MPE Mathematical Problems in Engineering 1563-5147 1024-123X Hindawi Publishing Corporation 746379 10.1155/2014/746379 746379 Research Article Robust D -Stability Controller Design for a Ducted Fan Unmanned Aerial Vehicle http://orcid.org/0000-0002-1774-0764 Ren Xiao-lu Wang Chang-hong Yi Guo-xing Sun Weichao Space Control and Inertial Technology Research Center Harbin Institute of Technology, Harbin 150001 China hit.edu.cn 2014 2242014 2014 14 02 2014 17 03 2014 22 4 2014 2014 Copyright © 2014 Xiao-lu Ren et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

This paper deals with the aerodynamic modeling of a small ducted fan UAV and the problem of attitude stabilization when the parameter of the vehicle is varied. The main aerodynamic model of the hovering flight UAV is first presented. Then, an attitude control is designed from a linearization of the dynamic model around the hovering flight, which is based on the H output feedback control theory with D-stability. Simulation results show that such method has good robustness to the attitude system. They can meet the requirements of attitude control and verify further the feasibility of such a control strategy.

1. Introduction

The ducted fan unmanned aerial vehicle is a new popular UAV in the recent years. Almost all system forces and moments come from the ducted fan. This UAV with VTOL (vertical takeoff and landings) and hovering flight capability is shown in Figure 1.

Ducted fan UAV.

The ducted fan UAV can be adapted to a variety of complex environments and complete difficult flight missions [1, 2]. It is small with a compact and flexible layout. It is capable of maneuvering in any angle in addition to the normal hover and vertical takeoff and landing capabilities; therefore it can complete some special tasks [3, 4]. The peripheral duct not only enables the fan to be effectively protected, but also provides some aerodynamic lift for the vehicle's flight. The experiments show that the ducted unique geometry can provide some aerodynamic lift for the vehicle .

Compared with ordinary fixed-wing aircraft, ducted fan UAV has such advantages.

( 1 ) Flexibility. Fixed-wing aircraft needs ground support in takeoff and landing. So it cannot provide timely information assurance to combat team in the city and mountain area. The ducted fan UAV can do vertical takeoff and landing and hover to adapt to the complex environment of the city or mountain. It can hover and stare at the target and fly close to the building to provide the precision position of the target [6, 7].

( 2 ) Compact Structure and High Propulsive Force. Compared with unmanned helicopter, in the same power consumption, the ducted fan will produce greater tension than isolated propeller with the same diameter. The ducted fan UAV has more compact structure, less drag. And its flight attitude is closer to fixed-wing aircraft. So it can fly faster than the equivalent helicopter.

( 3 ) Low Noise, More Concealment, and Better Security. The propeller is placed in the ducted fan, which can obstruct the transmission of aerodynamic noise. So the noise intensity and propagation distance can be reduced to some extent. Meanwhile the engine is surrounded by the ducted fan, which can obstruct the engine’s thermal radiation, and then the vehicle’s thermal radiation characteristic can be reduced. So the ducted fan UAV has better concealment, which can provide better survivability to the vehicle in the battle.

Compared with traditional fixed-wing aircraft, the flow field distribution of the ducted fan UAV is more complex, which makes the vehicle’s modeling difficult. The attitude control is the key problem in the ducted fan UAV control, and it will affect the vehicle’s other actions. In the flight, because the vehicle’s flow field is complex , the nonlinear effect is obvious , and these will be big problems in the vehicle’s control. In references [18, 19], an evolutionary pinning control is put toward which is applied in unmanned aerial vehicle coordination. Reference  used a PID control method. The PID method has a simple control structure and a low calculation amount, but the parameter adjustment is too cumbersome, and it has a poor adaptability for the coupling between the axes. Reference  used a robust H method. The H control has good inhibitory effect on the outside interference, but it is conservative to the uncertainty of the system. Reference  used a feedback linearization method, which has a good antijamming capability, but complex calculations.

