Advances in Unmanned Aerial Vehicle Technologies

ARTICLES ISIUS 2008 -Nanjing © 2008 KEYNOTE SPEECH submit stencil Advances in Unmanned Aerial Vehicles Technologies Agus Budiyono1 1 Smart Robot Center, Department of Aerospace Information Engineering, Konkuk University, Seoul, Korea. Previously with Center for Unmanned System Studies, Institut Teknologi Bandung, Indonesia The utilization of unmanned vehicles has become increasingly more popular today and been successfully demonstrated for various civil and military applications. The unmanned aerial vehicles (UAVs) have shown applications in different areas including crop yield prediction, land use surveys in rural and urban regions, traffic surveillance and weather research. The unmanned small scale helicopters are particularly suitable for demanding problems which requires accurate low-speed maneuver and hovering capabilities such as detailed area mapping. Generally a certain level of autonomous flight capability is required for the vehicle to achieve its mission. The basic autonomy level is to maintain its stability following a desired path under embedded guidance, navigation and control algorithm. The UAV technology trends indicate that to cope with the more stringent operation requirements, the UAVs should rely less and less on the skill of the ground pilot and progressively more on the autonomous capabilities dictated by a reliable onboard computer system. To systematically develop and enhance flight autonomy, a rotary wing UAV (RUAV) or model helicopter has been proposed and used as a flying test-bed at various major research centers. The ability of the helicopter to operate in the hovering mode makes it an ideal platform for a step-by-step autonomous capability development. On the other hand, a small helicopter exhibits not only increased sensitivity to control inputs and disturbances, but also a much richer dynamics compared to conventional unmanned aerial vehicles (UAVs). The paper surveys recent advances in modeling, control and navigation of autonomous unmanned aerial vehicles. Without loss of generality, an autonomous small scale helicopter research program is taken as a case study. Approaches to modeling and control for such a vehicle are presented and discussed. Future directions in the advancement of UAV technologies are identified and key barriers highlighted. Unmanned aerial vehicle, model identification, control, navigation, trajectory generation I. Introduction A widely used definition of UAV is an aerial vehicle (including fixed-wing, rotary-wing or airship platform) which can sustain its flight along a prescribed path without an on-board pilot. The UAV technology has proven applications in many areas such as environmental monitoring and protection, meteorological surveillance and weather research, agriculture, mineral exploration and exploitation, aerial target system, airborne surveillance for military land operations, and reconnaissance missions. The unmanned small scale helicopters enjoy no requirement for runway and are particularly suitable for demanding problems such as traffic or volcanic areas surveillance, detailed area mapping, video footage recordings and crop dusting or spraying. Table 1 lists applications of contemporary UAVs in different areas. A recent progress in the supporting technologies has enabled the development of semi to fully autonomous UAV. This includes the availability of compact, lightweight, affordable motion detecting sensors essential to the flight control system and compact lightweight low-cost computing power for autonomous flight control. A wide varieties of autonomous UAV platforms have been developed and flown ranging from fixed-wing to rotary wing platforms, International Symposium on Intelligent Unmanned System | October 15-18, 2008 | several minutes to hours/day in endurance and 100 grams to 800 kg in weight. From all types of UAVs, the small scale rotorcraft-based vehicle has been considered one that exhibits the most complex dynamic properties. From the perspective of control area, the RUAV demonstrates literally all challenges that have attracted enormous interests from industry and academia alike. The challenging problems include higher bandwidth, hybrid modes, non-holonomic, under-actuation, multi input multi output (MIMO), and non-minimum phase. The paper discusses the advances in UAV technologies from the perspective of modeling and control of rotorcraft-based aerial vehicles. II. Background: Science and Technology A. Survey of UAVs The viability of UAV as a multipurpose research vehicle has driven great interest since recent decades. The basic technology building blocks responsible for the current advances include airframes, propulsion systems, payloads, safety or protection systems, launch and recovery, data processor, ground control station, navigation and guidance, and autonomous flight controllers. The following brief survey is focused on the area of navigation, guidance and control of UAVs. Various control design for UAVs has been proposed ranging from linear to nonlinear synthesis, time invariant to parameter varying, and conventional PID to intelligent control approaches. The developed controllers have been implemented for different aerial platforms: airship (blimp), fixed-wing UAV, small scale helicopter, quad-rotors, and MAV. The research on autonomous airship is reported in (Azinheira, 2008) where the authors proposed a nonlinear control approach for the path-tracking of an autonomous underactuated airship. A backstepping controller is designed from the airship nonlinear dynamic model including wind disturbances, and further enhanced to consider actuators saturation. The hover control using the same approach for such a vehicle is presented in (Azinheira and Moutinho, 2008). A number of investigations have been conducted for control and stabilization of quadrotor UAV. In (Raffo, 2008), a robust control strategy to solve the path tracking problem for such a vehicle was designed in consideration of external disturbances like aerodynamic moments. A state parameter control based on Euler angles and open loop positions state observer was proposed by Mokhtari and Benallegue (2004). The work was continued in (Mokhtari, 2005) in which a mixed robust feedback linearization with linear GH∞ controller was applied. An actuator saturation and constrain on 2 state space output are introduced to analyze the worst case of control law design. A different approach was proposed in (Madani, 2007) where a backstepping control running parallel with a sliding mode observer for a quadrotor vehicle. The sliding mode observer works as an observer of the quadrotor velocities and estimator of the external disturbances such as wind and parameter uncertainties. In (Escareno et.al., 2008), the authors proposed a three-rotor configuration which incorporates certain structural advantages in order to improve the attitude