Aircraft-Pilot System Modeling and Pilot Control Behavior Research for Airdrop Task

1st IFAC Conference on Cyber-Physical & Human-Systems 1st IFAC Conference on Cyber-Physical & Human-Systems 1st IFAC Conference on Cyber-Physical & Human-Systems December 7-9, 2016. Florianopolis, Brazil Available online at www.sciencedirect.com December 7-9, 2016. Florianopolis, Brazil 1st IFAC Conference on Cyber-Physical December 7-9, 2016. Florianopolis, Brazil& Human-Systems December 7-9, 2016. Florianopolis, Brazil

ScienceDirect

IFAC-PapersOnLine 49-32 (2016) 177–182

Aircraft-Pilot System Modeling and Pilot Control Behavior Aircraft-Pilot System Modeling and Pilot Control Behavior Aircraft-Pilot System Modeling and Pilot Control Behavior Aircraft-Pilot System Modeling and Pilot Control Behavior Research for Airdrop Task Research for Airdrop Task Research for Airdrop Task Research Airdrop TaskSun Liguo Tan Wenqian, Li Zongyuan,for Qu Xiangju, Wang Yanyang, Tan Wenqian, Li Zongyuan, Qu Xiangju, Wang Yanyang, Sun Liguo Tan Wenqian, Li Zongyuan, Qu Xiangju, Wang Yanyang, Sun Liguo School of Aeronautic Science and Engineering, Beihang University, 100083, China Tan Wenqian, Li Zongyuan, Qu Xiangju, Wang Yanyang,Beijing Sun Liguo School of Aeronautic Science and Engineering, Beihang University, Beijing 100083, China School of Aeronautic Science and Engineering, Beihang University, Beijing 100083, China School of Aeronautic Science and Engineering, Beihang University, Beijing 100083, China