This paper uses H output feedback control with D-stability for ducted fan UAV attitude control system with parameter uncertainty. The status of the system may be difficult to observe in the practical engineering. So the output variable which can be observed is used as feedback variable in this paper. Therefore, the output feedback control system has good physical realizability. Using H control method can resolve the uncertainties of the system but cannot provide location information of the closed-loop poles. H control method used in this paper which is combined with regional pole assignment not only ensures the robustness of control system, but also places the poles of the closed-loop system in a specified region. Dynamic performance of the UAV can be satisfied. This controller is designed by solving a set of linear matrix inequalities. Simulation results show this controller not only guarantees the closed-loop system to be D-stable, but also achieves the given H disturbance index. The vehicle can meet the requirements of attitude control and has good practicality, which illustrates the effectiveness of the proposed method.

2. UAV Attitude Modeling 2.1. Rotor Aerodynamics

Compared with the fixed-wing aircraft, the ducted fan vehicle has a more complicated air environment. The modeling of the fan is the main difficulty to the modeling of the whole vehicle.

There are two theories to study the fan, the momentum theory, and the blade element theory. This paper uses the two theories above to present the fan’s aerodynamics equations.

The thrust can be expressed as  (1) T fan = ( v b - v i ) · ρ · ω r · a · n · d · R 2 4 , where ω r represents angular velocity of rotor, ρ is free stream density, R is the radius of the rotor, a is rotor blade lift curve slope, n represents the number of rotor blades, d is the rotor blade chord, v b represents the flow through the rotor, and v i is rotor induced velocity. The far field velocity can be expressed as (2) | v | = v x 2 + v y 2 + ( v z - v i ) 2 .

The induced velocity can be now expressed as follows: (3) v i = T 2 ρ π R 2 | v | .

The airflow along the z -axis can be expressed as (4) v b = v z + 1 2 ω r R K twist , where K twist is the twist of the blades.

2.2. Duct

The airflow tends to follow the direction of the duct contour due to the Coandă effect. As a result, the wake will have a larger area than that of a helicopter. Comparing with a traditional helicopter, the duct provides additional thrust which can be attributed to the pressure distributions above and below the propeller. As shown in , the total thrust is the sum of the propeller or fan thrust and the force experienced by the duct: (5) T = T fan + T duct , T fan T = 1 2 a d , where a d represents the ratio between the area of the wake and the propeller disk area.

The force on the vehicle due to the ducted fan can now be expressed as (6) F rotor = [ 0 0 - T ] .

The rotor twisting moment can be expressed as (7) M rotor = [ 0 0 - P ω r ] , where P is rotor output power.

2.3. Control Vane

The rudder control torque is (8) L = 1 2 ρ v i 2 S r C L r ( δ ) , D = 1 2 ρ v i 2 S r C D r ( δ ) , where S r is rudder surface area and C L r , and C D r are dimensionless lift coefficient and resistance coefficient, which are related to the deflection angle of the rudder.

The rudder surface distribution is shown in Figure 2; the control volume is δ 1 = δ p + δ r , δ 2 = δ q - δ r , δ 3 = δ p - δ r , and δ 4 = δ q + δ r , where δ p , δ q , and δ r are roll, pitch, and yaw control volume and δ 1 , δ 2 , δ 3 , and δ 4 are the deflection angles of the rudder.

Configuration of control vane.

2.4. Gyroscopic Moment

The expression for gyroscopic moment vector is given as (9) M gyro = n I b ω r [ - ω y ω x 0 ] , where I b is the rotational inertia of the blade.

3. Modeling

The force and moment vectors can be expressed as (10) F = F aero + F rotor + F vane , M = M aero + M rotor + M vane + M gyro .

The UAV dynamics and kinematics equations are given as (11) v ˙ x = - ( v z ω y - v y ω z ) + F x m - g sin θ , v ˙ y = - ( v x ω z - v z ω x ) + F y m - g cos θ sin ϕ , v ˙ z = - ( v y ω x - v x ω y ) + F z m - g cos θ cos ϕ , I x x ω ˙ x = ( I y y - I z z ) ω y ω z + I x z ( ω ˙ z + ω z ω y ) + M x , I y y ω ˙ y = ( I z z - I x x ) ω z ω x + I x z ( ω z 2 - ω x 2 ) + M y , I z z ω ˙ z = ( I x x - I y y ) ω x ω y + I x z ( ω ˙ x + ω y ω z ) + M z , ϕ ˙ = ω x + ω y sin ϕ tan θ + ω z cos ϕ tan ϕ , θ ˙ = ω y cos ϕ - ω z sin ϕ , ψ ˙ = ω y sin ϕ sec θ + ω z cos φ sec θ , where ϕ , θ , and ψ represent the vehicle roll, pitch, and yaw attitudes; ω x , ω y , and ω z represent vehicle roll rate, pitch rate, and yaw rate; v x , v y , and v z represent body-fixed velocity in the body-fixed x , y , z directions.

