7/30/2019 AIAA-46204-831

1/6

Engineering NotesSynthetic-Waypoint Guidance

Algorithm for Following a Desired

Flight Trajectory

Eran D. B. Medagoda and Peter W. Gibbens

University of Sydney,

Sydney, New South Wales 2126, Australia

DOI: 10.2514/1.46204

Nomenclature

N = proportional gainR = closing range

R

= virtual rangeTH = time horizonVa = true airspeedVw = synthetic-waypoint speedxc = output commandC = commanded climb angleref = reference climb angleT = sample time = heading errorC = commanded heading angleref = reference heading angle

I. Introduction

THE development of aircraft guidance, navigation, and controlsystems has been a long-standing research area. Numerous

methods relating to the enhancement of aircraft performance undervarious mission parameters have been developed in response to aneed for more reliable and robust guidance systems. Current guid-ance systems applied to commercial, civilian, and unmanned aircraftrely onthe knowledge of aflight path, specified by waypointslocatedin inertial space. Most missions are considered successful when thevehicle reaches thedesignated waypoint at which new commands areissued to the vehicle to proceed to the next waypoint. Two commontypes of conventional aircraft guidance are the direct-to-waypoint(DTW) and track-to-waypoint (TTW) methods in relation to path-following between designated waypoints. The DTW method simplyissues heading commands to the vehicle based on the angular differ-ence betweenthe waypoint andvehicle.When thevehiclereaches the

waypoint, the control system issues a new command to guide theaircraft to the next waypoint. The TTW method aims to follow thetrack between waypoints. In this guidance methodthe control systemaims to minimize the lateral offset between the prescribed flight pathand the aircrafts position, issuing heading commands that return thevehicle to the nominalflight path. The track method therefore placesthe additional constraint on aflight path that the vehicle must follow

in order to reach the waypoint, rather than simply reaching thewaypoint. However, both methods are far from optimal. This is evi-

dent in how theaircraft transitions between flight paths after reachinga waypoint. During theseflight-path transitions, the aircraft will oftenovershoot the desiredflight path to correctits track, particularlywhenthe flight-path transition angle is acute. Various control strategieshave been investigated to alleviate or minimize flight-path devia-tions. Such strategies include applying modern control methods suchas receding-horizon control [1,2] and model predictive control [3] toanticipateflight-pathchanges andtake control action before reachinga goal while maintaining adequate vehicle flight performance.

Missile guidance and control systems operateon similar principlesto commercial, civilian, and unmanned aircraft guidance and controlalgorithms. The primary mission for missile systems is to intercept amoving target using information about the relative position andvelocitybetween thepursuer andtarget. Oneof thefirst methods usedin missile guidance was pursuit guidance (PG) [47]. The method

operates by forcingthe angular displacement error between a pursuerand its target to zero. Control commands scaled by a proportionalfactor of the current error are then issued to direct the pursuer alongthe line of sight (LOS) between the pursuer and target. PG solutions,however, do not consider the path taken or the levels of systemperformance required by the pursuer in reaching the target, resultingin a far-from-optimal solution. To address this problem of sub-optimality, additional parameters have been introduced to enhancemissile performance. One method includes taking into account themotion of the commanded line of sight between the pursuer andtarget [4,710], issuing lateral acceleration commands based ontracking error and tracking error rate to the target. This approach hasbeen shown to improve overall interceptor performance comparedwith conventional PG [4,7]. Another such methodaims to modifythe

level of control the guidance algorithm possesses over the vehicle byadjusting the level of proportional gain. This is achieved by gainscheduling [11] to select gain values based on current interceptorstates.

In addition to modifying internal missile guidance and controlparameters such as variable gains and LOS rate estimation, missionperformance can be enhanced by manipulating the trajectory takenby the pursuer to the targets. A good example of such a method isdiscussed in [12], in which a missile aims to exploit the aerodynamicbenefits of high-altitude flight by tracking a virtual target at someinitially high altitude that is not necessarily along thetrajectory to thetrue target.