Abstract Abstract Abstract Adverse aircraft-pilot coupling during the process of transport airdrop can lower the level of flight safety. Pilot’s Adverse aircraft-pilot coupling during the process airdrop can lower the level of flight Pilot’s Abstract Adverse aircraft-pilot theaircraft-pilot process of of transport transport airdrop lower levelof ofaircraft-pilot flight safety. safety. system Pilot’s rational control behaviorcoupling can avoidduring adverse coupling. In thiscan paper, thethe model rational control behavior can avoid adverse aircraft-pilot coupling. In this paper, the model of aircraft-pilot system Adverse aircraft-pilot coupling during the process of transport airdrop can lower the level of flight safety. Pilot’s rational control behavior avoidtask adverse coupling. In at this paper, thedifferent model ofpilot aircraft-pilot system consisting of motion bodycan model, modelaircraft-pilot and pilot model is built first. Three control methods consisting of motion body model, task model and pilot model is built at first. Three different pilot control methods rational control behavior can avoid adverse aircraft-pilot coupling. In this paper, the model of aircraft-pilot system consisting of motion body model, task model pilot model built at first. Three different pilot control are investigated for airdrop. The first one is and continuous pitchis compensatory control, the second one is methods elevator are investigated for The first one is continuous pitch compensatory control, the second one elevator consisting of motion body model, task model pilot model built at first. Three different pilot control are investigated for airdrop. airdrop. is and continuous pitchisbased compensatory control, the one is is methods elevator maintenance control, and the The thirdfirst oneone is precognitive control on the knowledge of second flight tendency during maintenance control, and the third one is precognitive control based on the knowledge of flight tendency during are investigated for airdrop. The first one is continuous pitch compensatory control, the second one is elevator maintenance the third one is precognitive control based Models on the knowledge of flight tendency during airdrop. Two control, differentand precognitive control methods are designed. of those pilot control methods are airdrop. Two different precognitive control methods are designed. Models of those pilot control methods are maintenance control, and the third one is precognitive control based on the knowledge of flight tendency during airdrop. Two different precognitive control methods are designed. Models of those pilot control methods area constituted and numerical simulations are conducted. In addition, for exploration on feasibility of those methods, constituted and numerical simulations are conducted. In addition, for exploration on feasibility of those methods, airdrop. Two different precognitive control methods are designed. Models of those pilot control methods areaa constituted and numerical simulations are conducted. In addition, exploration on feasibility of thoseexperiments methods, series of pilot-in-loop real-time experiments are performed in aforfixed flight simulator. Real-time series of pilot-in-loop real-time experiments are performed in a fixed flight simulator. Real-time experiments constituted and numerical simulations are conducted. In addition, for exploration on feasibility of those methods, a series pilot-in-loop real-time experiments performed a fixedeffects flight of simulator. Real-time experiments match of well with numerical simulation results are which validate in different those methods on airdrop. The match well with simulation results which validate different those methods on airdrop. The series of pilot-in-loop real-time experiments are performed a fixedeffects flight of simulator. Real-time experiments match with numerical numerical simulation results validate in different effects of those The aircraftwell performs most satisfactorily under thewhich two precognitive control methods. The methods study is on of airdrop. considerable aircraft performs most the two precognitive control methods. The methods study of considerable match well with numerical simulationunder results validate different effects of those The aircraft performs most satisfactorily satisfactorily under thewhich two precognitive methods. study is is on of airdrop. considerable significance to investigate of which kind of pilot control method to control be applied during The airdrop. significance to of of control method be during airdrop. aircraft performs most satisfactorily under the two precognitive methods. study is of considerable significance to investigate investigate of which which kind kind of pilot pilot control method to to control be applied applied during The airdrop. significance to investigate of aircraft-pilot which kind ofsystem control method to Hosting be applied airdrop. Keywords: transport airdrop, modeling, aircraft-pilot coupling, pilot control . © 2016, IFAC (International Federation ofpilot Automatic Control) byduring Elsevier Ltd. Allbehavior rights reserved. Keywords: Keywords: transport transport airdrop, airdrop, aircraft-pilot aircraft-pilot system system modeling, modeling, aircraft-pilot aircraft-pilot coupling, coupling, pilot pilot control control behavior behavior .. Keywords: transport airdrop, aircraft-pilot system modeling, aircraft-pilot coupling, controlasbehavior . during problems. Leaving thepilot controls they are 1. Introduction ii problems. Leaving the controls as they are during 1. Introduction i problems. Leaving the controls as they are during 1. Transport Introduction airdrop is recommended as a result. There is not an aircraft is an effective vehicle for dealing i airdrop is recommended as a result. Thereare is not an Transport aircraft is an effective vehicle for dealing problems. Leaving the controls as they during 1. Introduction airdrop is method recommended result. There not an Transport aircraft is an airdrop effectiveisvehicle effective given as bya Thomas J. tois control with emergency. Cargo one offor thedealing main effective method given by Thomas J. to control with emergency. Cargo airdrop is one of the main airdrop is recommended as a result. There is not an Transport aircraft is an effective vehicle for dealing effective method given by Thomas J. to control with emergency. Cargo airdrop is one of the main aircraft. However, the violent oscillation brought by applications of military transport aircraft. It also has aircraft. However, the violent oscillation by applications of military transport aircraft. alsomain has effective method by Thomas J. brought toa simple, control with emergency. Cargo airdrop one paper ofIt aircraft. needs However, thecontrolled. violent oscillation brought by applications of military transport Itthe also has airdrop to begiven Therefore, practical applications in civil field.isaircraft. This focuses airdrop needs to be Therefore, aa simple, practical applications in civil field. This paper focuses aircraft. However, violent brought by applications of military transport aircraft. It also has airdrop needs tomethod bethecontrolled. controlled. Therefore, simple, practical in[1] civil field. Thisaltitude paper focuses easy mastered will beoscillation put forward in this on heavyapplications cargo airdrop . Since high airdrop [1] easy mastered method will be Therefore, put forward in this on heavyapplications cargo airdrop . Since high altitude airdrop [1] airdrop needs to be controlled. a simple, practical in civil field. This paper focuses [2] easy mastered method will be put forward in this ononly heavy cargo airdrop . Since high altitude airdrop paper to minimize the deviations of flight states. is available for light equipment or parachutists , [2] [1] paper to the states. is available for or parachutists [2], easy mastered method will be of putflight forward ononly heavy cargo airdrop .equipment Since airdrop paper to minimize minimize the deviations deviations of flight states.in this is only available for light light equipment oraltitude parachutists low altitude airdrop reduces the high vertical velocity of, [2] low altitude airdrop reduces the vertical velocity of paper to minimize the deviations of flight states. is only available for light equipment or parachutists , 2. Modeling of aircraft-pilot system for airdrop low airdrop reduces the of vertical velocity cargoaltitude and ensures good condition cargo as well asofa 2. Modeling of aircraft-pilot system for airdrop cargo and ensures good condition of cargo as well as aa 2. In Modeling ofstudy aircraft-pilot system for airdrop low altitude airdrop reduces the vertical velocity of order to the characteristics of aircraft-pilot cargo and ensures good condition of cargo as well as higher precision of airdrop field. Thus for heavy cargo order to the of aircraft-pilot 2. In Modeling aircraft-pilot system for higher precision ofgood airdrop field. Thus for heavy In order(APC) toofstudy study the characteristics characteristics aircraft-pilot cargo ensures ofaltitude cargo asairdrop wellcargo as isa coupling during the process of airdrop airdrop, the higherand precision of airdrop field.low Thus for heavy cargo airdrop, low altitude orcondition super coupling (APC) during the process of airdrop, In order to study the characteristics aircraft-pilot airdrop, low altitude or super low altitude airdrop is coupling (APC) during the process of airdrop, the the higher precision of airdrop field. Thus for heavy cargo aircraft-pilot system is modeled at first. airdrop, low altitude or super low altitude airdrop is the first choice. However, low altitude airdrop may be aircraft-pilot system is modeled at first. coupling (APC) during the process of airdrop, the the first choice. However, low altitude airdrop may be aircraft-pilot system is modeled at first. airdrop, low altitude or super low altitude airdrop is The structure of aircraft-pilot system model shown the first choice. However, low altitude airdrop may be affected by complex atmosphere disturbance， which The structure of aircraft-pilot system model shown aircraft-pilot system is modeled at first. affected by complex atmosphere disturbance， which The structure of aircraft-pilot system model shown the first However, altitude airdrop may be in figure 1 includes three elements: motion body affected by complex atmosphere disturbance， which has greatchoice. disadvantage on low flight. A rational control of in figure 11 includes three elements: motion body The structure of aircraft-pilot system Each model shown has greatby disadvantage on flight. disturbance， A rational control of in figure includes three motion body affected complex which model, task model and pilotelements: model. model’s has great disadvantage on flight. A safety. rational control of pilot can enhance theatmosphere level of flight model, task model and pilot model. Each model’s in figure 1 includes three elements: motion body pilot can enhance the level of flight safety. model, task model and pilot model. Each model’s hasThe great disadvantage on flight. A rational control of function is introduced in detail. pilot can enhance the level of flight safety. mass, as well as the moment of inertia and the function is in model, task model and pilot model. Each model’s Thecan mass, as well as theofmoment of inertia and the function is introduced introduced in detail. detail. pilot enhance the level flight safety. The mass, as well as the moment of inertia and the centroid of the carrier aircraft, changes during function is introduced in detail. 