Hovering and low speed flight is the ducted fan UAV’s main flight state. Using small perturbation theory and the Taylor expansion, the nonlinear equations of the system can be linearized: (12) x ˙ = A x + B u , y = C x + D u , where x = [ v x , v y , v z , ω x , ω y , ω z , ϕ , θ , ψ ] T , u = [ δ p , δ q , δ r ] T , and δ p , δ q , and δ r represent the control input.

3.1. Robust <inline-formula> <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="M68"> <mml:mrow> <mml:mi>D</mml:mi></mml:mrow> </mml:math></inline-formula>-Stability Analysis in UAV Control System

Fully considering the uncertainty of the system, control object can be expressed as (13) x ˙ = ( A + Δ A ) x ( t ) + B 1 ω ( t ) + ( B 2 + Δ B ) u ( t ) , z = C 1 x ( t ) + D u ( t ) , y = C 2 x ( t ) , where x ( t ) R n is the state vector of generalized object, u ( t ) R m is the control input, ω ( t ) R p is the interference signal, z ( t ) R q is the output to adjust,    y ( t ) R f is the output to measure, A is n × n real constant matrix, B 1 is n × p real constant matrix, B 2 is n × m real constant matrix, C 1 is q × n real constant matrix, D is q × m real constant matrix, C 2 is f × n real constant matrix, Δ A , Δ B represent the uncertainty of the real parameters in system state matrix and input matrix, ( Δ A Δ B ) = H F ( E 1 E 2 ) , H , E 1 , and E 2 represent the constant matrix which have the corresponding dimension, and F represents unknown matrices and meets F F T I .

Definition 1.

Given performance index γ > 0 , region D ( - q , r ) , as shown in Figure 3, the feedback controller (14) is said to be H output feedback controller with D -stability, if the following conditions are established.

Region D .

If the uncertainties of the control system meet ( Δ A Δ B ) = H F ( E 1 E 2 ) and F F T I , then the close-loop system is internal stability.

The poles of the augmented closed-loop system are placed in region D shown in Figure 3.

T z ω is minimized, where T z ω ( s ) is the closed-loop transfer function from ω to z ; meanwhile, given γ > 0 , T z ω < γ , (14) x ˙ c = A c x c + B c y , u = C c x c + D c y .

Augmented closed-loop system represents as (15) x ˙ cl = A cl x cl + B cl ω , z = C cl x cl ,

where x cl = ( x x c ) T is the state vector of the augmented closed-loop system and system matrixes are represented, respectively, as (16) A cl = [ A + Δ A + ( B 2 + Δ B ) D c C 2 ( B 2 + Δ B ) C c B c C 2 A c ] , B cl = [ B 1 0 ] , C cl = [ C 1 + D D c C 2 D C c ] .

Lemma 2.

Let U , W , and V be any given real matrixes which have the corresponding dimension. For a given symmetric matrix M , the following statements are equivalent.

For any given matrix V which meets V T V I , U , W ,  and V satisfy (17) M + U V W + W T V T U T < 0 .

There exists a real number ε > 0 which makes (18) M + ε - 1 U U T + ε W T W < 0 .

Theorem 3.

Given upper bound γ > 0 and region D ( - q , r ) , for uncertainty system (13), if there exists symmetric positive matrix    P = P T > 0 and real number ε > 0 for which (19) [ N * B cl P C cl T ( A - cl + α I ) P P E - T ε H - B cl T - γ 2 0 0 0 0 C cl P 0 - I 0 0 0 P ( A - cl + α I ) T 0 0 - r P P E - T 0 E - P 0 0 E - P - ε I 0 ε H - T 0 0 0 0 - ε I ] < 0 .