stabilization. The control strategy is robust with respect to dynamic couplings and to the adverse torques produced by the gyroscopic-effect and propeller’s drag. The research on autonomous flight using model helicopters as a test-bed has been performed by a large number of teams all over the world. The MIT UAV team successfully developed an autonomous aerobatic helicopter in (Gravilets, 2003). The development relied on the modeling framework of the miniature helicopter dynamics. A methodology for designing model-based control strategies for autonomous aerobatic maneuver was proposed and validated experimentally. Referring to previous work by Mettler (Mettler et.al., 2002) at Carnegie Mellon Robotics Institute, the basis for a simplified modeling framework was considered to stem from the fact that the dynamics of small-scale helicopters is dominated by the rotor response. The real-time control system was developed using a Hardware-In-the-Loop (HIL) simulation system which allows high fidelity representation of the signal’s time-dependence in real time navigation scheme At Georgia Tech, the Open Control Platform (OCP)—a new object-oriented real time operating software architecture— has been used onboard the GTMAX UAV helicopter to compensate for the simulated in-flight failure of a low level flight control system. The viability of designing inexpensive architecture, along with a relatively simple processor, will pave the way for the extremely low-cost flight control and guidance systems. Another novel contribution was the use of Pseudo Control Hedging (PCH) in the adaptive flight control scheme for improving tracking performance of a small helicopter. Using this architecture, a consolidated reference command that includes position, velocity, attitude and angular rate may be provided to the control system. At UC Berkeley, the research on an autonomous helicopter has been conducted as reported in Koo and Sastry(1998), Koo et.al.(2001) and Kim et.al.(2003). A helicopter mathematical model is first established with the lump-parameter approach. The control models of the Agus Budiyono International Symposium on Intelligent Unmanned System | October 15-18, 2008 The discussion on this paper is centered on model-based control design and navigation system technology in the framework of recent advances in UAVs elaborated in the following order. In the section below (II), the system and technology background of UAVs are presented including a brief survey of contemporary UAVs, summary of lessons from the research on RUAV modeling and controls, and identification of trends in UAV technology. Section III presents the review of modeling of RUAV using combined first principle and time-domain identification. Nonlinear dynamic modeling is presented based on first principle approach using X-cell 60 small scale helicopter as a test bed. A method for linearization procedure is elaborated to provide an analytical model for the implementation of linear control. Section IV is focused on discussion on simulation, control and guidance for UAV. Some approaches for control synthesis are demonstrated for illustration. The last section (V) identifies emerging technologies in the area of aerial robotic. Concluding remarks on the challenges and future directions are made in final section (VI). B. Lessons learned from CentrUMS-ITB UAV Program The research on RUAV at the Center for Unmanned Systems Studies (CentrUMS)-ITB was carried out by using a fully instrumented X-cell 60 SE model helicopter similar to one used by MIT team as shown in Fig. 1. The mini helicopter is characterized by a hinge-less rotor with a diameter of 0.775 m and mass of 8 kg. The X-Cell blades both for main and tail rotors use symmetric airfoils. The vehicle has been used by a number of research centers as published in a number of literatures (Gravilets,2003; Bogdanov,2003; Bogdanov,2004). Therefore comparison and validation can be achieved from the available published results. Using the test bed, studies on modeling and control of RUAV were conducted. A great deal of effort was focused on developing nonlinear model based on first principle approach. The nonlinear model was implemented in Simulink/Matlab with parameters are measured independently or obtained from literatures. Flight tests were conducted to validate the model. Various control synthesis were studied for performance comparison. The important lesson learnt from the experience is that a small scale helicopter is a intricate and unstable platform; to utilize it for a useful re- search test-bed there is a compelling need for development of mathematical model that capture the key dynamics of the vehicle with reasonable level of complexity for the purpose of control design. A number of key results are presented in Section IV. Figure 1: Instrumented X-Cell 60 SE- CentrUMS-ITB C. Trends in UAV Research More stringent mission requirements have driven the UAVs to have a higher level of autonomy dictated by a reliable onboard computer system. The metric for UAV level of autonomy is given in Table 2 (Sholes, 2006). Some key areas in current state-of-the-art aerial robotic technologies are responsible for enabling AUVs to achieve its required level of autonomy. Current status of UAV research activities in these areas can be summarized as the following: 1. State estimation algorithm. To achieve better performance, multiple sensors are typically fused together using EKF in a sensor fusion algorithm. Propagated IMU-data can be fused with discrete updates from GPS and altimeter. Several design examples are provided in (Johnson and Kannan, 2002). Recent study includes the use of nonlinear adaptive observers for estimating speed of UAV from IMU measurements only without the aid of GPS (Khadidja, 2007). 2. Simultaneous Localization and Mapping. An unmanned aerial vehicle (UAV) is tasked to explore an unknown environment and to map the features it finds, but must do so without the use of infrastructure-based localization systems such as GPS, or any a priori ter- Agus Budiyono International Symposium on Intelligent Unmanned System | October 15-18, 2008 3 ARTICLES RUAVs are then derived by the application of a time-domain parametric identification method to the flight data of target RUAVs. The classical control theory and modern linear robust control theory are applied to the identified model. The proposed controller are validated in a nonlinear simulation environment and tested in a series of test flights (Shim, 2000). rain data. A statistical estimation technique allows for the simultaneous estimation of the location of the UAV as well as the location of the features it sees. 3. The use of GPS as attitude sensor. The need for reduced complexity avionics system has driven the research on the use of single GPS for obtaining attitude estimate (Kornfeld, 1998). 5. Integrated modeling. Linear model is obtained by using combination of first principle results and time or frequency domain identification scheme. 