2.1. Motion Body Model centroid of the carrier aircraft, changes during The mass, as well as the of inertia and the 2.1. Motion Body Model centroid of of the carrier aircraft, changes during the extraction cargo, andmoment eventually leads to 2.1. Motion BodyBody Model The Motion Model includes aircraft model extraction of cargo, and eventually leads to centroid of the carrier aircraft, changes during the The Motion Body Model includes aircraft model 2.1. Motion Body Model [14-15] extraction cargo, ofand leads to considerableofdeviations flighteventually states. Many studies The Motion Body Model aircraft aircraft model and cargo model . Forincludes investigating [14-15] considerable deviations of flight states. Many studies extraction of cargo, and eventually leads to and cargo model . For investigating aircraft [14-15] The Motion Body Model includes aircraft model considerable deviations of flight states. Many studies have been conducted to this field, and dynamic and cargo model . For investigating aircraft characteristics during airdrop, the aircraft’s force [14-15] have been conducted to flight this field, and considerable deviations of states. researches Manydynamic studies [3-4] characteristics during airdrop, the aircraft’s force and cargo model . For investigating aircraft have beenareconducted to this those field, and dynamic functions built to enhance [3-4]. characteristics during airdrop, the aircraft’s force equations and moment equations are built. The cargo’s functions are built to enhance those researches . [3-4] have been conducted to this field, and dynamic equations and moment equations are built. The cargo’s characteristics during airdrop, the aircraft’s force functions are built to enhance those researches . Numerical simulation of airdrop provides a better equations and moment equations are built. The cargo’s [3-4] translation motion will affect aircraft’s flight, so the Numerical simulation of airdrop provides aa better functions built enhance those researches translation motion will affect aircraft’s flight, so the equations and moment equations are built. The cargo’s Numerical are simulation of airdrop provides better understanding of the to whole airdrop process, and leads. translation motion will affect aircraft’s flight, so the cargo’s force equations are also required. understanding of the whole airdrop process, and leads Numerical simulation of airdrop provides a better [5-6] cargo’s force equations are also required. translation motion will affect aircraft’s flight, so the understanding of the whole airdrop process, and leads to a significant reduction of flight tests and costs[5-6]. cargo’s force equations are also required. Above models are given in Figure 1, where vectors to significant of flight tests and and costs [5-6]. understanding ofreduction the whole process, Above models are given in Figure 1, where vectors cargo’s force equations are also required. to aapresent, significant reduction ofairdrop flight tests costsleads . At most of the studies focus onand analysis and ω r , Θ , v Above models are given in Figure 1, where vectors [5-6] and ωa represent aircraft’s location, At most of studies focus on analysis and. raa ,Above Θa , va models to apresent, significant reduction of flight tests and costs and represent aircraft’s are given invelocity. Figure 1,Vectors wherelocation, vectors At present, most of the the studies on analysis ωaa angular ra , Θaa , vaa velocity simulation of flight response orfocus cargo motion. In and the rc and and and represent aircraft’s location, attitude, simulation of flight response orfocus cargoon motion. In and the rc and At present, most of the studies analysis ωa angular ra , Θa , va velocity attitude, and velocity. Vectors and represent aircraft’s location, rc and simulation of flight response or cargo the motion. In the early 1990s, Baixian Fu computed motion of δe vc meanvelocity attitude, and angular velocity. Vectors cargo’s location and velocity. Variety early 1990s, Baixian Fu computed the motion of δe v simulation of flight response or cargo motion. In the rc and mean cargo’s location and velocity. Variety c attitude, velocity and angular velocity. Vectors early 1990s, Baixian Fu computed the motion of δis v continuous airdrop, and studied the responses mean cargo’s location and velocity. Variety means elevator deflection of aircraft, which e c continuous airdrop, and studied the of early 1990s, Baixian Fu computed the motion δis v [7]responses means elevator deflection of aircraft, which cargo’s location and velocity. Variety e c meanelevator continuous airdrop, andof studied the [7]responses of aircraft through a series simulations . The effects F means deflection of aircraft, which is controlled by pilot. Variety means the traction c aircraft through a series of simulations . The effects [7] F continuous airdrop, andcargo responses of controlled by pilot. Variety the traction means elevator deflection ofas aircraft, which is c means aircraft through a series of studied simulations effects F of different factors, i.e. mass,thepull coefficient, controlled by pilot. Variety means the traction [7]. The c force of parachute to the cargo the cargo moves out of different factors, i.e. cargo mass, pull coefficient, aircraft through aand series of simulations effects Fas force of parachute to the cargo cargo out controlled bywith pilot. Variety means traction c the of factors, i.e. cargo mass,condition, pull. The coefficient, CGdifferent coefficient initial flight on the force parachute to cargo of as cargothemoves moves out of the of aircraft thethe traction athe parachute. CG coefficient and initial flight condition, on of different factors, i.e. cargo mass, pull coefficient, [8,9] the of the aircraft with the traction of a parachute. force of parachute to the cargo as the cargo moves out CG coefficient and initial flight condition, on the responses of flight states have also been studied[8,9]. of the aircraft with the traction of a parachute. responses of flight states haveflight also been studiedon [8,9]. the CG and initial condition, of the aircraft with the traction of a parachute. 2.2. Task Model responses of flight states alsostates been studied . A coefficient marked change ofhave flight may lead to [8,9] 2.2. Task A change of flight may lead responses of flight statespilot alsostates been studied 2.2. Task Model Model includes external condition and Task A marked marked change ofhave flight states may lead. to to sudden change of operation. Unexpected Task includes external condition and 2.2. Task Model sudden change of pilot operation. Unexpected A marked change of flight states may lead to Task Model external and reference inputs includes of aircraft flight. condition The external sudden change of pilot operation. Unexpected Aircraft-Pilot Coupling (APC) may happen as a reference inputs of aircraft flight. The external Task Model includes external condition and Aircraft-Pilot Coupling (APC) may happen as a sudden change of pilot operation. Unexpected [10-12] reference inputs of aircraft flight. The external condition gives all environment factors related to Aircraft-Pilot Coupling (APC) may happen as a result[10-12]. Thomas J. studied the interaction between condition gives all environment factors related to reference inputs of aircraft flight. The external result the interaction between [10-12]. Thomas Aircraft-Pilot Coupling (APC) may happen as a [13]J. studied condition gives all environment factors related to airdrop, for instance, airdrop field R , weather and result[10-12] . Thomas the interaction aircraft and pilot[13]J.. studied He presents a cargo between airdrop airdrop, for instance, airdrop field , weather and R condition gives all environment factors related to aircraft and pilot . He presents a cargo airdrop [13] result . Thomas J. studied the interaction between R W airdrop, for instance, airdrop field , weather and atmospheric turbulence . The pilot decides whether aircraft and pilot . He presents a cargo airdrop simulation, which [13] is a real-time environment in order W . Thefield atmospheric turbulence pilot whether airdrop, for instance, airdrop , weather and Rdecides simulation, which is a real-time environment in order aircraft and pilot . He presents a cargo airdrop W atmospheric turbulence . The pilot decides whether to start airdrop or not basing on both external simulation, which is a real-time environment in order to include pilot interaction. The conclusion is that only to start or not both atmospheric The pilot whether to include interaction. The is that only simulation, which is a real-time environment in order raon , Θdecides , va , ω aexternal to start airdrop airdrop or flight notW .basing basing on both external condition andturbulence aircraft states . to include pilot pilot interaction. The conclusion conclusion that only experienced pilot would probably be able tois bring the raon , Θaaboth , va , ω aexternal condition and aircraft flight states . to The startreference airdrop or not basing experienced pilot would probably be able to bring the ω to include pilot interaction. The conclusion is that only r , Θ , v , r , Θ condition and aircraft flight states a a r a r. , v r , ω r inputs of aircraft raflight experienced pilot would able tostate. bring the aircraft quickly into probably a new betrim An , Θa , var,rω, Θ The reference inputs of aircraft flight condition and aircraft flight states . , vr , ω r a a r aircraft quickly into a new trim state. An experienced pilot would probably ablethe tostate. bring the rr , Θairdrop. reference inputsofofthe aircraft flight areThe ideal flight states aircraft during r , vr , ω r aircraft quickly into a new trim An inexperienced pilot, however, can be make situation rr , Θairdrop. , vr , ω r are ideal flight states of the aircraft during The reference inputs of aircraft flight inexperienced pilot, however, can make the situation r aircraft quickly into a new trim state. An are ideal flight states of the aircraft during airdrop. The pilot controls the aircraft flight according to these inexperienced however, can make situation even worse bypilot, trying to compensate thethe deviations, The pilot controls the aircraft flight according to these are ideal flight states of the aircraft during airdrop. even worse by trying to compensate the deviations, inexperienced however, can make situation The pilot controls reference inputs. the aircraft flight according to these even worse lead bypilot, trying to compensate thethe deviations, and might to Pilot Induced Oscillation (PIO) reference inputs. The pilot controls the aircraft flight according to these and might lead to Pilot Induced Oscillation (PIO) even worse by trying to compensate the deviations, reference inputs. and might lead to Pilot Induced Oscillation (PIO) reference inputs. and might lead to Pilot Induced Oscillation (PIO)