Then ( 1 ) the poles of the augmented closed-loop system are placed in region D ( - q , r ) ; ( 2 ) the close-loop system is stability internal and T z ω < γ , where (20) α = q - r , N * = ( A - cl + α I ) P + P ( A - cl + α I ) T , A - cl = [ A + B 2 D c C 2 B 2 C c B c C 2 A c ] , H - = [ H H 0 0 ] , F - = [ F 0 0 F ] , E - = [ E 1 0 E 2 D c C 2 E 2 C c ] .

Proof.

Consider (21) A cl = A - cl + Δ A cl , (22) Δ A cl = [ Δ A + Δ B D c C 2 Δ B C c 0 0 ] = [ H F E 1 + H F E 2 D c C 2 H F E 2 C c 0 0 ] = H - F - E - , (23) [ N * B cl P C cl T ( A - cl + α I ) P B cl T - γ 2 0 0 C cl P 0 - I 0 P ( A - cl + α I ) T 0 0 - r P ] + [ Δ A cl P + P Δ A cl T 0 0 Δ A cl P 0 0 0 0 0 0 0 0 P Δ A cl T 0 0 0 ] < 0 , (24) Z - = [ N * B cl P C cl T ( A - cl + α I ) P B cl T - γ 2 0 0 C cl P 0 - I 0 P ( A - cl + α I ) T 0 0 - r P ] where (23) is equivalent to (25) Z - + [ H - F - E - P + P ( H - F - E - ) T 0 0 H - F - E - P 0 0 0 0 0 0 0 0 P ( H - F - E - ) T 0 0 0 ] < 0 .

Then (26) Z - + [ H - 0 0 0 ] F - [ E - P 0 0 E - P ] + [ P T E - T 0 0 P T E - T ] F - T [ H - 0 0 0 ] < 0 .

Let (27) M - = [ H - 0 0 0 ] , N - = [ E - P 0 0 E - P ] , where (26) is equivalent to (28) Z - + M - F - N - + N - T F - T M - T < 0 .

According to Lemma 2, (25) is founded when all F meet F F T I , only if there exists a positive constant ε , let (29) Z - + ε M - M - T + ε - 1 N - T N - < 0 . Which is equivalent, by the Scher complement, to (30) [ Z - N - T M - N - - ε I 0 M - T 0 - ε - 1 I ] < 0 . Let inequality (30) premultiply and postmultiply diag ( I I ε I ) .

The proof is completed.

Since inequality (19) contains the nonlinear terms, it cannot be calculated directly by LMI. So the variable substitution should be performed to the inequality.

Lemma 4.

There exists symmetry positive definite matrix X ; the necessary and sufficient condition that meets both (31) and (32) is (31) X = [ R M M T * ] , (32) X - 1 = [ S N N T * ] , (33) [ R I I S ] > 0 .

Theorem 5.

Given H performance index γ > 0 , positive real number ε , and region D ( - q , r ) , if there exist corresponding dimension real matrixes R = R T and S = S T , matrixes A ~ c , B ~ c , C ~ c , and D ~ c satisfy inequalities (33) and (34): (34) [ Φ 11 Φ 12 B 1 Φ 14 Φ 15 Φ 16 R E 1 T Φ 18 ε H ε H Φ 12 T Φ 22 S B 1 Φ 24 Φ 25 Φ 26 E 1 T Φ 28 ε S H ε S H B 1 T B 1 T S - γ 2 I 0 0 0 0 0 0 0 Φ 14 T Φ 24 T 0 - I 0 0 0 0 0 0 Φ 15 T Φ 25 T 0 0 - r R - r I R E 1 T Φ 18 0 0 Φ 16 T Φ 26 T 0 0 - r I - r S E 1 T Φ 28 0 0 E 1 R E 1 0 0 E 1 R E 1 - ε I 0 0 0 Φ 18 T Φ 28 T 0 0 Φ 18 T Φ 28 T 0 - ε I 0 0 ε H T ε H T S 0 0 0 0 0 0 - ε I 0 ε H T ε H T S 0 0 0 0 0 0 0 - ε I ] < 0 , (35) Φ 11 = A R + R A T + B 2 C ~ c + C ~ c T B 2 T + 2 α R , Φ 12 = A + A ~ c T + B 2 D ~ c C 2 + 2 α I , Φ 22 = S A + A T S + B ~ c C 2 + C 2 T B c T + 2 α S , Φ 14 = R C 1 T + C ~ c T D T , Φ 24 = C 1 T + C 2 T D c T D T , Φ 15 = A R + B 2 C ~ c + α R , Φ 16 = A + B 2 D ~ c C 2 + α I , Φ 25 = A ~ c + α I , Φ 26 = S A + B ~ c C 2 + α S , Φ 18 = C ~ c T E 2 T , Φ 28 = C 2 T D ~ c T E 2 T . Then all poles of the closed-loop system are placed within the circular region D ( - q , r ) , and the system is stable and T z ω ( s ) < γ .