This Note discusses the development of a guidance law fusingthe virtual-target concepts with those of pursuit guidance forimplementation into an aircraftguidance system. This Note develops

a path-following aircraft guidance algorithm that pursues syntheticwaypoints using only a small set of guidance parameters, extendingthe virtual-target concept to complete aircraft guidance. The path isdefined by the track between a minimal set of waypoints at specifiedlocations, removing the need for a smooth path to be defined or theneed for complicated path-switching logic or trajectory planningwhen a waypoint is reached. Thesynthetic waypoint travelsalongthepath between waypoints, with the trailing aircraft traveling a smoothpath generated through its own dynamics in following the syntheticwaypoint. The guidance law is tested by varying guidance param-eters, thus assessing vehicle sensitivity to and overall systemperformance of parameter variations. The following sections discussthe basic concepts of missile and aircraft guidance and provide adetailed description of the structure of the synthetic-waypoint guid-ancealgorithm. A discussion on the implementation of the algorithminto the underlying aircraft control system will also be presented,followedby an analysis of theperformance of theguidance algorithmin nonlinear simulation.

Received 30 June 2009; revision received 3 November 2009; accepted forpublication 9 November2009. Copyright 2009by the American Institute ofAeronautics and Astronautics, Inc. All rights reserved. Copies of this papermaybe made forpersonal or internal use, on condition that the copier pay the$10.00 per-copy fee to the Copyright Clearance Center, Inc., 222 RosewoodDrive, Danvers, MA 01923; include the code 0731-5090/10 and $10.00 incorrespondence with the CCC.

Graduate Research Student, School of Aerospace, Mechanical andMechatronic Engineering; e.medagoda@aeromech.usyd.edu.au.Senior Lecturer, School of Aerospace, Mechanical and Mechatronic

Engineering; pwg@aeromech.usyd.edu.au.

JOURNAL OF GUIDANCE, CONTROL, AND DYNAMICSVol. 33, No. 2, MarchApril 2010

601

http://-/?-http://-/?-http://dx.doi.org/10.2514/1.46204http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://dx.doi.org/10.2514/1.46204http://-/?-http://-/?-

7/30/2019 AIAA-46204-831

2/6

II. Algorithm Overview

A. Pursuit Guidance

The theory of pursuit guidance is based on maintaining a directheading toward a targetby driving theanglebetween theforward axisof the pursuer and the LOS to the target to zero [7]. This is achievedby commanding the pursuing vehicle to maneuver at a rate equal totheLOS rate. As thetargetmoves, thepursuer also moves to maintaina direct line to the target for interception. In its simplest form, thepursuit guidance algorithm can be expressed mathematically as

xC N_ (1)

where xC is the output command, and N is a proportional gain

constant that multiplies the LOS rate _. Figure 1 shows a typicalpursuit engagement scenariobetween thepursuer andtarget. It canbeseen that the pursuer aims to maintain a direct heading toward thetarget by aligning its direction of travel with the LOS to the target, bydriving the LOS sight error angle to zero. As the target moves, sodoes the LOS, forcing the pursuer to adjust accordingly. As theclosing range between the pursuing vehicle and the target reduces,the pursuer converges on the targets position.

The greatest benefit of applying pursuit guidance is in its sim-

plicity. Only and _ need to be measured in order to generate

sufficient control commands for an interception course. The gainterm N is the only guidance parameter that needs to be defined,determining the input states level of magnification. The gain termcan be varied to allow for adequate flight performance, whileenabling the vehicle to converge upon the commanded trajectory.However, the selection ofN is heavily dependent on the aircraftcurrent state, which results in varying levels of performance acrosstheflight envelope ifNremains unchanged. As a result, a schedule ofgains needs to be specified to allow for consistent aircraft perform-ance within the flight envelope.

B. Synthetic-Waypoint Guidance

1. Engagement Dynamics

In the SWG algorithm, aflight path is defined that designates the

trajectory that the aircraft is to follow. The waypoints defining theflight path are specified in inertial space with reference to afixed localframe. Following this, a synthetic waypoint is placed at the locationofthefirstwaypoint. Theoperationof theSWG algorithmis based ontracking the synthetic waypoint that travels along the designated

flight path. The waypoint is considered synthetic, as its position onthe flight path is a projection of the position at which the aircraftintends to be within a specified time horizon. The time horizon is auser-defined time interval by which the aircraft trails the syntheticwaypoint. As the aircraft approaches the synthetic waypoint on theflight path, the synthetic waypoint repositions itself, forcing thevehicle to assume a new heading in pursuit.

