Original Article

Proc IMechE Part G:

J Aerospace Engineering
Aircraft electric anti-skid braking and 0(0) 1–19
! IMechE 2019

combined direction control system using Article reuse guidelines:

sagepub.com/journals-permissions

co-simulation and experimental methods DOI: 10.1177/0954410019852825

journals.sagepub.com/home/pig

Qiaozhi Yin, Hong Nie , Xiaohui Wei and Ming Zhang

Abstract
The work reported in this paper concentrates on the design and application of an electric ground control system
combining the braking and steering mechanisms on a small unmanned aerial vehicle. A virtual prototype of a small
unmanned aerial vehicle is built with a multibody dynamic software LMS Virtual.Lab Motion. An electric anti-skid braking
system and a new combined direction control system considering the sensors models are established with MATLAB/
Simulink. Optimizations are carried out using a global optimization command patternsearch first and then a local
optimization method fminsearch for fine-tuning to design the dynamic allocation for the direction rectifying weight
coefficients. Then a co-simulation method is introduced to study the ground maneuver performance so as to investigate
the interaction of each subsystem via the interfaces between the two softwares. The anti-skid braking simulation verifies
that the aircraft can stop smoothly and efficiently. The combined rectification control simulations in three different
conditions verify the system stability and robustness. In addition, an anti-skid braking and a direction-control experiment
are conducted. Results show that the experimental results fit well with the simulation and that the yaw angle can be
corrected effectively under the designed control systems.

Keywords
Co-simulation, virtual prototype, electric brake mechanism, direction control, optimization

Date received: 16 October 2017; accepted: 2 May 2019

build mathematical models. Barnes and Yager6 pre-

Introduction
sented the modeling method of the vehicle dynamics
Small unmanned aerial vehicle (UAV) has become considering the tire forces. Song et al.7 established a
one of the research focuses in the aeronautical area three-degree-of-freedom (DOF) dynamic model of a
in recent years due to the small size, low cost, wide taxiing flying-wing aircraft in the ground plane.
application range, no casualty, and high maneuver- Abzug8 built an aircraft six-DOF mathematical
ability.1 Most attention was devoted to the flight char- dynamic model during landing rollout, and Zhang
acteristics in the past, but as the requirements of the and Zhou9 adopted the Newton’s second law to estab-
UAV ground maneuvers are improved,2 and also as lish a complete nonlinear UAV taxiing model with
accidents occur frequently during the take-off and MATLAB/Simulink. Ro10 built a three-DOF landing
landing phases,3 it appears particularly important to gear dynamic model including the strut, tire, and
study the UAV ground motion performance. After wheel modules using the Lagrangian formulation.
the concept of ‘‘more electric’’ is proposed,4 applying Evans et al.11 showed the development of a complete
fly-by-wire technique in lieu of hydraulic systems to tricycle landing gear simulation model taking into
the braking and direction correction control systems consideration the shock absorber and tire forces.
will bring the advantages including low weight, secur- Then as the virtual prototype technique developed,
ity, reliability, high precision, and quick response.5 A
new direction correction control law is put forward, State Key Laboratory of Mechanics and Control of Mechanical
and the experiments of the UAV ground electric oper- Structures, Nanjing University of Aeronautics and Astronautics, PR
ating systems are conducted, providing with effective China
references and instructions for designing UAV ground
Corresponding author:
maneuvers control systems. Hong Nie, Nanjing University of Aeronautics and Astronautics, No. 29
On the aspect of aircraft ground dynamic model- Yudao Street, Qinhuai District, Nanjing, Jiangsu 210016, China.
ing, a number of studies have been carried out to Email: hnie@nuaa.edu.cn
2 Proc IMechE Part G: J Aerospace Engineering 0(0)