6. Trajectory generation using maneuver automaton. Vehicle motion is described by library of motion primitives (Frazzoly et.al, 2005). The trajectory between two positions and vehicle states is found by searching the sequence of motion primitives which will best satisfy an objective function. One important application of guidance system is collision avoidance between vehicle at its tight and structured environment or between vehicles operating in formation or multi agent system. Safety verification. Safety verification or reachability analysis aims to show that starting at some initial conditions, a systems cannot evolve to some unsafe regions in the state space. Unsafe region for UAV application can be defined in the context of proximity to obstacles, fuel availability (endurance), un-flyable zone and/or communication range. A new concept called barrier certificate is being used for safety verification of hybrid systems. III. Modeling of RUAV The requirement for successful navigation and guidance task is stabilization of vehicle platform. Viewed as a multi-loop system, guidance and navigation is represented by the outer-loop and control and stabilization the inner loop. The design starts from the most inner loop outward. In this context, to control small scale helicopter as unstable platform with complex dynamics require sufficiently accurate model. This section elaborates the modeling technique and the corresponding model-based control synthesis. 4 UAV LEVEL OF AUTONOMY Vision for guidance. Computer vision is used as a feedback sensor in a control loop for an autonomous flight system. (Amidi et.al, 1998). More recent example is precision targeting without using secondary actuation or add-on gimbal system. 4. 7. TABLE I ; ; ; ; V ; . A. Methods of Modeling The approach to helicopter modeling can be in general divided into two distinct methods. The first approach is known as first principle modeling based on direct physical understanding of forces and moments balance of the vehicle. The challenge of this approach is the complexity of the mathematical model involved along with the need for rigorous validation. The method is primarily suitable for one with a strong background in flight physics. The second method based on system identification (Tischler and Cauffman, 1992; Mettler et.al., 2002, Tischler and Remple, 2006) basically arises from the difficulty of the former approach. The frequency domain identification starts with the estimation of frequency response from flight data recorder from an instrumented flight-test vehicle. The parameterized dynamic model can then be developed in the form of a linear state-space model using physical insight and frequency-response analysis. The identification can also be conducted in time-domain. In what follows, the author argues that, any modeling should start from adequate basis in first-principle. In practice, the above two methods can be used in an integrated scheme for developing an accurate small scale rotorcraft vehicle model for the purpose of control design. The modeling based on neural networks with appropriate structure and training method can be viewed as a viable alternative. Agus Budiyono International Symposium on Intelligent Unmanned System | October 15-18, 2008 B. Equation of Motion of RUAV The motion of a vehicle in three-dimensional space can be represented by the position of the center of mass and the Euler angles for the vehicle rotation with respect to the inertial frame of reference. The Euler-Newton equations are derived from the law of conservation of linear and angular momentum. Assuming that vehicle mass is m and inertial tensor I, the equations of motion are given by: j j dV =F m dt I j d ωj j I =M dt I ARTICLES Meanwhile, a new modeling scheme based on Linear Parameter Varying (LPV) identification is attractive for RUAV application. ∑ X = m ( u$ − rv + qw ) + mg sin θ ∑ Y = m ( ru + v$ − pw) − mg sin φ cosθ ∑ Z = m ( −qu + pv − w$ ) − mg cos φ cosθ ∑ L = I p$ − ( I − I ) qr ∑ M = I q$ − ( I − I ) pr ∑ N = I r$ − ( I − I ) pq xx yy yy zz zz (4) zz xx xx yy The forces and moments components consist of contribution from main rotor, tail rotor, fuselage, horizontal fin and vertical fin. 1) Main Rotor: The main rotor thrust equations are expressed as: TMR = ρ ( ΩR)MR (π R2 ) CT MR 2 (5) MR (1) j where F = [ X Y Z ]T is the vector of external forces actj ing on the helicopter center of gravity and M = [ L M N ]T is the vector of external moments. For helicopter, the external forces and moments consists of forces generated by the main rotor, tail rotor; aerodynamics forces from fuselage, horizontal fin and vertical fin and gravitational force. For computational convenience, the Euler-Newton equations describing the rigid-body dynamics of the helicopter is then represented with respect to body coordinate system by using the kinematic principles of moving coordinate frame of reference as the following: j j j j mV$ + m(ω × V ) = F j j j j I ω$ + (ω × I ω ) = M 1 ⎡1 ⎛1 1 2 ⎞ CT MR = aMRσ MR ⎢ ( µz MR − λ0MR ) + ⎜ + µMR ⎟θ0 2 2 ⎝3 2 ⎠ ⎣ (2) (6) and the inflow ratio, advance ratio and normal airflow component are respectively given by λ0MR ≡ ( ΩR )MR wiMR µ MR ≡ Here Here the vector V = [u v w]T and ωj = [ p q r ]T are the fuselage velocities and angular rates in the body coordinate system, respectively. For the helicopter moving in six degrees of freedom, the above equations produce six differential equations describing the vehicle’s translational motion and angular motion about its three reference axes. From here, we can express the mathematical expression for external forces and moments of the helicopter as a function of the control inputs and the vehicle states. j where the thrust coefficient is given by = 2η w µ ua2 + va2 µ z MR ( ΩR )MR σ , a and ηw are + ( λ0MR − µ z MR ) CT MR 2 MR 2 wa ≡ ( ΩR )MR (7) solidity ratio, lift curve slope and coefficient of non-ideal wake contraction of the main rotor. The above equations must be solved iteratively to obtain the thrust. The main rotor torque can be approximated as a resultant of induced torque due to generated thrust, and torque due to profile drag on the blade. QMR = ρ ( ΩR )MR (π R 2 ) 2 MR RMR CQ MR (8) where the torque coefficient is given by 1 ⎛ 7 2 ⎞ CQMR = σ MR ⎜ 1 + µ MR ⎟ CD0 MR + ( λ0MR − µ z MR ) CT MR 8 ⎝ 3 ⎠ (9) and CD0 is the profile drag coefficient of the main rotor. The representation of the main rotor tip path plane dynamics is given by Agus Budiyono International Symposium on Intelligent Unmanned System | October 15-18, 2008 5 τ ea$1s = −a1s + ∂a1s ∂a ua wa + 1s −τ eq + AδLo ∂µMR ( ΩR) MR ∂µz MR ( ΩR) MR τ eb$1s = −b1s − where a Bδ lat va ∂b1s − τ e p + Bδ Lat δ Lat ∂µMR ( ΩR ) MR X VF = 0 (10) and Aδ steady-state lateral and longitulong dinal gains from the cyclic inputs to the main rotor flap angles; δ lat and δ long are the lateral and longitudinal cyclic control inputs; τ e is the effective rotor time constant for a rotor with the stabilizer bar. 2) Tail Rotor: The tail rotor thrust can be computed by the following equation: TTR = mYδ r δ r + mYv vTR (11) And the normal velocity component to the tail rotor is vTR = va − lTR r + hTR p (12) The tail rotor torque is computed using similar equations for main rotor with tail rotor parameters substituted into the main rotor parameter. 