Proof.

By Lemma 4, there exists a symmetric positive definite matrix P = [ R M M T U ] and inverse matrix P - 1 = [ S N N T V ] , in which R , and S are symmetric positive definite matrixes, as P P - 1 = I , and can get M N T = I - R S ; because [ R I I S ] > 0 , then I - R S is invertible and then M ,  and   N are full-ranked; let (36) Γ 1 = [ R I M T 0 ] , Γ 2 = [ I S 0 N T ] . Obviously (37) P Γ 2 = Γ 1 , Γ 2 T P Γ 2 = [ R I I S ] . Then matrix P = Γ 2 Γ 2 - 1 ; furthermore P > 0 .

Let inequality (34) premultiply diag ( Γ 2 T , I , I , Γ 2 T , I , I ) and postmultiply diag ( Γ 2 , I , I , Γ 2 , I , I ) .

We can have the following inequality: (38) [ Θ 11 Θ 12 B 1 Θ 14 Θ 15 Θ 16 R E 1 T Θ 18 ε H ε H Θ 12 T Θ 22 S B 1 Θ 24 Θ 25 Θ 26 E 1 T Θ 28 ε S H ε S H B 1 T B 1 T S - γ 2 I 0 0 0 0 0 0 0 Θ 14 T Θ 24 T 0 - I 0 0 0 0 0 0 Θ 15 T Θ 25 T 0 0 - r R - r I R E 1 T Θ 18 0 0 Θ 16 T Θ 26 T 0 0 - r I - r S E 1 T Θ 28 0 0 E 1 R E 1 0 0 E 1 R E 1 - ε I 0 0 0 Θ 18 T Θ 28 T 0 0 Θ 18 T Θ 28 T 0 - ε I 0 0 ε H T ε H T S 0 0 0 0 0 0 - ε I 0 ε H T ε H T S 0 0 0 0 0 0 0 - ε I ] < 0 , where (39) Θ 11 = A R + R A T + B 2 D c C 2 R + B 2 C c M T + R C 2 T D c T B 2 T + M C c T B 2 T + 2 α R , Θ 12 = A + B 2 D c C 2 + R A T S + R C 2 T D c T B 2 T S + R C 2 T B c T N T + M C c T B 2 T S + M A c T N T + 2 α I , Θ 22 = S A + A T S + S B 2 D c C 2 + N B c C 2 + C 2 T B c T N T + C 2 T D c T B 2 T S + 2 α S , Θ 14 = R C 1 T + R C 2 T D c T D T + M C c T D T , Θ 24 = C 1 T + C 2 T D c T D T , Θ 15 = A R + B 2 D c C 2 R + B 2 C c M T + α R , Θ 16 = A + B 2 D c C 2 + α I , Θ 25 = S A R + S B 2 D c C 2 R + N B c C 2 R + S B 2 C c M T + N A c M T + α I , Θ 26 = S A + S B 2 D c C 2 + N B c C 2 + α S , Θ 18 = R C 2 T D c T E 2 T + M C c T E 2 T , Θ 28 = C 2 T D c T E 2 T . Since the inequality contains the nonlinear terms, the variable substitution should be performed.

Define variable substitution as (40) A ~ c = S A R + N A c T M T + S B 2 C c M T + S B 2 D c C 2 R + N B c C 2 R , B ~ c = S B 2 D c + N B c , C ~ c = C c M T + D c C 2 R , D ~ c = D c . Substitutioning inequality (40) into inequality (38) results in inequality (34). The proof is completed.