2405-8963 © 2016, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved. Copyright@ 177 Control. Peer review 2016 underIFAC responsibility of International Federation of Automatic Copyright@ 2016 IFAC 177 Copyright@ 2016 IFAC 177 10.1016/j.ifacol.2016.12.210 Copyright@ 2016 IFAC 177

2016 IFAC CPHS 178 December 7-9, 2016. Florianopolis, Brazil Tan Wenqian et al. / IFAC-PapersOnLine 49-32 (2016) 177–182

Task Model

Pilot Model

R, W

External Conditions

v a , ωa , ra , Θa

Startup Logic of

Motion Body Model Fc

Cargo

Airdrop rc

C

v r , ωr , rr , Θr

-

v e , ωe , re , Θe

θe

θr

-

v a , ωa rc , v c

Switch Logic of Command Conversion

θe

Pilot Control Model

δe

v a , ωa , ra , Θa

Aircraft

θa

Fig 1 Structure of aircraft-pilot system model for airdrop

The task for pilot on the cargo movement stage is to restrain oscillations of flight states. θ r is the reference pitch during that stage. The pilot controls aircraft also depending on θe . The pilot control model describes the pilot’s control behavior basing on task instructions. Compensatory control, precognitive control or any other control methods can be used here. And that is the very part to be investigated in this paper.