3) Fuselage: For hover and low speed forward flight, the rotor downwash is deflected by the forward and side velocity. This deflection creates a force opposing the movement. The fuselage forces of the helicopter can be expressed as 1 X fus = − ρ S x fusV∞ ua 2 1 Yfus = − ρ S y fusV∞ va (13) 2 1 Z fus = − ρ S z fusV∞ ( wa − wiMR ) 2 4) Horizontal tail: The horizontal tail generates lift and a stabilizing pitching moment around the center of gravity. This will also compensate the destabilizing effect of the main rotor flapping due to vertical speed. The horizontal tail fin forces and moments of the helicopter referenced to body coordinate system are X HF = 0 YHF = 0 1 Z HF = − ρ S HF ( CLα HF ua + wHF ) wHF 2 1 2 Z HF = ρ S HF ( ua2 + wHF ) 2 (14) 5) Vertical tail: The vertical tail forces can be approximated by the following expression 6 1 YVF = − ρ S VF ( CLα VFV∞ VF + vVF ) vVF 2 1 2 YVF = ρ SVF (V∞2 VF + vVF ) 2 (15) C. First Principle Model The detailed equations of motion as presented previously are the basis for first principle modeling. It is a bottom-up physical modeling. A study by Weilenmann (1994) was an attempt to use first-principle approach to model the helicopter dynamics. The modeling however was limited only to hovering condition. Some simplified version of helicopter model existed including the Minimum-Complexity Helicopter Simulation Math Model (Heffley and Mnich, 1988) spanning from the previous work by Heffley et.al.(1979 and 1986). In 2003, Gavrilets (Gavrilets, 2003) presented a nonlinear model helicopter based on first principle approach used for an aerobatic maneuver control. The work however does not present workable procedures for developing linear model for the purpose of control design. The step-by-step development of linear model requires the calculation of a trim condition around which the vehicle motion will be linearized. The trim conditions for the helicopter are chosen operating points within which we solve j the equilibrium condition f ( xj , uj ) = 0 by first setting the states to the values which characterize the corresponding flight condition. For the case of RUAV, the solution of trim condition is achieved through an iterative process. The notion of stability derivatives used in the modeling arises from Taylor’s series expansion of external forces and moments around an equilibrium condition where only first order effects are retained. The external forces and moments are thus expressed in terms of product of derivatives and the rigid-body vehicle states and control inputs. The linearized equations of motion can finally be expressed in the form of state space readily usable for control synthesis. For more detail explanation, the readers are referred to (Budiyono, 2007b). As needed, the first principle model can also be refined by the system identification technique as presented in the following section. D. Identification Modeling The first principle approach typically requires the detail knowledge regarding the system behavior. The use of system identification modeling either in time or frequency domain on the other hand is more practical. The system identification approach requires experimental input-output data collected from the flight tests of the vehicle. Thus the flying test-bed must be outfitted with adequate instruments Agus Budiyono International Symposium on Intelligent Unmanned System | October 15-18, 2008 Given the time domain description of a system: y (t ) = G ( q )u (t ) + H ( q )e(t ) (16) and by observing the input (u) and output (y) data, the error, e(t) can be computed as: e(t ) = H −1 (q)[ y (t ) − G (q)u (t )] (17) PEM uses optimization to minimize the cost function, defined by: V N (G , H ) = ∑ e (t ) N 2 t =1 (18) The result of combined first principle and identification modeling is illustrated in Fig. 2. The figure shows the forward velocity flight data (solid thick line) compared with the first principle model (solid thin line) and identification model (dashed line). The figure shows that the fitness ratio of the flight data for first principle and identification model is 19.87% and 24.34% respectively. ARTICLES to measure both state and control variables. To utilize experimental data to build a parameterized model however, a model structure and decent initial conditions in the optimization scheme would be required to achieve convergence. The model structure and its initial value in this case can be provided by prediction of first principle calculation. In structured parameterization scheme, Predication Error Minimization (PEM) method can be utilized to estimate the parameters. With the method, the parameters of a model are chosen so that the difference between predicted output of the model and the measured output is minimized with the following process. TABLE II STABILITY DERIVATIVES COMPARISON BETWEEN FIRST PRINCIPLE PREDICTION AND IDENTIFICATION Yv Yr Yb1s Lu Lw Lv Lp Lb1s Nw Nv Np Nr Ba1s First Principle Prediction -0.3471 -16.5191 10.1395 -0.0106 0.1098 -0.2486 -40.8739 408.5485 1.0103 2.5045 0.1406 -0.9758 Identification -0.8652 -16.286 134.74 -0.03 0.0703 -0.217 1.3026 320.53 1.3669 2.1817 -1.2065 -0.695 0 / τe 0.0656 Further comparison between the first principle prediction and identification result is given in Table II. E. Linear Parameter Varying Identification All previous modeling schemes boil down to the development of linear model associated with a certain flight condition as shown in Fig. 3. The design of global nonlinear control is then predicated on the notion of gain scheduling. The drawback of this approach is that control designs based on linearized dynamics might become deteriorated when it is applied beyond the vicinity of equilibrium. In contrast, LPV control technique explicitly takes into account the change in performance due to real-time parameter variations. Therefore, this control technique gives a promising potential in designing control systems which is robust over the entire operating envelope. RUAVs FLIGHT CONDITIONS Piourette Accelerate Maneuvers Hover Cruise Deccelerate Ascend Descend Figure 2: Comparison of first principle and ID result Figure 3: RUAVs flight conditions Agus Budiyono International Symposium on Intelligent Unmanned System | October 15-18, 2008 7 25 Estimated Response Plant Response 20 u (m/s) 15 10 5 0 -5 0 50 100 150 200 250 300 time (seconds) Figure 4: Result of LPV identification for forward speed The LPV identification scheme employs recursive