Figure 2 illustrates the overall dynamics of a lateral engagementbetween the aircraft and the synthetic waypoint for a no-wind

situation. Initially (Fig. 2a), the aircraftis a distanceR fromthefl

ightpath definedby theLOS between thevehicle andsynthetic waypoint.The termC defines the commanded heading angle required by theaircraftto headtoward the synthetic waypoints current position. Thesynthetic waypoint is initially located at point 1. As the closing rangeR reduces during the engagement midcourse (Fig. 2b), the syntheticwaypoint beginsto move along theflightpathbetweenpoints 1 and 2along a reference path with heading ref defined by the flight-pathgeometry. As the synthetic waypoint moves, its speed graduallyincreases, forcing the aircraft to change its heading with appropriatecontrol to maintain a direct pursuit. As the closing range reducesfurther until a desired range is achieved, the speed of the syntheticwaypoint along the flight path increases to match the speed of theaircraft. As a result, the aircraft continually pursues the waypointalong theflight path (Fig. 2c). The desired range that thevehicle aims

to achieve behind the waypoint is a function of the aircraft currentspeed and time horizon. This will be discussed in further sections.

2. Guidance Laws

For the SWG algorithm to operate, certain constraints need to beapplied to the dynamic behavior of the synthetic waypoint. Suchconstraints are applied to ensure correct tracking between the syn-thetic waypoint and the aircraft. The laws that define the dynamics ofthe waypoint are as follows.

The synthetic waypoint can only travel along the flight pathbetween designated inertial waypoints. This constraint is necessaryto ensurethat thesynthetic waypoints movement is along thedefinedflight path. If this condition was not obeyed, then the guidancealgorithmwould be tracking a point that would leadthe aircraft away

from the flight path. Mathematically, this condition is expressed as

Xw

Yw

Zw

264

375 i

XwYwZw

24

35i1

Vw

cos ref cosrefcos ref sinref sin ref

24

35T (2)

where Xw, Yw, and Zw define the three-dimensional position of themoving synthetic waypoint in inertial space;Vw defines the speed ofthe synthetic waypoint;ref and ref define the reference heading andclimb angle of theflight path between fixed waypoints, respectively;and Tdefines the integration sample used to update the position ofthe synthetic waypoint.

The minimum distance between the synthetic waypoint andthe aircraft, referred to as the virtual range, must be set within the

guidance algorithm. This constraint implies that aircraft can never becloser to the synthetic waypoint than the specified minimum rangethat will be defined as the virtual range R. Mathematically, thiscondition implies that

R R (3)

where R defines the closing range between the aircraft and currentposition of the synthetic waypoint.

The desired virtual range for a given flight condition is defined bythe vehicles current airspeed and time horizon. This constrainteffectively defines the level of performance achievable by the air-craft. Mathematically, the virtual range can be expressed as

R VaTH (4)

where Va represents the aircraft currentairspeed andTHspecifies thedesired time horizon for the vehicle to initiate a response to flight-path changes. The time horizon is analogous to prediction horizons

target

pursuer

X

Y

Fig. 1 Two-dimensional pursuer/target engagement.

602 J. GUIDANCE, VOL. 33, NO. 2: ENGINEERING NOTES

http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-

7/30/2019 AIAA-46204-831

3/6

discussed in most predictive control applications , which specifies aprediction horizon over which control action is taken to achieve afuture goal. The time horizon in this application refers to the level ofset-point prediction introduced into the system.

The speed of the waypoint is dictated by the ratio of the virtualrange and current closing range multiplied by the aircraft currentairspeed. This constraint is crucial to the overall guidance algorithmand ensures that the aircraft is always in pursuit of the syntheticwaypoint. Mathematically, this condition is expressed as

Vw VaR

R(5)

where Vw represents the speed of the synthetic waypoint along thereference flight path. This constraint deviates from the typicalinterceptor/target engagement scenarios in which the dynamics ofthe interceptor and target are independent. In this case, the dynamicsof the synthetic waypoint and aircraft are intrinsically linked throughVa and R.