2.3. Pilot Model Pilot Model describes pilot’s control behavior during the process of airdrop. As shown in Figure 1, Pilot Model includes startup logic of airdrop, switch logic of command and pilot control model. The startup logic of airdrop and the switch logic of command describe pilot’s discrete decision-making behavior. During the airdrop pilot should decide the startup time T1 for airdrop. Both the external conditions R , W and aircraft flight states ra , Θa , va , ω a are analyzed whether they are suitable for airdrop. The airdrop startup command is dictated when time is right. With the start of airdrop, the fixing device of cargo is released. The cargo moves out of the aircraft with the traction force of parachute Fc . The switch logic of command decides the task command given to the pilot during different airdrop stages. The process of airdrop can be divided into three stages according to time sequence. The command of airdrop C and the position of cargo rc are used to distinguish each stage. The relationship between the time sequence of airdrop and the pilot’s control task command is shown in Figure 2. Task Command

3.1 Compensatory control Compensatory control of pitch is introduced firstly that the pilot will do precise tracking of pitch during airdrop. Classical pilot model includes McRuer Model, Hess’s Structural Pilot Model and Pilot Optimal Control Model (OCM). All of these models are quasi-linear models. During the process of airdrop, the characteristics of APC change suddenly when the cargo moves out. Above quasi-linear pilot model cannot be used to describe pilot’s nonlinear control behavior. While the aircraft characteristics changes with the motion of cargo, a time-varying structural model is essential. Such kind of pilot model and aircraft-pilot system have been studied by Ronald A. Hess [16,17]. At the airlift stage, a Hess’s structural pilot model[18,19] Yp1 is used to describe pilot’s continuous compensatory control behavior. During the cargo movement stage, the characteristics of aircraft change continually with cargo’s motion. Generally, the cargo will move out of aircraft in 5 seconds which is a short time. Although pilot control behavior exhibits nonlinear characteristics, the structural model for airlift stage Yp1 is also used here. At the empty flight stage, another structural pilot model Yp2 is applied. The structure of pilot control model is shown in Figure 3, of which Yp1 is the structural model for airlift stage and also be used during cargo movement stage, while Yp2 is the structural model for empty aircraft stage. An adaptive time τ is needed for pilot to adapt well to new features of empty aircraft. So the time that S switches from 1 to 2 is set as T2 + τ , where τ =2s [20].

Empty Aircraft Flight

Flight State Maintenance

Cargo Movement

Pitch Control Flight State Maintenance

3. Investigation of pilot control methods for airdrop The startup logic of airdrop and the switch logic of command have been described in detail in Section 2.3. The pilot control model will be studied here.

Airlift

T1 T2 Time Fig 2 Relationship between the time sequence of airdrop and the pilot control task command

During the airlift stage, the aircraft remains initial flight states while the cargo is fixed in its initial position. Pilot’s control task in this stage is to maintain straight level flight. The cargo movement stage begins when the pilot dictates the start of airdrop at time T1 . Traction parachute is released so as to pull the cargo toward the rear. Pilot’s task in this stage is to avoid severe oscillation of flight state. The empty aircraft stage comes when the cargo moves out of the aircraft at time T2 . Then the pilot brings the aircraft back into a new straight level flight. Pilot’s task in this stage is to maintain the new flight state. At airlift stage and empty aircraft stage, pilot should control the aircraft to maintain straight level flight. The errors re , Θe , ve , ω e between the reference inputs rr , Θr , v r , ω r and the real flight state ra , Θa , va , ω a are calculated and given to the pilot. Based on the errors re , Θe , ve , ω e , the pilot creates attitude control command θe and controls flight according to it.

θe

S

1 2

Yp1

δe

Yp2 Fig 3 Pitch compensatory control model

178

2016 IFAC CPHS December 7-9, 2016. Florianopolis, Brazil Tan Wenqian et al. / IFAC-PapersOnLine 49-32 (2016) 177–182

known that one of the causes of PIO is pilot’s precise tracking behavior. If the gain of pilot control is higher, the aircraft-pilot system is more unstable. Additionally, the pilot controls flight states through the operation of elevator. Theoretically, a larger deviation of pitch needs a higher deflection of elevator to maintain flight states. Maintaining elevator deflection is one method to decrease pilot control gain. Hence, elevator maintenance is taken into account when designing pilot control methods. As shown in figure 5, an additional maintenance module is placed after the pilot structural models. Variety δ e0 is defined as the value of maintenance.

At the airlift stage and the empty aircraft flight stage, the pilot controls the aircraft’s pitch attitude to maintain straight level flight. Table 1 shows the parameters of the initial and final equalization flight states for airdrop task. These parameters are regarded as the reference inputs of aircraft flight. Table 1 Initial and final flight parameters Final flight Initial flight Item Item parameters parameters 100 100 V2 (m ⋅ s −1 ) V1 (m ⋅ s −1 ) H2 / m H1 / m 800 800 α1 /(°) α 2 /(°) 3.352 2.040 θ1 /(°) θ 2 /(°) 3.352 2.040 δ e.2 /(°) δ e.1 /(°) 1.931 5.719

θe

Yp1

Compensatory control

δ e′

−δ e0

Yp2

20

δ

e

8-deg maintenance for elevator 10-deg maintenance for elevator 12-deg maintenance for elevator Without elevator maintenance

10 q/ °/s

0 -5 -10

0 -10

10

20

Time/s (s)

30

40

50 -20 0

（a）Response of the pitch-rate 20

10

25

15

θ/ °

5 0

5 0

(s) Time/s

30

40

30

40

50

10

-5 0

20

(s) Time/s

8-deg maintenance for elevator 10-deg maintenance for elevator 12-deg maintenance for elevator Without elevator maintenance

20 10

10

20

（a）Response of the pitch-rate

Compensatory control

15

θ/ °

δ e0

The influence of elevator maintenance on responses of aircraft is investigated. The control stick provides the interface between pilot and aircraft which means maintenance of elevator can be translated to maintenance of stick displacement in practice. In this airdrop task, maintenances of elevator δ e0 are set as 8°, 10° and 12° separately. Responses of aircraft are shown in Figure 6, comparing with no elevator maintenance. Elevator deflection curve is shown in Figure 7.