least square technique implemented on the LPV system represented by dynamics of helicopter during a transition. The airspeed as the scheduling of parameter trajectory is not assumed to vary slowly. The exclusion of slow parameter change requirement allows for the application of the algorithm for aggressive maneuvering capability without the need of expensive computation. Fig. 4 shows the result of LPV identification for varying forward speed. More detail account can be found in (Budiyono, 2008b). IV. Simulation, Control and Guidance To date various control techniques have been designed for rotorcraft vehicles ranging from classical single-output-single-output PID controller (Shim, 2000) to nonlinear (Koo, 1998; Boussios, 1998; Devasia, 1999; Buskey et.al., 2001, Harbick, 2004) and from non-aggressive flight (Corke et.al., 2000; Castillo et.al., 2005) to aggressive flight (Gavrilets et.al., 2001). To cover a wide region in the flight envelope, a gain schedule technique is typically employed as in Shamma and Athans (1991). A control using state-dependent Riccati equation was proposed by Bogdanov and Wan (2003) and Bogdanov et.al.(2003). The scheme was implemented on X-Cell helicopter (Bogdanov et.al, 2004). A control synthesis based on behavioral approach was suggested by Fagg et.al. (1993) and Buskey et.al. (2002,2003). Fuzzy (Jang and Sun, 1995) and adaptive control have been also synthesized for control of RUAV (Hovakimyan et.al. 2000; Johnson and Kannan, 2002; Kannan and Johnson, 2002; Kim et.al., 2002; Kutay et.al., 2002, Sanchez et.al., 2005). Bagnell and Schneider (2001) proposed a control using reinforcement learning. A Lyapunov control design was proposed by Mazenc et.al. 8 (2003). Overall, there exists a tendency in the area of RUAVs that more research has been done in control design methodologies than in developing dynamics model. The author argues that modeling is prerequisite of good control design. In order that a control system can be successfully designed and implemented for a vehicle (system), the dynamics characteristics of the vehicle must be well-understood. In line with this argument, Mettler (2003) viewed that the tendency to get around modeling efforts by searching for perfect control methodology is not productive and can even lead to inaccurate or misleading conclusions regarding the applicability or performance of certain control techniques. Flight simulation based on the developed model can be used to complement flight testing (Johnson et.al., 1996; Johnson and DeBitetto, 1997; Munzinger, 1998; Perhinschi and Prasad, 1998; Johnson and Fontaine, 2002; Johnson and Mishra, 2002; Lee and Horn, 2005). Guidance can be viewed as the most outer loop of multi-loop control system. A. Simulation environment for UAV Research in control engineering regularly produces new theoretical insights and algorithms that promise substantial improvement over the state of the practice. However, it is only a small fraction of this research that ultimately sees practical application (Samad et.al, 2004). The area of control for UAVs is not an exception. The need to close the gap between theory and application of control to UAVs in real operating conditions has been addressed by creating simulation environment where actual time-dependent signals are taken into account. Implementation and testing of control systems by a hardware-in-the-loop (HIL) simulation is increasingly being required for the design as it becomes a very versatile tool in acquiring ’real’ data without taking a risk of losing any expensive instrumented UAVs. HIL simulation is characterized by the operation of real components in connection with real-time simulated components. Usually, the control system hardware and software is the real system while the controlled plant can be either fully or partially simulated. The ‘high-confidence’ control can be achieved by developing increasingly higher fidelity models and simulations through successive improvements. It should be ensured that the plant model is a sufficiently accurate approximation of reality and that assumptions about disturbances and the operational environment are valid. The implementation of HILS for various RUAVs at Smart Robot Center (Konkuk University) is illustrated in Fig. 5. Agus Budiyono International Symposium on Intelligent Unmanned System | October 15-18, 2008 ARTICLES Motion table δlong IMU Iron Bird δ lat φ,θ,ψ δ col RT simulink execution Bidirectional communication Host PC δ ped xPC Windows RS232 Flight Gear Cockpit view RS232 PC‐104 Flight control exec Sensor signal processor GCS communication Matlab/Simulink 6 DOF heli nonlinear mdl RTW/XPC Target appl. Health monitoring Interactive autopiloting Ground control GUI High level controller Hardware In the Loop (HIL) Simulator Figure 5:Simulation environment for UAV control synthesis B. Control Synthesis Given the sufficiently accurate model, the control synthesis of RUAV can be conducted and validated within real-time simulation environment. Various control techniques have been developed thus far in Budiyono (2005a, 2005b) and Budiyono et.al. (2004, 2005, 2007a). Referring to the taxonomy of flight conditions of RUAV (Fig. 3), the control design can be classified into the following different approaches: 1. Classical control. Since the problem of RUAV control is a MIMO problem, the design procedure of classical approach is to be conducted in cascaded multi-loop SISO system starting from the innermost loop outward. The cascaded multi-loop SISO approach however has limitations in its implementation. To implement this control approach for a small scale helicopter, a pitch and roll attitude control system is often subordinated to a, respectively, longitudinal and lateral velocity control system in a nested architecture. The requirement for this technique to work is that the inner attitude control loop must have a higher bandwidth than the outer velocity control loop. While this is va- lid for a relatively large unmanned helicopter such as Yamaha R-50, for a class of high-performance helicopters, such as the X-Cell 60, or helicopters where this bandwidth separation is not sufficient, a simultaneous design will be necessary (Mettler, 2003). The simultaneous design is provided by modern control synthesis. 2. Modern MIMO control. To control a model helicopter as a complex MIMO system, an approach that can synthesize a control algorithm to make the helicopter meet performance criteria while satisfying some physical constraints is required. To address a MIMO problem, LQR and H ∞ are the most popular control design procedures. These methods however also have drawbacks that can inhibit a practical implementation. They include dealing with higher than necessary order of controller, non-existence of formal parameter tuning and weight selection procedures, possible exclusion of good controllers, and difficulty in integrating state variable constraints (Manabe, 2002). 