some researchers employed this method to establish combination of a differential brake and a rudder to
landing gear and aircraft dynamic models. A multi- rectify the deviation, where the rectifying weight
body dynamic method is able to simulate multi-body coefficient varied linearly with the aircraft velocity
dynamic systems with complex relative motions and changing. Hu25 coordinated a rudder and a nose-
highly nonlinear force laws. It possesses powerful ana- wheel steering system to correct the direction. The
lytical capabilities and can conduct static analysis, rectifying weight coefficient is designed, which
structure finite element analysis, kinematic, and changes proportionally to the square of the velocity.
dynamic analysis of every component. Wei et al.12 Dong et al.26 combined the nose-wheel steering and
built a virtual prototype of a half-axle landing gear the differential brake and designed a direction correc-
with detailed shock absorber forces with a dynamic tion distribution curve as an approximate S type to
software LMS Virtual.Lab Motion. Khapane13 built help the system running smoothly. However, the rect-
an aircraft model with accurate dynamic ground loads ifying weight coefficients above were all given accord-
for its ground maneuvers in a multibody simulation ing to experts’ experiences. Comparing with the
tool SIMPACK. Zhang et al.14 established a nose- previous works regarding to direction control, opti-
wheel steering system with LMS Imagine.Lab mizations are employed in this study to design the
AMESim to study the steering performance. dynamic allocation for the rectifying weight coeffi-
However, control modules are usually comparatively cients of the combined direction rectification control
simple in multi-body dynamic softwares. Therefore, law in order to reduce the UAV maximum lateral
some scholars use MATLAB15–17 to establish control displacement.
system models to study aircraft ground maneuver On account of the rapid development and the com-
characteristics. In this study, an advanced co-simula- plexity of an aircraft electric ground taxiing control
tion method is adopted via the interfaces between a systems, as well as the environmental and devices
virtual prototype model with finite element analysis influences, it is more accurate to employ the experi-
with a multibody dynamic software and a complete ment platforms and methods to carry out the research
ground operating control system with MATLAB to of the ground taxiing control systems and to study the
analyze the UAV ground performance during the roll- UAV ground performance. Zhang27 used the hard-
out process. ware-in-the-loop simulation technology as a test
Comparing with traditional hydraulic braking sys- method for a wheel brake system. Lin et al.28 designed
tems, an electric drive mechanism has great advan- a dual-redundancy electric braking system and
tages including response rapidity, security, good adopted the hardware in a loop simulation platform
robustness, and light weight. In 2005, Danielson18 to verify the stability and efficiency of the braking
applied its electric braking system to NASA’s system. Lu et al.29 used a four-wheel-drive fully elec-
Global Hawk and carried out a flight test. In 2010, tric vehicle prototype to carry out the experiments and
Anderson et al.19 and Hanlon20 designed an aircraft to evaluate the two examples of integrated yaw rate
electric braking mechanism and an electric braking and sideslip control systems. The development of this
actuator, respectively. Wei et al.21 established an elec- study is that the experiments of the electric braking
tric anti-skid braking system based on fuzzy–PID con- and new designed direction control systems are car-
troller with parameter self-adjustment feature. Zhang ried out on a small UAV demonstrator on the ground.
and Li22 introduced the principle of an aircraft electric In this paper, the ground taxiing control system
braking system and designed a BP neural network and the ground maneuver performance of a small
braking control system. After successful use of electric UAV are studied. In ‘‘Aircraft virtual prototype mod-
mechanisms for anti-skid braking systems, the appli- eling’’ section, the virtual prototype of a small UAV is
cation is extended to differential brake, cooperating built with a multibody dynamic software LMS
with electric nose steering gear to keep direction sta- Virtual.Lab Motion. Then the landing gear static
bility on a real small UAV in this study. simulations and the corresponding static tests are con-
In regard to the ground taxing direction correction ducted to prove that the landing gear model is accur-
system, if a yawing movement, resulting from the ate and the static strength meets the demand. In
crosswind, rough runway surface, initial landing yaw ‘‘Electric breaking system design’’ section, a dual-
angle, or asymmetry of the UAV structure, is not cor- redundancy electric braking mechanism, whose
rected in time, the aircraft may veer off a runway or volume and weight are strictly constrained on the
roll over, causing serious accidents. Yan and Wu23 small UAV, is designed and processed. Then an elec-
built a ground dynamic mathematical model of a tric braking system test is carried out. In ‘‘Rectifying
three-wheel UAV considering the lateral motion and control law design’’ section, a new direction rectifica-
adopted the rudder to correct the taxiing direction tion control system combining the differential brake
during the high-speed period, while the nose-wheel and the nose-wheel steering is put forward. The opti-
steering system is used during the low-speed mizations are conducted, and the polynomial fitting
period. Wang and Zhou24 built a UAV dynamic method is used to obtain the weight coefficient curve.
model and a direction correction control system In ‘‘Co-simulation and experiment design’’ section, a
with MATLAB/Simulink and then employed the co-simulation method and an experimental process
Yin et al. 3