5

-15 0

δ e′

Fig 5 Elevator maintenance control model

10

q/ °/s

1

S

2

At the cargo movement stage, smoothly flight is hoped. The pilot obtains the attitude control command θe directly and controls the pitch attitude in order to eliminate the error of pitch θe . The mass ratio of cargo to aircraft λ affects the APC characters of airdrop. In this paper, the mass ratio λ is set as 0.169. The pitch compensatory control model is added to the aircraft-pilot system model and numerical simulation is conducted to test the effectiveness of the control method. Simulation results are as follows, where T1 =10 sec. 15

179

50

-5 0

（b）Response of the pitch

Fig 4 Responses of flight states with compensatory control

10

20

(s) Time/s

30

40

50

（b）Response of the pitch

Fig 6

Figure 4 shows the responses of aircraft during airdrop. With the motion of cargo, the aircraft will nose up and the pitch will increases at cargo movement stage. In order to eliminate the pitch deviation quickly, the pilot pushes the control stick to obtain a negative elevator deflection. This control behavior leads to the aircraft’s nosing down, and as a result, the pilot pulls the control stick to obtain a positive elevator. During the process of airdrop, the pilot controls the stick repeatedly. Therefore, the aircraft flight states oscillate violently. It is very easy to bring pilot induced oscillation (PIO), just like illustrated by Tomas in reference [13]. Therefore, a more effective pilot control method is essential for the sake of safety.

The effect of elevator maintenance on airdrop 20

8-deg maintenance for elevator 10-deg maintenance for elevator 12-deg maintenance for elevator Without elevator maintenance

δe/ °

15 10 5 0 0

Fig 7

10

20

(s) Time/s

30

40

50

Elevator deflection of elevator maintenance

The main parameters of the responses of aircraft are given in Table 2. Ve max in Table 2 is the peak value of velocity deviation during airdrop. qmax is the maximum of pitch rate which is the peak value of nose-up rate. qmin is the minimum of pitch rate which is the peak value of nose-down rate. θ e max is the peak value of pitch deviation. H e max is the maximum of altitude deviation.

3.2 Elevator maintenance control The results above show that pilot’s precise tracking of pitch during airdrop will lead to serious oscillations of flight states, and may also lead to PIO. It is well 179

2016 IFAC CPHS 180 December 7-9, 2016. Florianopolis, Brazil Tan Wenqian et al. / IFAC-PapersOnLine 49-32 (2016) 177–182

Table 2 Comparison of elevator deflection maintenance 8° 10° 12° No maintenance Item Ve max /(m ⋅ s −1 ) 10.0 6.9 5.7 4.7 qmax / (° ⋅ s −1 ) 14.4 13.1 12.4 12.4 qmin / (° ⋅ s −1 ) -5.7 -7.4 -9.3 -14.0 θ e max / (°) 16.6 15.0 14.3 13.7 H e max / m 96 62 49 37

Flight curves in Figure 6 and data in Table 2 show us that maintaining elevator has some kind of influence on aircraft. For instance, Ve max rises to 10m/s from 4.7m/s, and H e max rises to 96m from 36m when the maintenance of elevator is set as 8°. The loss of velocity and altitude is acceptable in the simulated mission. Figure 6 reveal that maintaining elevator has good effects on pitch since the qmin decreases from 14°/s to 5.7°/s and the oscillation of pitch has been dampened. One can conclude from Figure 8 that maintaining elevator does not have noticeable influence on the motion of cargo. 0

Vc/ m/s

Fig 9 Pilot model for precognitive control method 1

Through a series of simulations, the relationship between ∆δ e and the ratio of mass of cargo to aircraft λ are as shown in Figure 10. 5

Δδe/ °

4

-4

3 2 1 0 0

0.05

0.1 λ

-8

10

10

20

(s) Time/s

30

40

50 q/ °/s

Though elevator maintenance control method has better effects than compensatory control, the simulation results are not that satisfying cause the deviations of flight states remains too large. Other control methods are needed to be designed.

0 -5 -10 0

10

20

30 Time/s (s)

40

50

（a）Response of the pitch-rate

3.3 Precognitive control For an airdrop task, the start time of airdrop, the fixed position of cargo and cargo motion are known by pilot in advance. Hence an experienced pilot has a practical understanding of flight tendency during airdrop, i.e. the pitch will increase with the start of airdrop if there is no control. Precognitive control is introduced to decrease that change.

10

Precognitive control 1

θ/ °

5

0

-5 0

3.3.1 Precognitive control method 1 Pilot precognitive control method 1 for airdrop is designed as follows. First, do compensatory control on the airlift stage. Then increase elevator to δ e0 with the start of airdrop. The elevator kept constant when the cargo moves inside the cabin. The elevator turns back to pitch compensatory control to bring the aircraft into a new trim state when the cargo moves out. Formula (1) describes the control method. C = 0 and K = 0 C = 1 and K = 0 C = 1 and K = 1

Precognitive control 1

5

Fig 8 Effect of elevator maintenance on cargo motion

Yp1 × θ e  δ e = δ e0 Y × θ  p2 e

0.2

In order to compare with the results of compensatory control method and elevator maintenance control method, airdrop simulation is conducted with λ = 0.169 , and the responses of aircraft are shown in Figure 11. The elevator curve is shown in figure 12.

-6

-10 0

0.15

Fig 10 Relationship between ∆δ e and λ

8-deg maintenance for elevator 10-deg maintenance for elevator 12-deg maintenance for elevator Without elevator maintenance

-2

of pitch ∆θ max is set as optimization function.

