
ϕ(s) = δa (s)+ dϕ (s)
s(s + aϕ1 ) aϕ2 2
• Goal of inner loop: To control roll angle and its rate

• Gain design: Based on desired response of closed-loop system
kpϕ aϕ2
Hϕ/ϕc (s) =
s2 + (aϕ1 + aϕ2 kdϕ )s + kpϕ aϕ2
UAV Lateral Autopilot: Roll Attitude Loop Design
• Desired second order closed-loop transfer function
ϕc (s) ωn2 ϕ
= 2
ϕ(s) s + 2ζϕ ωnϕ s + ωn2 ϕ
• How to find this desired closed-loop transfer function?

• Equating coefficients results in
ωn2 ϕ = kpϕ aϕ2 , 2ζϕ ωnϕ = aϕ1 + aϕ2 kdϕ
• Gain kpϕ is chosen as per saturation bound on aileron deflections

r
δamax δamax
kp ϕ = sign(aϕ2 ) =⇒ ωnϕ = |aϕ2 |
eϕmax eϕmax
• Gain kdϕ based on gain kpϕ , for chosen ζϕ to satisfy other specifications
2ζϕ ωnϕ − aϕ1

kd ϕ =
aϕ2
UAV Lateral Autopilot: Roll Attitude Loop Design Issues
• Zero steady-state tracking error in roll without an integrator (Type-1 open

loop TF)
• Recall final value theorem and system type
• Presence of disturbance at the summing junction
• Neglected dynamics, wind gust, or turbulence

aϕ2 1
ϕ(s) = δa (s)+ dϕ (s)
s(s + aϕ1 ) aϕ2 2
UAV Lateral Autopilot: Roll Attitude Loop Design Issues
• On solving for ϕ(s)

1
ϕ(s) = d ϕ2

aϕ2 kdϕ
+ ϕc
• What would be the steady-state error for constant step disturbance

dϕ2 = A/s?
• Using final-value theorem, steady-state error is A/(aϕ2 kpϕ ), a nonzero
value.
• For constant orbit, p, q, r and thus dϕ2 are constant.
• How to remove this steady-state error?
• To eliminate steady-state error, an integrator is required.
• How to see if an integrator reject disturbances?
UAV Lateral Autopilot: Roll Attitude Loop Design using Integrator

s
ϕ= d ϕ2
s3 + (aϕ1 + aϕ2 kdϕ )s2 + kpϕ aϕ2 s + aϕ2 kiϕ
 ki

ϕ
aϕ2 kpϕ s + kpϕ
+  ϕc
s3 + (aϕ1 + aϕ2 kdϕ )s2 + kpϕ aϕ2 s + aϕ2 kiϕ
UAV Lateral Autopilot: Roll Attitude Loop Design using Integrator
A
• Steady-state error: Zero for constant dϕ2 , aϕ2 ki
for ramp inputs
ϕ
• kiϕ can be chosen using root-locus technique
• Characteristic equation of closed-loop system
s3 + (aϕ1 + aϕ2 kdϕ )s2 + kpϕ aϕ2 s + aϕ2 kiϕ = 0
• How to draw root locus?

!
aϕ2
1 + ki ϕ =0
s s2 + (aϕ1 + aϕ2 kdϕ )s + kpϕ aϕ2
ki ϕ c
δa = kpϕ (ϕc − ϕ) + (ϕ − ϕ) − kdϕ p
s
Proper gain for better performance
UAV Autopilot: Course Hold
What next? Course hold outer loop
UAV Lateral Autopilot: Course Hold
• With proper tuning of gains, Hϕ/ϕc ≈ 1 over frequencies between 0 to ωnϕ

• Objective: Design kpχ and kiχ s.t. χ → χc asymptotically
• Transfer functions relating χ, χc , and dχ
g g g
s kpχ s + kix
Vg Vg Vg c
χ= g g dχ + 2 g g χ
2
s + kpχ s + kiχ s + kpχ s + kiχ
Vg Vg Vg Vg
• Using final value theorem, with constant dχ and χc , χ → χc

• Desired transfer function from χc to χ
2ζχ ωnχ s + ωn2 χ

Hχ =
s2 + 2ζχ ωnχ s + ωn2 χ
• Gains selection: Based on desired ζχ and ωnχ

• Selection of ωnχ and ζχ
UAV Autopilot: Course Hold
• Due to zero in numerator, standard intuition for selection of ζ does not

hold.
• Larger ζ results in larger bandwidth and smaller overshoot.
UAV Lateral Autopilot: Course Hold
• On comparing coefficients, we get
ωn2 χ Vg
ωn2 χ = kiχ g/Vg =⇒ kiχ =
g
2ζχ ωnχ Vg
2ζχ ωnχ = kpχ g/Vg =⇒ kpχ =
g
• To ensure sufficient bandwidth

ωnϕ
ωnχ = , Wχ ≥ 5
Wχ

• Large bandwidth separation is better How can this be ensured?
⇒ Slower response of χ-loop, lower ωnχ
⇒ Faster response of ϕ-loop, higher ωnϕ
• More control demand for a faster response
ki χ c
ϕc = kpχ (χc − χ) + (χ − χ)
s
UAV Lateral Autopilot: Sideslip Hold
• Can we maintain constant or zero sideslip angle?