III. Aircraft Control

Aircraft flight control is driven by track and climb-angle com-mands calculated from the positional difference between the aircraft

and synthetic waypoint through observation of LOS dynamics. Thecommands are then transformed into control inputsto aircraft controlsystem. Theformulation of thecontrol systemwas performedusingalinear quadratic regulator [13] design. It is important to note that thedesigned controller was kept constant throughout the analysis, as it isthe effect of modifying SWG parameters on vehicle stability andperformance that is being assessed, not the overall effectiveness ofthe controller. Ideally, a new controller would be designed toaccommodate for SWG parameter changes.

Figures 3aand 3b show the relative position of the aircraft and thesynthetic waypoint along the lateral and longitudinal axes. Thecoordinates (Xa, Ya, and Za) and (Xw, Yw, and Zw) define the currentpositions of the aircraft and synthetic waypoint, respectively, in theinertial frame.

Consequently, the commanded heading and climb angles to the

current synthetic-waypoint position can be evaluated as

c arctan

Ry

Rx

(6)

c arcsin

RzR

(7)

where

Rx Xw Xa (8)

Ry Yw Ya (9)

Rz Zw Za (10)

C is the commanded heading andC is the commanded climb angleto the synthetic waypoint from the aircraft current position. Note thatEq. (6) operates a four-quadrant inverse.

Figure 4 shows the complete closed-loop system summarized inblock-diagram form. The guidance block within the loop executesthe SWG algorithm, receiving vehicle states from the aircraftdynamics to generate guidance commands that are also fed into thecontrol and navigation system. The navigation system containsthe defined flight-path information used to update the position of thesynthetic waypoint, which is then fed back into the guidance blockalong with aircraft states. The control block receives guidance com-mands to generate control actions based on the current aircraft statesrelative to the synthetic waypoint. The controls generated are thenused to guide the aircraft along the desired trajectory.

X

Y

C

ref

R

Vw = 0

1

2

X

a)

b)

c)

Y

ref

R

Vw Va

1

2

C

X

Y

ref

RVw = Va

1

2

C

Fig. 2 Synthetic-waypoint engagement dynamics: a) initial engage-ment, b) midcourse, and c) flight-path acquisition.

J. GUIDANCE, VOL. 33, NO. 2: ENGINEERING NOTES 603

http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-

7/30/2019 AIAA-46204-831

4/6

IV. Simulation Results

This section demonstrates the SWG algorithm within a nonlinearaircraft simulation. The flight trajectory taken by the vehicle wasdefined in three dimensions and was known by the vehicle a priori.

The simulated aircraft is classed as afighter trainer and was trimmedat an airspeed of75 m=s (150 kt) at an altitude of 500 m.

Figure 5 shows the layout of the defined flight path in the inertialframe and the trajectory taken by the aircraft under different timehorizons. The defined flight path consists of transition pointsrequiring varying degrees of control activity to follow the desiredtrajectory. All the legs of the defined flight path were maintained at aconstant altitude of 500 m. From the aircraft response, the trajectorytaken by the vehicle follows the defined flight path closely in all

cases, with minimalfl

ight-path divergence using SWG.Figures 6 and 7 shows an enhanced view around an acute-angletransition point. For the purposes of the simulation, the bank angle ofthe vehicle was limited to 45. Even though the flight path itselfundergoes rapid changes in heading, the trajectory taken by thevehicle is smooth and rapidly reacquires the new flight leg once themaneuver has been completed. This is due to the fact that the syn-thetic waypoint transitions along a new flight path before reachingthe true waypoint located at the apex. Since the aircraft is respondingto changes in the synthetic-waypoint position, it begins to transitionto the new flight leg before the apex is reached, avoiding rapidtransition between legs, requiring high levels of control activity thatmay exceed the performance capabilities of the aircraft, particularlyduring acute flight-path changes (Fig. 7). In this region, it can beobserved that the level of aircraft response is influenced by the time

horizon. For the short time horizon (7.5 s), the new flight path is

Z

X

ref

R

2

(Xa,Za)

(Xw,Zw)

C

X

a)

b)

Y

C

ref

R

2

(Xa,Ya)

(Xw,Yw)

Fig. 3 Relative position of synthetic waypoint and aircraft in inertialframe: a) lateral and b) longitudinal.