10

20

30 Time/s (s)

40

50

（b）Response of the pitch

Fig 11 Responses of flight states with precognitive control method 1 7 6

δe/ °

5 4 3 2 1 0

(1)

10

20

(s) t/s

30

40

50

Fig 12 Elevator deflection of precognitive control method 1

Comparing with the first two control methods, all flight states have been controlled effectively under precognitive control method 1. However, the initial change of pitch is kind of large cause the value of δ e0 is a little big. Hence, precognitive control method 2 is designed as an adjustment.

Yp1 and Yp2 are the same with what we’ve got

above. C represents the airdrop signal. C=0 at airlift stage and C=1 with the start of airdrop. K represents cargo’s position while when K=0 means the cargo is inside the aircraft and when the cargo moves out of the cabin, K=1. A pilot model is built to realize the control method, shown in Figure 9. According to the method, δ e0 should be given in advance. The mass of cargo is taken into account when design the value of δ e0 . The maximum of error

3.3.2 Pilot precognitive control method 2 Pilot precognitive control method 2 for airdrop is designed as follows. First, do compensatory control

180

2016 IFAC CPHS December 7-9, 2016. Florianopolis, Brazil Tan Wenqian et al. / IFAC-PapersOnLine 49-32 (2016) 177–182

on the airlift stage. Then increase elevator to δ e1 , a smaller value compared with δ e0 designed above, with the start of airdrop. The aircraft will nose down respectively and nose up with the movement of cargo. Increase elevator to δ e2 at the point of nose up and keep that until the cargo moves out. Turn back to pitch compensatory control to bring the aircraft into a new trim state when the cargo moves out. Formula (2) describes the control method. Yp1 × θ e  δ δ e =  e1 δ e 2  Yp 2 × θ e

C = 0 and K = 0 and L = 0 C = 1 and K = 0 and L = 0 C = 1 and K = 0 and L = 1 C = 1 and K = 1 and L = 1

10

δe/ °

8

2 0

10

Ve max / (m ⋅ s −1 ) qmax / (° ⋅ s −1 ) qmin / (° ⋅ s −1 ) θ e max / (°) H e max / m

6 Δδe2/ °

4 2

0.1

0 0

0.15

0.05

0.1 λ

λ

0.15

0.2

and λ （b） ∆δ e2 and λ Fig 14 Relationship between ∆δ e1 , ∆δ e2 and λ （a） ∆δ e1

q/ °/s

Compen

Elevator

Precog-

Precog-

maintenance

control 1

control 2

9.0

7.7

2.6

2.1

12.8

12.3

9.0

8.3

-12.8

-5.2

-4.6

-4.9

12.6

12.6

7.4

5.1

84.8

70.1

27.6

16.1

Compensatory control Elevator maintenance Precognitive control 1 Precognitive control 2

110

V/ m/s

105 100

Precognitive control 2

95

5

90 0

0

10

20

(s) Time/s

30

40

50

（a）Response of the velocity 15

-5

Compensatory control Elevator maintenance Precognitive control 1 Precognitive control 2

10 10

20

(s) Time/s

30

40

5

50

q/ °/s

-10 0

50

- control

115

Airdrop simulation is conducted where λ = 0.169 , and the responses of aircraft are shown in Figure 15. The elevator curve is shown in figure 16. 10

40

Table 3 Comparison of flight data Item

0.05

30

4. Real-time experiment with pilot-in-loop In order to test those methods in practical use, pilot-in-loop real-time experiments are conducted. These experiments were fulfilled by use the BUAA (Beijing University of Aeronautics and Astronautics) fixed-based flight simulator for investigation of manual control task. The closed-loop pilot-aircraft system is shown in Figure 1 and the pilot model is substituted by a real pilot. Experiment results are shown in Figure 17, which are in agreement with those methods designed above. Table 3 gives the comparison about the aircraft flight states between those methods.

3

0 0

(s) Time/s

The pitch problem of precognitive method 1 has been improved while the other flight states are basically remain the same. However, the disadvantage of precognitive method 2 compared with method 1 is its complexity which may lead to execution problem in practice. Comparing with the first two control methods, the aircraft performs most satisfactorily under the two precognitive control methods since all flight states have been controlled effectively.

(2)

Two fit curves which describe the relationship between ∆δ e1 and λ , else the relationship between ∆δ e2 and λ are created respectively. The values of ∆δ e1 and ∆δ e2 for any λ can be picked in the Figures 14.

1

20

Fig 16 Elevator deflection of precognitive control method 2

Fig 13 Pilot model for precognitive control method 2

Δδe1/ °

6 4

Here L represents the signal of pitch. The initial value of L is 0 and L turns to be 1 at the point when the aircraft start to nose up during the movement of cargo in the cabin. An pilot model is built to realize the control method, shown in figure 13. According to the method, δ e1 and δ e2 should be given in advance.