• Yes, using rudder inputs
• Transfer function from β c to β

aβ2 (kpβ s + kiβ )
Hβ/β c (s) = , Hβ/β c (0) = 1
s2 + (aβ1 + aβ2 kpβ )s + aβ2 kiβ
• Characteristic equation for desired closed-loop poles
s2 + 2ζβ ωnβ s + ωn2 β = 0
UAV Lateral Autopilot: Sideslip Hold
• On equating the coefficients of characteristic equations
2ζβ ωnβ = aβ1 + aβ2 kpβ , ωn2 β = aβ2 kiβ
• Maximum error in sideslip (emax

β ) and maximum allowable rudder
deflection (δrmax )
• Gain with such constraints
δrmax
kpβ = sign(aβ2 )
emax
β
2
ωn2 β

1 aβ1 + aβ2 kpβ
ki β = =
aβ2 aβ2 2ζβ
• Output of sideslip loop (rudder deflection)
kiβ
δr = −kpβ β − β
s
UAV Longitudinal Autopilot
• Complicated longitudinal autopilot as compared to lateral one

• Role of airspeed in the longitudinal dynamics
• Objective: Regulate airspeed and altitude using throttle and elevator
inputs
• Regulation of altitude and airspeed: Role of altitude error
• Take-off zone: full throttle is commanded and θ → θc using the elevator.
• Climb zone: to maximize the climb rate
⇒ To maximize the climb rate, full throttle is commanded and airspeed is

regulated using pitch angle.
⇒ If airspeed increases above its nominal value, then aircraft is caused to pitch
up, leading to increase in climb rate and decrease in airspeed.
⇒ Similarly, if the airspeed drops below nominal value, aircraft is pitched

down, resulting in increase of airspeed but also decreases climb rate.
⇒ Regulating airspeed using pitch attitude effectively avoids stall conditions.
⇒ No regulation of airspeed with pitch altitude immediately after take-off
• Descend zone: Similar to climb zone except zero throttle
⇒ Avoidance of stall conditions by regulating airspeed using pitch angle
⇒ Maximization of descent rate at a given airspeed
• Altitude hold zone
⇒ Airspeed: by adjusting throttle
⇒ Altitude: by commanding pitch attitude.
• Longitudinal autopilot: Necessary feedback loops
⇒ Pitch attitude hold using elevator
⇒ Airspeed hold using throttle
⇒ Airspeed hold using pitch attitude
⇒ Altitude hold using pitch attitude
UAV Longitudinal Autopilot: Pitch Attitude Hold
• Pitch attitude hold loop: Similar to roll attitude hold loop

′
θC δe δe q θ
aθ3 s 1
kp θ
+ eθ + s2 +a θ1 s+aθ2 s
− −
kd θ

kpθ aθ3
Hθ/θc (s) =
s2 + (aθ1 + kdθ aθ3 )s + (aθ2 + kpθ aθ3 )
• What about DC gain of this transfer function? nonunity DC gain

• Desired canonical second order transfer function
kθDC ωn2 θ
s2 + 2ζθ ωnθ s + ωn2 θ
• Equating coefficients of characteristic equation
ωn2 θ = aθ2 + kpθ aθ3 , 2ζθ ωnθ = aθ1 + kdθ aθ3
• Gain kpθ chosen to satisfy actuator constraint, then

δemax
kpθ = sign(aθ3 )
emax
θ
• Consideration of sign of aθ3 as it depends on Cmδ e (typically <0)

• Bandwidth limit of pitch loop
r
δemax
ω nθ = aθ2 + |aθ3 |
emax
θ
• Selection of gain kdθ

2ζθ ωnθ − aθ1
kdθ =
aθ3
• What about DC gain of inner loop transfer function as kpθ → ∞?

kpθ aθ3
kθDC = →1
aθ2 + kpθ aθ3
• What about the DC gain value with nominal value of kpθ ?

• Significantly less than one, for typical gain values of kpθ
• Integral feedback to ensure unit DC gain
• Drawback: Limitation on bandwidth by use of integral term
• Homeworks to check these two facts
• Elevator deflection input (Output of pitch attitude hold loop)
δe = kpθ (θc − θ) − kdθ q
UAV Longitudinal Autopilot: Altitude Hold using Pitch
• Altitude-hold autopilot: Successive-loop-closure strategy with

pitch-attitude-hold autopilot as an inner loop
ki h
s
+ ′
q
hc θC δe δe aθ 3 s θ h
kph kpθ s2 +aθ1 s+aθ2
1
s
Va
+ + eθ + s
+ + − −
kd θ
• How to deal with attitude hold design?