Guidance AircraftControl

Navigation

Guidance AircraftControl

Navigation

Fig. 4 Complete guidance loop structure block diagram.

-5000 0 5000 10000 15000 20000

-5000

0

5000

10000

15000FlightPath

NortherlyPosition(m)

Easterly Position (m)

TH

= 7.5

TH

= 10

TH

= 15

Flight Path

Fig. 5 Defined flight path and aircraft trajectory.

1.2 1.25 1.3 1.35 1.4 1.45 1.5

x 104

3500

4000

4500

5000

5500

6000

6500FlightPath

NortherlyPosition(m)

Easterly Position (m)

TH

= 7.5

TH

= 10

TH

= 15

Flight Path

Fig. 6 Vehicle trajectory at transition point.

604 J. GUIDANCE, VOL. 33, NO. 2: ENGINEERING NOTES

http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-

7/30/2019 AIAA-46204-831

5/6

acquired faster than in the longer-time-horizon cases (10 and 15 s).However, this is at the expense of closed-loop stability, with theaircraft requiring a greater level of activity to return to theflight path.

Figure 8 compares the virtual range disturbance between theaircraft and the synthetic waypoint over the duration of the flight for

the different time-horizon cases. It can be observed that the virtualranges converge to the desired value formulated in Eq. (4) (562.5 mforthe7.5scase,750mforthe10scase,and1125mforthe15scasegiven a constant airspeed of 75 m=s). The overall disturbance invirtual range remains relatively small for time-horizon changes,suggesting that the aircraft is maintaining a constant distance to thesynthetic waypoint throughout the flight.

Figure 9 shows the trajectory taken by the aircraft in following anacute-angle turn while being subjected to a steady 20 kt northeasterlyand20 kt southwesterly wind. The time horizon was set atTH 10 sat an airspeed of 75 m=s. By observing the responses in both thewind and no-wind cases, it can be seen that the wind affects theaircraft trajectory during a maneuver, but the algorithm is stableenoughto handlesuch disturbances andstillfollows thedesiredflightpath closely throughout the duration of the flight. When the wind isacting from the NE, the aircraft is pushed to the SW, restricting theaircrafts divergence fromtheflight path when compared with theno-wind case. When the wind is acting from the SW, the aircraft is

pushed to the NE, pushing the aircraft away from the desired flightpath. However, the aircraft is still able to reacquire the current legof the designated flight path once the maneuver is completed,demonstrating the SWG methods robustness to steady atmosphericdisturbances.

V. Conclusions

This Note has demonstrated the development and implementationof a synthetic-waypoint guidance algorithm into a 6-DOF aircraftmodel. It has been shown through nonlinear simulations that theperformance of an aircraft in following a specified trajectory isheavily dependent on guidance parameter changes: namely, timehorizon. The results presented in the nonlinear simulations providedinsight into the performance of the aircraft under various time-horizon conditions as well as demonstrating the algorithms abilityto handle atmospheric disturbances. It demonstrated that an increasein time horizon provided an improved stability in flight-path con-vergence at the expense of response time.

References

[1] Frew, E. W., Langelaan, J., and Sungmoon, J., Adaptive RecedingHorizon Control for Vision-Based Navigation of Small UnmannedAircraft, American Control Conference, Inst. of Electrical andElectronics Engineers, Piscataway, NJ, 2006, p. 6.

[2] Keviczky, T., and Balas, G. J., Flight Test of a Receding HorizonController for Autonomous UAV guidance, American ControlConference, Vol. 5, Inst. of Electrical and Electronics Engineers,Piscataway, NJ, 2005, pp. 35183523.

[3] Sprinkle, J., Eklund, J.M., Kim, H. J., and Sastry, S., Encoding AerialPursuit/Evasion Games with Fixed Wing Aircraft into a NonlinearModel Predictive Tracking Controller, 43rd IEEE Conference onDecision and Control, Vol. 3, Inst. of Electrical and ElectronicsEngineers, Piscataway, NJ, 2004, pp. 26092614.

[4] Zarchan, P., Tactical and Strategic Missile Guidance, 3rd ed., Vol.176,AIAA, Reston, VA, 1997.