3.2. Robust <inline-formula> <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="M180"> <mml:mrow> <mml:mi>D</mml:mi></mml:mrow> </mml:math></inline-formula>-Stability Dynamic Output Feedback Controller Design

Based on the above analysis, the nonlinear inequalities are converted to linear inequalities by variable substitution. From Theorem 5, for the system with parameter uncertainty, given H performance index γ > 0 , positive real number ε > 0 , and circular region D ( - q , r ) , then the output feedback controller with H - D stability is designed as follows.

Get a feasible solution of linear matrix inequalities R , S , A ~ c , B ~ c , C ~ c , and D ~ c .

Verify that the matrix I - R S is reversible. If reversible, make full rank decomposition where M N T = I - R S .

According to variable substitution equation (40), we can get (41) D c = D ~ c , C c = ( C ~ c - D ~ c C 2 R ) M - T , B c = N - 1 ( B ~ c - S B 2 D ~ c ) , A c = N - 1 [ A ~ c - S ( A + B 2 D ~ c C 2 ) R ] M - T - B c C 2 R M - T - B c C 2 R M - 1 S B 2 C c ,

where ( A c , B c , C c , D c ) is the output feedback controller.

3.3. Simulation of Ducted Fan UAV Attitude Control

Based on the above model, the dynamic output feedback controller of the closed-loop system can be built. According to Theorem 5, the reconstruction of control law of systems is designed to place the closed-loop system poles into the circle region D ( - 5 , 4.5 ) ; let γ = 0.92 , ε = 1 .

Perturbation parameters H = [ 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 ] T and ( E 1 E 2 ) = [ 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 ] .

Then satisfy F T ( t ) F ( t ) I , according to previous methods, the matrixes in control law reconfigurable systems are (42) A c = [ - 22.64 - 0.2732 - 0.09441 - 1.568 - 0.4542 - 0.2776 0.5601 0.601 0.1774 0.06462 - 8.671 0.6483 16.53 15.78 3.741 0.5845 1.741 0.7923 - 0.8520 - 2.610 - 0.8109 5.448 - 0.1541 0.8185 - 0.9721 0.3199 0.3014 1.983 5.107 2.086 - 11.26 6.820 - 0.9133 0.8767 - 0.1671 - 0.5441 - 2.984 0.04552 - 3.048 - 1.52 - 23.44 - 2.365 - 1.378 - 0.9767 0.06624 - 7.701 2.591 - 5.136 - 10.90 - 23.26 - 3.992 - 0.2918 - 0.5453 - 0.4525 - 6.050 476.2 28.74 93.85 171.2 38.17 - 1.440 10.07 8.601 114.2 - 590.1 - 39.47 - 118.6 - 239.3 - 50.20 - 0.7837 - 14.37 - 10.601 - 19.19 - 2.605 - 17.23 - 19.65 - 50.55 - 5.615 - 0.6332 - 1.285 - 1.281 ] , B c = [ - 0.02361 - 0.01612 - 0.2993 - 0.1272 - 0.1637 - 0.9465 0.04412 0.3329 - 0.3412 0.1168 - 0.3121 0.1372 0.9954 0.4566 0.8167 0.8109 - 0.4101 - 0.2643 - 0.04551 - 0.04553 - 0.6308 0.4096 - 0.7956 0.4148 - 01628 0.4089 0.9448 - 0.03100 0.1068 - 0.1281 - 0.7726 0.3786 - 0.3129 0.3200 - 0.1454 - 0.2687 - 0.03587 - 0.09831 - 0.3434 - 0.9012 - 0.5590 - 0.1777 - 0.7323 0.7785 - 0.1931 0.5540 - 0.2112 0.4984 0.3361 0.8961 - 0.5204 - 0.7233 0.3424 0.5452 1.985 - 110.9 3.859 16.64 4.160 15.07 - 29.60 - 3.180 - 2.672 - 9.969 137.8 - 4.437 - 19.92 - 4.290 - 17.78 36.75 4.460 3.364 1.338 0.9083 - 2.200 - 0.2673 1.703 - 0.3880 2.000 0.7796 1.650 ] , C c = [ 301.8 - 4.545 10.80 51.19 95.25 15.79 - 6.855 - 6.062 - 1.385 - 297.4 4.901 - 4.647 - 46.91 - 48.35 - 11.11 9.613 8.017 1.247 0.4401 3.211 0.2304 - 7.527 - 3.683 - 1.338 - 0.1323 - 0.5338 - 0.3353 ] , D c = [ 1.320 0.7305 1.317 0.6735 6.697 0.4660 - 0.3071 - 0.8931 0.3908 - 0.4978 0.4364 1.101 0.9777 5.969 0.3108 0.6912 0.3042 0.05631 - 0.3687 0.9844 - 0.2012 - 4.332 - 0.8031 - 3.081 0.9728 0.04252 0.09001 ] . Closed-loop poles distribution as shown in Figure 4.