2

181

（a）Response of the pitch-rate

0 -5

10

Precognitive control 2

-10 -15 0

θ/ °

5

10

20

(s) Time/s

30

40

（b）Response of the pitch-rate 0

-5 0

10

20

(s) Time/s

30

40

50

（b）Response of the pitch

Fig 15 Responses of flight states with precognitive control method 2

181

50

2016 IFAC CPHS 182 December 7-9, 2016. Florianopolis, Brazil Tan Wenqian et al. / IFAC-PapersOnLine 49-32 (2016) 177–182

20

Compensatory control Elevator maintenance Precognitive control 1 Precognitive control 2

θ/ °

15 10

[4] [5]

5 0 -5 0

10

20

(s) Time/s

30

40

50

（c）Response of the pitch 1000 950

H/ m

[6]

Compensatory control Elevator maintenance Precognitive control 1 Precognitive control 2

900 850

[7]

800 750 0

10

20

30 Time/s (s)

40

[8]

50

（d）Response of the altitude

Fig 17 Responses of flight states

[9]

Experiment results match well with numerical simulations, which illustrate that precise pitch compensatory control will lead to violent oscillations of flight states. The peak value of nose-down rate decreases when elevator maintenance is introduced, which dampens the oscillation effectively. Precognitive control executed by the pilot is based on the knowledge of flight tendency during airdrop. Both two precognitive methods are effective and each has its own advantages and disadvantages. Precognitive control can be easily mastered according to the real-time experiments and is comparatively more effective than the first two methods as both attitude oscillation and flight trajectory are well controlled.

[10] [11]

[12] [13]

5. Conclusions (1) An aircraft-pilot system model is built with detailedly defined motion body model, task model and pilot model inside which can be used for investigating aircraft-pilot coupling during heavy cargo airdrop. (2) Cargo’s motion has remarkable influence on aircraft since the reaction of aircraft is violent. Precise pitch compensatory control is not suitable for airdrop and may lead to PIO problems. (3) Three different control methods are investigated to control flight states during airdrop. There are obvious differences between those methods on flight. Precognitive control dampens the oscillations of flight states most effectively and simplifies complex command. (4) Values of elevator operation for precognitive control are not the same when cargo mass changes. The relationship is given in the paper. The study presents an effective control method which minimizes flight deviations and dampens oscillations during heavy cargo airdrop. The result is of considerable significance to be applied in practical use.

[14] [15] [16]

[17] [18] [19]

References [1] A. A. Arnold. The development of ultra low altitude [2] [3]

[20]

dropping of heavy stores. Aerodynamic deceleration systems conference, 1966; 237-241. K. J. Desabrais, Justin Riley, James Sadeck, et al. Low-cost high-altitude low-opening cargo airdrop systems. Journal of Aircraft, 2012; 49(1): 349-354. Zhang H N, Zhang P T, Cheng W H. Research of two different modeling methods of airdrop. Science Technology and Engineering, 2012; 12(16):

182

3896-3900[Chinese]. Han Y H, Lu Y Ｐ. Dynamics for heavy cargo airdrop and control design. Flight Dynamics, 2011; 29(3): 28-31[Chinese]. P. A. Cuthbert, Gary L. Conley. A desktop application to simulate cargo drop tests. 18th AIAA Aerodynamic decelerator systems technology conference and seminar. Monterey, California: American Institude of Areonautics and Astronautics ; 2005.1623-1636. P. A. Cuthbert. Validation of a cargo airdrop software simulator. 17th AIAA Aerodynamic decelerator systems technology conference and seminar. Monterey, California: American Institude of Areonautics and Astronautics ; 2003. 2133-2143. Fu B X. The longitudinal dynamic analysis of Y-8 airlpane in continuous airdrop. Flight Dynamics, 1993; 11(1): 80-87[Chinese]. Chen J, Shi Z K. Aircraft modeling and simulation with cargo moving inside. Chinese Journal of Aeronautics, 2008; 22(2009): 191-197. Ke P, Yang C X, Yang X Ｓ. Extraction phase simulation of cargo airdrop system. Chinese Journal of Aeronautics, 2006; 19(4): 315-321. Tian F L, Gao Z Ｈ. Current status and prospects of aircraft pilot coupling research. Flight Dynamics, 2005; 23(1): 9-13[Chinese]. Ronald A. Hess. A Unified Theory for Aircraft Handling Qualities and Adverse Aircraft-Pilot Coupling. American Institude of Areonautics and Astronautics, 1997; 454-465. Mark R. Anderson. Pilot-induced oscillations involving multiple nonlinearities. American Institude of Areonautics and Astronautics, 1997; 192-200. Thomas Jann. Coupled simulation of cargo airdrop from a generic military transport aircraft. 21st AIAA Aerodynamic decelerator systems technology conference and seminar. Dublin, Ireland: American Institude of Areonautics and Astronautics, Dublin, Ireland: American Institude of Areonautics and Astronautics ; 2011. 2566-2585. Yang M S, Qu X J. Flight dynamic modeling and simulation for transport airdrop. Flight Dynamics, 2010; 21(2): 9-12[Chinese]. Meng B, Wu H X. On characteristic modeling of a class of flight vehicles’ attitude dynamics. Sci China Tech Sci, 2010; 53(8): 2074-2080. Ronald A. Hess. A Preliminary Study of Human Pilot Dynamics in the Control of Time-Varying Systems. AIAA Modeling and Simulation Technologies Conference. Portland, Oregon: American Institude of Areonautics and Astronautics ; 2011. 6554-6595. Ronald A. Hess. Modeling Pilot Control Behavior with Sudden Changes in Vehicle Dynamics. Journal of Aircraft, 2009; 46(5), 1584-1592 Qu X J, Fang Z P. Identification the parameters of pilot structural model. Acta Aeronautica et Astronautica Sinica, 1996; 17(7): 63-66 [Chinese]. Ronald A. Hess. Structural model of the adaptive human pilot. J of Guidance and Control, 1980; 3(5): 416-423. Tan W Q, Zhang W, Qu X J. Analysis of APC characteristics of control mode switching induced by active side stick. 21st AIAA Atmospheric flight mechanics conference. Minneapolis, Minnesota: American Institude of Areonautics and Astronautics; 2012. 4502-4512.