[5] Taek Lyul, S., and Tae Yoon, U., Practical Guidance for HomingMissiles with Bearings-Only Measurements, IEEE Transactions onAerospace and Electronic Systems, Vol. 32, No. 1, 1996, pp. 434443.doi:10.1109/7.481284

[6] Shneydor, N., Missile Guidance and Pursuit: Kinematics, Dynamicsand Control, Horwood, Chichester, England, U.K., 1998.

[7] Lin, C. F., Modern Navigation, Guidance, and Control Processing,PrenticeHall, Englewood Cliffs, NJ, 1991.

[8] Gyu Taek, L., and Jang Gyu, L., Improved Command to Line of Sight

for Homing Guidance, IEEE Transactions on Aerospace andElectronic Systems, Vol. 31, No. 1, 1995, pp. 506510.doi:10.1109/7.366337

[9] Waldmann, J., Line-of-Sight Rate Estimation and Linearizing Control

0 500 1000 1500 2000 2500 3000 3500 4000 4500 50001

1.05

1.1

1.15

1.2

1.25

1.3

1.35

1.4

1.45

1.5x 10

4 FlightPath

Northerly

Position(m)

Easterly Position (m)

TH

= 7.5

TH

= 10

TH = 15

Flight Path

Fig. 7 Vehicle trajectory at acute transition point.

0 100 200 300 400 500 600 700 800-3

-2

-1

0

1

2

3Virtual Range Disturbance

Range(m)

Time (s)

TH

= 7.5

TH

= 10

TH

= 15

Fig. 8 Virtual range disturbance between aircraft and syntheticwaypoint.

0 500 1000 1500 2000 2500 3000 3500 4000 4500 50001

1.05

1.1

1.15

1.2

1.25

1.3

1.35

1.4

1.45

1.5x 10

4 FlightPath

Northerly

Position(m)

Easterly Position (m)

No wind

20 kt NE wind

20 kt SW wind

Flight Path

Fig. 9 Vehicle trajectory at acute transition point subject to 20 kt NEand 20 kt SW wind.

J. GUIDANCE, VOL. 33, NO. 2: ENGINEERING NOTES 605

http://-/?-http://-/?-http://-/?-http://dx.doi.org/10.1109/7.481284http://dx.doi.org/10.1109/7.366337http://dx.doi.org/10.1109/7.366337http://dx.doi.org/10.1109/7.481284http://-/?-http://-/?-http://-/?-

7/30/2019 AIAA-46204-831

6/6

of an Imaging Seeker in a Tactical Missile Guided by ProportionalNavigation, IEEE Transactions on Control Systems Technology,Vol. 10, No. 4, 2002, pp. 556567.doi:10.1109/TCST.2002.1014675

[10] Taek Lyul, S., and Tae Yoon, U., CLOS IRTH CompositeGuidance, IEEE Transactions on Aerospace and Electronic Systems,Vol. 33, No. 4, 1997, pp. 13391344.doi:10.1109/7.625135

[11] Yu, J., Liu, L., Zhao, H., and Xu, C., Robust Gain-Scheduled Con-troller Design for Air Defense Missile, Chinese Control Conference,

Inst. of Electrical and Electronics Engineers, Piscataway, NJ, 2006,pp. 713718.doi:10.1109/CHICC.2006.280726

[12] Raju, P. A., and Ghose, D., Empirical Virtual Sliding Target GuidanceLaw Design: An Aerodynamic Approach, IEEE Transactions onAerospace and Electronic Systems, Vol. 39, No. 4, 2003, pp. 11791190.doi:10.1109/TAES.2003.1261120

[13] Nelson, R. C., Flight Stability and Automatic Control, 2 ed., Vol. 1,McGrawHill, New York, 1998.

606 J. GUIDANCE, VOL. 33, NO. 2: ENGINEERING NOTES

http://dx.doi.org/10.1109/TCST.2002.1014675http://dx.doi.org/10.1109/7.625135http://dx.doi.org/10.1109/CHICC.2006.280726http://dx.doi.org/10.1109/TAES.2003.1261120http://dx.doi.org/10.1109/TAES.2003.1261120http://dx.doi.org/10.1109/CHICC.2006.280726http://dx.doi.org/10.1109/7.625135http://dx.doi.org/10.1109/TCST.2002.1014675