• Can we replace the inner loop with unity, as before? No Why?
• DC gain of KθDC , θ = KθDC θc
dh
1
ki h Va
s
+ +
hc + Va
h
kph kθDC s
+ + θC θ
−

KθDC Va kph (s + kih /kph )
h(s) = hc (s)+
s2 + KθDC Va kph s + KθDC Va kih

s
+ dh (s)
s2 + KθDC Va kph s + KθDC Va kih
• What about DC gain and disturbance rejection?

• DC gain of one and rejection of constant disturbance
• Independent closed-loop transfer function on vehicle parameters, except Va
ωnθ
• Selection of gains kph and kih to ensure ωnh = , Wh ∈ [5 15]
Wh
ωn2 h
• Desired closed-loop transfer function
s2 + 2ζnh ωnh s + ωn2 h
• Equating coefficients
2ζnh ωnh
2ζnh ωnh = KθDC Va kph =⇒ kph =
KθDC Va
ωn2 h
ωn2 h = KθDC Va kih =⇒ kih =
KθDC Va
• Pitch command inputs (Output of altitude-hold-with pitch loop)
kih c
θc = kph (hc − h) + (h − h)
s
Reference
Randal Beard and Timothy W. McLain, “Small Unmanned Aircraft:

Theory and Practice”, Princeton University Press, 2012.

Guidance and Control of Unmanned Autonomous Vehicles: Shashi Ranjan Kumar

Uploaded by

Copyright:

Available Formats

Guidance and Control of Unmanned Autonomous Vehicles: Shashi Ranjan Kumar

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Guidance and Control of Unmanned Autonomous Vehicles: Shashi Ranjan Kumar

Uploaded by

Copyright:

Available Formats

Guidance and Control of Unmanned Autonomous Vehicles

Shashi Ranjan Kumar

• Successive loop closure: Performance limitation of the system due to the

• What are other reasons for performance degradation?

⇒ Physical bounds on ailerons restrict roll rate

⇒ Bound on total lateral acceleration due to aircraft stall at higher angle of

• Goal of autopilot design: To design bandwidth of the inner loop to be as

• It allows outer loops to ensure sufficient bandwidth separation of the

• Consider second order system

• Can we achieve any bandwidth at our choice? No

• Natural frequency of closed-loop system under saturation constraints

• How to design lateral autopilot?

• Roll rate damping: kdϕ

• Linear design model:

• Goal of inner loop: To control roll angle and its rate

• Desired second order closed-loop transfer function

• How to find this desired closed-loop transfer function?

ωn2 ϕ = kpϕ aϕ2 , 2ζϕ ωnϕ = aϕ1 + aϕ2 kdϕ

• Gain kpϕ is chosen as per saturation bound on aileron deflections

2ζϕ ωnϕ − aϕ1

• Zero steady-state tracking error in roll without an integrator (Type-1 open

• On solving for ϕ(s)

• What would be the steady-state error for constant step disturbance

• How to draw root locus?

Proper gain for better performance

What next? Course hold outer loop

• With proper tuning of gains, Hϕ/ϕc ≈ 1 over frequencies between 0 to ωnϕ

• Using final value theorem, with constant dχ and χc , χ → χc

2ζχ ωnχ s + ωn2 χ

• Gains selection: Based on desired ζχ and ωnχ

• Due to zero in numerator, standard intuition for selection of ζ does not

• Larger ζ results in larger bandwidth and smaller overshoot.

• On comparing coefficients, we get

• Can we maintain constant or zero sideslip angle?

• Transfer function from β c to β

• Characteristic equation for desired closed-loop poles

s2 + 2ζβ ωnβ s + ωn2 β = 0

• On equating the coefficients of characteristic equations

2ζβ ωnβ = aβ1 + aβ2 kpβ , ωn2 β = aβ2 kiβ

• Maximum error in sideslip (emax

• Complicated longitudinal autopilot as compared to lateral one

• Take-off zone: full throttle is commanded and θ → θc using the elevator.

• Climb zone: to maximize the climb rate

⇒ To maximize the climb rate, full throttle is commanded and airspeed is

⇒ Similarly, if the airspeed drops below nominal value, aircraft is pitched

⇒ Regulating airspeed using pitch attitude effectively avoids stall conditions.

⇒ No regulation of airspeed with pitch altitude immediately after take-off

• Descend zone: Similar to climb zone except zero throttle

⇒ Avoidance of stall conditions by regulating airspeed using pitch angle

⇒ Maximization of descent rate at a given airspeed

• Altitude hold zone

⇒ Airspeed: by adjusting throttle

⇒ Altitude: by commanding pitch attitude.

• Longitudinal autopilot: Necessary feedback loops

⇒ Pitch attitude hold using elevator

⇒ Airspeed hold using throttle

⇒ Airspeed hold using pitch attitude

⇒ Altitude hold using pitch attitude

• Pitch attitude hold loop: Similar to roll attitude hold loop