Closed-loop pole of the UAV control system.

As shown in Figure 4, in F T ( t ) F ( t ) I , the closed-loop system poles can be placed in a designated region within the left half-plane of the complex plane. Zero input response is displayed in Figures 5, 6, 7, 8, 9, and 10 which show that each state of the closed-loop system can quickly converge to zero. The system is asymptotically stable. Before and after the parameters change, the system’s dynamic response curve does not change significantly. The above results show that the dynamic output feedback controller effectively solves the system’s uncertainty problem.

Response curve of roll angle before and after parameter variation.

Response curve of pitch angle before and after parameter variation.

Response curve of yaw angle before and after parameter variation.

Response curve of x angle velocity before and after parameter variation.

Response curve of y angle velocity before and after parameter variation.

Response curve of z angle velocity before and after parameter variation.

From the simulation results in Figures 510, three-axis attitude angle and angular velocity can quickly maintain stability with strong robustness. The ducted fan UAV has system uncertainty during the flight. As can be seen from the response curve before and after the load chance, in the presence of uncertainty, system poles can be maintained within the designated region and has good stability and dynamic performance.

4. Conclusion

This paper has established the ducted fan UAV’s mathematical model and proposed a H -D stabilization control method for linear uncertainty. This method gives the condition that meets both the interference suppression index constraints and regional pole constraints, in the form of linear matrix inequality (LMI).

This control method is applied in the ducted fan UAV’s attitude control system. The simulation results demonstrate the effectiveness of this method. The experimental results show that this method has good robustness for both model linear uncertainty and controller uncertainty. This method can effectively suppress the crosswind interference to the control system. The regional pole constraints also give the system a good dynamic performance.

Conflict of Interests

There is no conflict of interests regarding the publication of this paper.