3. Algebraic control. The CDM is one of such approaches where control design process is based on Agus Budiyono International Symposium on Intelligent Unmanned System | October 15-18, 2008 9 coefficient diagram representing criteria of good design. The use CDM thus far has been limited to SISO or SIMO applications. Some trial designs for MIMO have been made (Manabe, 2002), but formal design procedures to implement CDM for MIMO has not been established yet. The typical approach in solving MIMO problem thus far has been to decompose MIMO problems into series of SISO or SIMO problems and proceed with design by standard CDM. The first attempt that demonstrates a successful implementation of CDM-based LQR technique without the need of decomposing a MIMO problem into a series of SISO or SIMO problems was presented in (Budiyono, 2007). Fig. 6 shows the result of design for step response of u and w subjected to 30% parameter variation. Speed w 25 20 15 ft/sec 10 5 0 -5 -10 -15 u (m/s) 0.6 0.4 nominal -30% in xu,xa,mq +30% in xu,xa,mq 10 20 30 40 50 60 70 80 90 50 60 70 80 90 t (s) 1 w (m/s) 0.8 0.6 0.4 0.2 0 -0.2 0 10 20 30 40 t (s) Figure 6: CDM-LQR control design 10 20 25 30 LPV approach. The control design is performed based on the model developed through LPV identification. Model Predictive Control (MPC) can be a good candidate for such an approach. A. Bio-inspired Technologies and Biorobotics 1 0 15 Issues pertaining to increased demand for higher performance and safety have pushed the UAV design beyond conventional approaches. Some emerging technologies can be summarized in the following paragraph. 0.8 -0.2 10 V. Emerging Technologies 1.2 0 5 Figure 7: Comparison of Switched Linear Control and LQR Hybrid approach. In the hybrid approach, each linear model in Fig. 3 can be considered as a hybrid automaton. To represent an RUAV flying over wider flight envelope therefore, the approach leads to a switching problem representing a change from one mode to another. A synthesis of switched control systems for model helicopter excited with external switches that bring changes of dynamics from hover to cruise by satisfying some constraint in the trajectories can thus be performed. Piecewise quadratic Lyapunov-like functions that leads to linear matrix inequalities (LMIs) for performance analysis and controller synthesis can be considered. State jumps of the controller responding to switched of plant dynamics are exploited to improve control performance (Sutarto et.al., 2006). The result is illustrated in Fig. 7 showing comparison between performance of LQR and Switched Linear Control. 0.2 0 Time (Second) 5. 4. LQR Output SLC State SLC The emerging field of unmanned system technologies largely relies on the ability of an onboard mechanism that replaces or imitates a human operator. To successfully design an unmanned system or vehicle therefore it is important to study the human intelligent at all levels: reasoning, perception, development and learning. Moreover, the compelling need to learn from nature stems from the fact that although the present conventional approach to engineering design may exceed nature in some regards, they are not superior to many designs in nature. Using conventional approach, present day UAVs can perform different control functions including altitude and speed hold, obstacle avoidance, terrain following navigation, and autonomous landing. Flying insects can perform all those and beyond, remarkably well using ingenious strategies for perception and navigation in three dimensions. Insects infer distances to potential obstacles and objects of interest from image motion cues that result from their own motion in the environment. The angular motion of texture in images is denoted generally as optic or optical flow. Computationally, a strategy based on optical flow is simpler than is stereosco- Agus Budiyono International Symposium on Intelligent Unmanned System | October 15-18, 2008 Recent studies also demonstrate that insects can perform extreme maneuvering capabilities far beyond those achieved by conventional UAVs. Flapping wing, morphing wing, formation flight, neuro-control and swarming are just a few examples of natural phenomena much related to UAVs’ advanced design features. More research should be consistently conducted for harvesting design principles from nature that would extend present UAV technologies out of its conventional boundaries. fication. Future challenges for advancing aerial robotics technology will be pivoted on exploitation of biomimetic principles for achieving higher performance and development of formal model and analysis tool to synthesize collaborative aerial robotics behavior. References 1. nomous helicopter research at Carnegie Mellon Robotics Institute, Proceedings of Heli Japan ’98, Gifu, Japan, Paper No: T7-3. 2. Cooperative multi-agents naturally lead to hybrid system abstraction. The hybrid model would capture both UAV dynamics and mode switching logic that supervises lower level control switches. It will be desirable in this regards to have a formal tool that can verify the performance and safety of such a system where high fidelity simulation can be conducted prior to flight tests. Future research direction in multi UAVs system should address this need. Azinheira, J.R et al. (2008), A backstepping controller for path-tracking of an underactuated autonomous airship, Int. J. Robust B. Multi UAV Systems One primary feature of high autonomy UAVs is their ability to perform coordination and cooperation functions. This capability is termed Level 5 and 6 in Table 2. Research in this area (collaborative sensing and exploration, synchronized motion planning, and formation or cooperative control) has been gaining more interests in recent past as shown for example in (Seiler, 2001) and Mot et al. (2002a, 2002b). A particular class of tasks for such multi-agent UAV systems involve surveillance of a region and tracking of targets cooperatively. Cooperative agents are typically desired to handle a particular task with higher robustness, higher performance (faster or more accurately) or task simply otherwise unattainable by single agent. UAVs formation control can be achieved through hierarchical (leader-follower) or non-hierarchical approach. Amidi, O., Kanade, T., and Miller, J. R. (1998) : Vision-based auto- Nonlinear Control 3. Azinheira, J.R and Moutinho, A (2008), Hover Control of an UAVWith Backstepping Design Including Input Saturations, IEEE Transactions On Control Systems Technology, Vol. 16, No. 3 4. Bagnell, J. A. and Schneider, J. G. (2001) : Autonomous Helicopter Control using Reinforcement Learning Policy Search Methods, Proceedings of the International Conference on Robotics and Automation 2001, IEEE, pp. 1615-1620. 5. Bogdanov, A. and Wan, E. (2003) : SDRE Control With Nonlinear Feedforward Compensation for A Small Unmanned Helicopter, AIAA, Paper No. : 2003-6512. 6. Bogdanov, A., Wan, E. and Harvey G (2004) : SDRE Flight Control For X-Cell and R-Max Autonomous Helicopters, Proceedings of the 43rd IEEE Conference on Decision and Control, IEEE, Atlantis, Paradise Island, Bahamas, pp. 1196- 1203. 7. Bogdanov, A., Carlsson, M., Harvey, G., Hunt, J., Kieburtz, D., Merwe, R. V. D. and Wan, E. (2003) : State-Dependent Riccati Equation Control of A Small Unmanned Helicopter, Proceedings of the AIAA Guidance Navigation and Control Conference, AIAA, Austin, TX, pp. 1120-1126. 8. Boussios, C. I. (1998) : An Approach for Nonlinear Control Design via Approximate Dynamic Programming, PhD thesis, Massachusetts Institute of Technology. 9. Budiyono, A. and Sutarto, H.Y. (2004) : Controller Design of a VTOL Aircraft: A Case Study of Coefficient Diagram Method to a Time-varying System, Regional Conference on Aeronautical Science, Technology and Industry, Bandung, Indonesia. VI. Concluding Remarks The paper discussed recent progress in the technology for unmanned aerial vehicles from the modeling, control and guidance perspectives. Dynamics of rotorcraft-based unmanned aerial vehicle is presented to describe the underlying principle of modeling for the control synthesis. The modeling based on first principle, system identification and LPV identification is presented briefly for illustration. A number of major trends in aerial robotics are discussed: state estimation algorithm, SLAM, vision for guidance, integrated modeling, maneuver automaton and safety veri- 10. Budiyono, A. (2005a) : Onboard Multivariable Controller Design for a Small Scale Helicopter Using Coefficient Diagram Method, International Conference on Emerging System Technology, Seoul, Korea. 11. Budiyono, A. (2005b) : Design and Development of Autonomous Uninhabited Air Vehicles at ITB: Challenges and Progress Status, Aerospace Indonesia Meeting, Bandung, Indonesia. 12. Budiyono, A. and Wibowo, S.S. (2007a) : Optimal Tracking Controller Design for A Small Scale Helicopter, Journal of Bionic Engineering, Vol 4, December. 13. Budiyono, A. et.al. (2007b) : First Principle Approach to Modeling of Small Scale Helicopter, in Proceedings of International Conference on Intelligent Unmanned Systems, Bali, Indonesia. Agus Budiyono International Symposium on Intelligent Unmanned System | October 15-18, 2008 11 ARTICLES py for avoiding hazards and following terrain (Thakoor, S. et.al. , 2003). 14. Budiyono, A. et.al (2008): Integrated Identification Modeling of 28. Harbick, K., Montgomery, J. F. and Sukhatme, G. S. (2004) : Planar Rotorcraft-based Unmanned Aerial Vehicle, accepted for IEEE Ro- Spline Trajectory Following for an Autonomous Helicopter, Journal bio, Bangkok of Advanced Computational Intelligence - Computational Intelli- 15. Budiyono, A and Sudiyanto, T (2008b): Linear Parameter Varying gence in Robotics and Automation, 8, pp. 237-242. Identification of Vertical-Longitudinal Dynamic of A Small Size 29. Heffley, R. K., Jewell, W. F., Lehman, J. M. and Winkle, R. A. V. Helicopter (XCell 60) Model, International Symposium on Intelli- (1979) : A Compilation and Analysis of Helicopter Handling Qualities Data, NASA Contractor Report 3144. gent Unmanned System, Nanjing 16. Buskey, G., Wyeth, G. and Roberts, J. (2001) : Autonomous Heli- 30. Heffley, R. K., Bourne, S. M., Curtiss Jr, H. C., Hindson, W. S. and copter Hover Using an Artificial Neural Network, International Hess, R. A. (1986) : Study of Helicopter Roll Control Effectiveness Conference on Robotics & Automation (ICRA 2001), Seoul, Korea, Criteria, NASA Contractor Report 177404. pp. 1635-1640. 31. Heffley, R. K. and Mnich, M. A. (1988) : Minimum Complexity 17. Buskey, G., Roberts, J. and Wyeth, G. (2002) : Online Learning of Autonomous Helicopter Control, Proceedings Australasian Conference on Robotics and Automation, Auckland, pp. 21-27. Helicopter Simulation Math Model, NASA Contractor Report 177476. 32. Hovakimyan, N., Kim, N., and Calise, A. J. (2000) : Adaptive output 18. Buskey, G., Roberts, J. and Wyeth, G. (2003) : A helicopter named feedback for high-bandwidth control of an unmanned helicopter, Dolly - Behavioral cloning for autonomous helicopter control, Pro- AIAA Guidance, Navigation and Control Conference, Paper No: ceedings Astralasian Conference on Robotics and Automation, Bris- AIAA-2000-4058. bane, pp. 36-41. 33. Jang, J. R. and Sun, C. T. (1995) : Neuro-Fuzzy Modeling and Con- 19. Castillo, C., Alvis, W., Castillo, M.-Effen, Valavanis, K. and Moreno, trol, Proceedings of The IEEE. pp. 378-406. W. (2005) : Small Scale Helicopter Analysis and Controller Design 34. Johnson, E. N. and Mishra, S. (2002) : Flight Simulation for the for Non-Aggressive Flights, Proceedings IEEE International Confe- Development of an Experimental UAV, Proceedings of the AIAA rence on SMC, Hawaii, pp. 3305- 3312. Modeling and Simulation Technologies Conference and Exhibit, 20. Corban, J. E., Calise, A. J., Prasad, J. V. R., Hur, J., and Kim, N. Monterey, California, Paper No: AIAA 2002-4975. (2002) : Flight evaluation of adaptive high bandwidth control me- 35. Johnson, E. and Kannan, S. (2002) : Adaptive flight control for an thods for unmanned helicopters, Proceedings of the AIAA Guidance, autonomous unmanned helicopter, AIAA Guidance, Navigation and Navigation and Control, American Institute of Aeronautics and As- Control Conference, AIAA-2002-4439, Monterey, California, Paper tronautics, pp. 645-651. No: AIAA 2002-4439. 21. Corke, P., Sikka, P. and Roberts J. (2000) : Height Estimation for an Autonomous Helicopter, International Symposium on Experimen- tal Robotics, pp. 101–110. 36. Johnson, E. and Fontaine, S. (2001) : Use of flight simulation to complement flight testing of low-cost UAVs, AIAA Modeling and Simulation Technologies Conference, Paper No. : AIAA 2001-4059. 22. Devasia, S. (1999) : Approximated Stable Inversion for Nonlinear 37. Johnson, E., DeBitetto, P., Trott, C., and Bosse, M. (1996) : The Systems with Nonhyperbolic Internal Dynamics, IEEE Transactions 1996 MIT/Boston University/Draper laboratory autonomous heli- on Automatic Control, 44, pp. 1419-1425. 23. Escareno, J. et.al. (2008): Triple Tilting Rotor mini-UAV: Modeling and Embedded Control of the Attitude, American Control Conference, Seattle, Washington, USA,June 11-13 24. Fagg, A. H., Lewis, M. A., Montgomery, J. F. and Bekey, G. A. copter system, 15th AIAA/IEEE Digital Avionics System Conference, 1, pp. 381-386. 38. Johnson, E. and DeBitetto, P. (1997) : Modeling and simulation for small autonomous helicopter development, AIAA Modeling and Simulation Technologies Conference, Monterey, California. (1993) : The USC Autonomous Flying Vehicle : An Experiment In 39. Kannan, S. and Johnson, E. (2002) : Adaptive Trajectory Based Real-time Behavior-Based Control, Proceedings of the 1993 Control For Autonomous Helicopters, AIAA Digital Avionics Con- IEEE/RSJ International Conference on Intelligent Robots and Systems, pp. 1173-1180. 