Ruzicka G. C. Strawn R. C. Meadowcroft E. T. Discrete-blade, Navier-Stokes computational fluid dynamics analysis of Ducted-Fan flow Journal of Aircraft 2005 42 5 1109 1117 2-s2.0-27144532157 Akturk A. Shavalikul A. Camci C. PIV measurements and computational study of a 5-inch ducted fan for V/STOL UAV applications Proceedings of the 47th AIAA Aerospace Sciences Meeting January 2009 Orlando, Fla, USA 2-s2.0-78549250715 Johnson E. N. Turbe M. A. Modeling, control, and flight testing of a small ducted-fan aircraft Journal of Guidance, Control, and Dynamics 2006 29 4 769 779 2-s2.0-33747075794 10.2514/1.16380 Haiming W. Huaming W. A General Summarize for the Ducted Fan Computational Methods Institute of Helicopter Technology, Nanjing University of Aeronautics and Astronautics Zhang X. Myklebust A. Gelhausen P. A geometric modeler for the conceptual design of ducted fan UAVs Proceedings of the AIAA 3rd Unmanned Unlimited Technical Conference September 2004 Chicago, Ill, USA AIAA 2004-6538 705 714 2-s2.0-19644397926 Lee J. K. Lee D. H. Kwon S. Performance prediction and design of a ducted fan system Proceedings of the 40th AIAA/ASME/SAE/ASEE Joint Propulsion Conference July 2004 Fort Lauderdale, Fla, USA AIAA 2004-4196 2-s2.0-77957847886 Wright T. Evaluation of the design parameters for optimum heavily loaded ducted fans Journal of Aircraft 1970 7 6 512 517 2-s2.0-0014870003 Guerrero I. Londenberg K. Gelhausen P. Myklebust A. A powered lift aerodynamic analysis for the design of ducted fan UAVs Proceedings of the 2nd AIAA Unmanned Unlimited Systems, Technologies, and Operations-Aerospace 2003 San Diego, Calif, USA AIAA 2003-6567 Chang I. C. Rajagopalan R. G. CFD analysis for ducted fans with validation Proceedings of the 21st Applied Aerodynamics Conference 2003 Orlando, Fla, USA AIAA 2003-4079 Ohanian O. J. III Gelhausen P. A. Inman D. J. A compact method for modeling the aerodynamics of ducted fan vehicles Proceedings of the 48th AIAA Aerospace Sciences Meeting Including the New Horizons Forum and Aerospace Exposition January 2010 Orlando, Fla, USA 2-s2.0-78649862769 Ohanian O. J. III Karni E. D. Kelly Londenberg W. Gelhausen P. A. Inman D. J. Ducted-fan force and moment control via steady and synthetic jets Proceedings of the 27th AIAA Applied Aerodynamics Conference June 2009 San Antonio, Tex, USA 2-s2.0-77957835184 Pflimlin J.-M. Binetti P. Souères P. Hamel T. Trouchet D. Modeling and attitude control analysis of a ducted-fan micro aerial vehicle Control Engineering Practice 2010 18 3 209 218 2-s2.0-77649180894 10.1016/j.conengprac.2009.09.009 Marconi L. Naldi R. Gentili L. Modelling and control of a flying robot interacting with the environment Automatica 2011 47 12 2571 2583 2-s2.0-81155151361 10.1016/j.automatica.2011.09.020 Naldi R. Marconi L. Optimal transition maneuvers for a class of V/STOL aircraft Automatica 2011 47 5 870 879 2-s2.0-79954875964 10.1016/j.automatica.2011.01.027 Peddle I. K. Jones T. Treurnicht J. Practical near hover flight control of a ducted fan (SLADe) Control Engineering Practice 2009 17 1 48 58 2-s2.0-53649085110 10.1016/j.conengprac.2008.05.004 Naldi R. Gentili L. Marconi L. Sala A. Design and experimental validation of a nonlinear control law for a ducted-fan miniature aerial vehicle Control Engineering Practice 2010 18 7 747 760 2-s2.0-77953613667 10.1016/j.conengprac.2010.02.007 Yu J. Jadbabaie A. Primbs J. Huang Y. Comparison of nonlinear control design techniques on a model of the Caltech ducted fan Automatica 2001 37 12 1971 1978 2-s2.0-0035546395 10.1016/S0005-1098(01)00149-2 Tang Y. Gao H. Zou W. Distributed synchronization in networks of agent systems with nonlinearities and random switchings IEEE Transactions on Cybernetics 2013 43 1 358 370 Tang Y. Gao H. Kurths J. Evolutionary pinning control and its application in UAV coordination IEEE Transactions on Industrial Informatics 2012 8 4 828 838 Kim H. J. Shim D. H. A flight control system for aerial robots: algorithms and experiments Control Engineering Practice 2003 11 12 1389 1400 2-s2.0-0346118763 10.1016/S0967-0661(03)00100-X Yang C.-D. Liu W.-H. Robust H decoupling hover control of uncertain nonlinear helicopter Proceedings of the AIAA Guidance, Navigation, and Control Conference and Exhibit June 2003 Austin, Tex, USA AIAA 2003-5489 3454 3459 2-s2.0-0142152904 Fritsch O. De Monte P. Buhl M. Lohmann B. Quasi-static feedback linearization for the translational dynamics of a quad rotor helicopter Proceedings of the American Control Conference 2012 P125 P130 Johnson E. N. Turbe M. A. Modeling, Control, and Flight Testing of a Small Ducted Fan Aircraft Georgia Institute of Technology Roberts A. D. Attitude Estimation and Control of a Ducted Fan VTOL UAV [M.S. thesis] 2007 Thunder Bay, Ontario, Canada Lakehead University