25. Frazzolli, E., Dahleh, M. A., and Feron, E. (2005) : Maneuver-Based Motion Planning for Nonlinear Systems With Symmetries, IEEE Transaction on Robotics, 21, pp. 1077—1091. 26. Gavrilets, V., Frazzoli, E., Mettler, B., Piedmonte, M., and Feron, E. (2001) : Aggressive maneuvering of small autonomous helicopters: a human-centered approach, International Journal of Robotics Research, pp. 795 - 807. 27. Gavrilets, V. (2003) : Autonomous Aerobatic Maneuvering of Miniature Helicopter, PhD thesis, Massachusetts Institute of Technology. ference, number 358, Irvine, CA, pp. 8.D.1-1 – 8.d.1-12. 40. Khadidja, et.al, (2007): Unmanned Aerial Vehicle Speed Estimation via Nonlinear Adaptive Observers, Proceedings of the 2007 American Control Conference 41. Kim, H. J., Shim, D. H. and Sastry, S. (2003) : A Flight Control System for Aerial Robots : Algorithms and Experiments, Control Engineering Practice, 11, pp. 1389-1400. 42. Kim, N., Calise, A. J., Hovakimyan, N., Prasad, J.V.R., and Corban, E. (2002) : Adaptive Output Feedback for High Bandwidth Flight Control, AIAA Journal of Guidance, Control, and Dynamics, 25(6), pp. 993-1002. 43. Koo, T. J., Pappas, G. J. and Sastry, S. (2001) : Mode Switching Synthesis for Reachability Specifications, Proceedings of the 4th In- 12 Agus Budiyono International Symposium on Intelligent Unmanned System | October 15-18, 2008 57. Mot, J. D., Kulkarni, V., Gentry, S., Gavrilets, V. and Feron, E. (2002a) : Coordinated Path Planning for a UAV Cluster, The First pp. 333 – 346. 44. Koo, T. J. and Sastry, S. (1998) : Output Tracking Control Design of AINS Symposium, UCLA, Los Angeles, CA. A Helicopter Model Based on Approximate Linearization, Proceed- 58. Mot, J. D., Kulkarni, V., Gentry, S. and Feron, E. (2002b) : Spatial ings Of the 37th IEEE Conference on Decision and Control, Tampa, Distribution Results for Efficient Multi-Agent Navigation. IEEE FL, pp. 3635-3640. Conference on Decision and. Control, 4, pp. 3776 – 3781. 45. Kornfeld, R. (1999) : The Impact of GPS Velocity Based Flight 59. Munzinger, C. (1998) : Development of A Real-Time Flight Simula- Control on Flight Instrumentation Architecture, PhD thesis, Massa- tor for An Experimental Model Helicopter, Diploma Thesis, Georgia chusetts Institute of Technology. Institute of Technology. 46. Kutay, A. T., Calise, A. J., Idan, M. and Hovakimyan, N. (2002) : 60. Perhinschi, M. G. and Prasad, J. V. R. (1998): A simulation model of Experimental Results on Adaptive Output Feedback Control Using A an autonomous helicopter, Proceedings of RPV/UAV Systems Bris- Laboratory Model Helicopter, Proceedings of the AIAA Guidance, Navigation and Control Conference, pp. 196-202. 47. La Civita, M., Messner, W. C. and Kanade, T. (2002) : Modeling of small-scale helicopters with integrated first-principles and system identification techniques, Proceedings Of the 58th Forum of the American Helicopter Society, Montreal, Canada, pp. 2505 - 2516. 48. Lee, D. and Horn, J. F. (2005) : Simulation of pilot workload for a helicopter operating in a turbulent ship airwake, Proceedings of the Institution of Mechanical Engineers part G, 219, pp. 445-458. 49. Madani, T and Bennelague, A. (2007): Sliding Mode Observer and Backstepping Control for a Quadrotor Unmanned Aerial Vehicles, American Control Conference, New York, USA, June 11-13 50. Mahony, R. and Hamel, T. (2004) : Robust trajectory tracking for a scale model autonomous helicopter, International Journal of Robust and Nonlinear Control, 14, pp. 1035-1059. 51. Manabe, S. (2002) : Application of Coefficient Diagram Method to MIMO Design in Aerospace, 15th Triennial World Congress, IFAC, Barcelona, Spain, T-Tu-MO62. 52. Mazenc , F., Mahony, R. E. and Lozano, R. (2003) : Forwarding control of scale model Autonomous helicopter: A Lyapunov control design, Proceedings of the 42nd IEEE Conference on Decision and Control , Maui, Hawaii USA, pp. 3960- 3965. 53. Mettler, B. (2003): Identification modeling and characteristics of miniature rotorcraft, Kluwer Academic Publisher tol International Conference and Exhibit, pp. 36.1-36.13 61. Raffo, G.V. et al. (2008): Backstepping/Nonlinear H∞ Control for Path Tracking of a QuadRotor Unmanned Aerial Vehicle, American Control Conference, Seattle, Washington, USA, June 11-13 62. Samad, T. et.al. (2004): High-Confidence Control: Ensuring Reliability in High-Performance Real-Time Systems, International Journal of Intelligent Systems, Vol. 19, 315–326 63. Sanchez, E. N., Becerra, H. M. and Velez, C. M. (2005) : Combining fuzzy and PID control for an unmanned helicopter, The 2005 North American Fuzzy Information Processing Society Annual Conference, Ann Arbor, Michigan, USA. 64. Seiler, P. J. (2001) : Coordinated Control of Unmanned Aerial Vehicles, PhD thesis, University of California, Berkeley. 65. Shamma, J. S. and Athans, M. (1991) : Gain Scheduling : Potential Hazards and Possible Remedies, American Control Conference, Boston, MA. 66. Shim, D. (2000) : Hierarchical Control System Synthesis for Rotorcraft-Based Unmanned Aerial Vehicles, PhD thesis, University of California, Berkeley. 67. Sholes, Eric. (2006): Evolution of a UAV Autonomy Classification Taxonomy, IEEEAC paper #1538, Version 3 68. Sutarto, H.Y., Budiyono A., Joelianto E., and Hiong, G. T. (2006) : Switched Linear Control of a Model Helicopter, International Conference on Automation, Robotics, Control and Vision, Singapore. 54. Mettler, B., Tischler, M., and Kanade, T. (2002) : System identifica- 69. Thakoor, S. et.al (2003): Review: The Benefits and Applications of tion modeling of a small-scale unmanned rotorcraft for flight control Bioinspired Flight Capabilities, Journal of Robotic Systems 20(12), design, Journal of the American Helicopter Society, 47, pp. 50 – 63. 687–706 55. Mokhtari, A and Bennelague, A (2004): Dynamic Feedback Control- 70. Tischler, M. B. and Cauffman, M. G. (1992) : Frequency-Response ler of Euler Angles and Wind parameters estimation for a Quadrotor Method for Rotorcraft System Identification : Flight Applications to Unmanned Aerial Vehicle, Proceedings of IEEE International Con- BO-105 Coupled Fuselage/Rotor Dynamics, Journal of the American ference on Robotics and Automation, New Orleans, LA, USA 56. Mokhtari, A, Bennelague, A. and Daachi, B (2005) : Robust Feedback Linearization and GH∞ Controller for a Quadrotor Unmanned Aerial Vehicle, IEEE/RSJ International Conference on Intelligent Robots and Systems Helicopter Society, 37/3: p. 3-17. 71. Tischler, M. B. and Remple, R. K. (2006) : Aircraft and Rotorcraft system Identification, AIAA Education Series. 72. Weillenmann, M. F. and Geering, H.P., (1994): Test Bench for Rotorcraft Hover Control, AIAA Journal of Guidance, Control and Dynamic, 17, pp. 729-736. . Agus Budiyono International Symposium on Intelligent Unmanned System | October 15-18, 2008 13 ARTICLES ternational Workshop on Hybrid Systems: Computation and Control,