are introduced in detail. In ‘‘Results and analysis of is fixed on the landing gear strut, while the wheel can
simulations and experiments’’ section, the designed spin on the axle. In addition, a rotational speed sensor
electric anti-skid braking control system and the is installed on the nose-wheel axle to calculate vx, the
combined direction control system are analyzed by aircraft velocity in x-direction in the body coordinate
the co-simulation method and experiments. The com- system (Sb  Ob Xb Yb Zb ).30 Due to the influence of the
parison of the results helps to verify the validity nose-wheel steering angle, there exists errors between
and stability of the ground operating control system. the nose-wheel speed and vx. However, the steering
The small UAV ground motion performance is also angle is small in this study, so that we can assume
studied. replacing vx with the nose-wheel speed is reasonable.
From the simulation and experimental results in
‘‘Results and analysis of simulations and experi-
Aircraft virtual prototype modeling ments’’ section, it can be seen that the steering angle
is always less than 2 , verifying that the assumption is
Aircraft as a multibody system
right.
The virtual prototype of a small UAV is built with a The MLG is fixed on the fuselage, and the direction
multibody dynamic software LMS Virtual.Lab motions of the main wheels are also fixed without
Motion shown in Figure 1. The UAV multibody steering function. Since the main wheels are always
dynamic model includes the fuselage, engine, nose rolling and slipping at the same time during a braking
landing gear (NLG) with a steering mechanism, and process, the speed sensors are mounted on the wheel
main landing gear (MLG) with an electric braking hubs to collect the rotational speeds and to calculate
system. The schematic diagram of the virtual UAV the slip ratio. The non-rotating brake stators are
during the taxiing and braking process is included in installed on the wheel axle, while the brake rotors
Figure 2. There are four main parts in Figure 2, are rotating with the wheel all the time. The relative
including Bodies, Force elements, Joints, and movement of the brake pads provides braking torque
Control nodes, which are all modeling parts in the not only to stop the UAV under the anti-skid brake
multibody dynamic software. Bodies contain the air- control, but also to correct the direction under the
craft fuselage, NLG components, and MLG compo- differential brake control.
nents. Force elements contain the shock absorber
forces in the landing gears, tires forces, and braking
Tire model
torques. Joints describe all the relative movements
and connection types among every component. For The nose tire model built with LMS Virtual.Lab
example, the NLG strut is connected to the fuselage Motion adopts a Complex Tire module based on
through a revolute joint with one DOF along the z- Moreland’s point contact theory.31 The lateral
direction, while the nose-wheel axle and the NLG force32 is one of the most interesting parts owing to
strut are connected using a bracket joint, so they are the nose tire steering and aircraft yawing movement.
fixed and 0 DOF is between them. In addition, control The nose tire force diagram is shown in Figure 3(a) in
nodes are the interfaces to exchange data between the which l is the steering angle that the steering gear can
multibody system in LMS Virtual.Lab Motion and control according to the UAV yaw angle and angular
the control systems in MATLAB/Simulink. velocity. The sideslip angle n of the nose tire is given
by Yan and Wu23
Landing gear system  
v y þ r  an
1
n ¼ l  tan ð1Þ
The landing gears used on the small UAV are of flat vx cos 
spring structures. The NLG subsystem with a steering
gear enables the nose wheel to steer and to correct the where vy is the y-direction aircraft velocity in the body
direction when taxiing on the ground. The wheel axle coordinate system, r is the yaw angular velocity, an is

Figure 1. Small UAV virtual prototype model.

4 Proc IMechE Part G: J Aerospace Engineering 0(0)

Figure 2. Small UAV multibody dynamic model schematic diagram.

Figure 3. Tire force analysis diagram: (a) nose tire force; (b) main tire force.

the distance between the nose wheel and the UAV of a braking tire mainly depends on the slip ratio, a
gravity center, and  is the attack angle. parameter used to reflect the braking tire slip level.
The lateral force Fy of the nose tire is calculated by The expression of the slip ratio  is
a piecewise function
vx  !  R g
 ¼ ð3Þ
K   n n 4c vx
Fy ¼ ð2Þ
ðFy Þmax n 4 c
where ! is the rotational speed of the main tire and Rg
where K is the cornering stiffness, c is the critical is the main tire rolling radius. The relationship
sideslip angle, and (Fy)max is the maximum lateral between the frictional coefficient  and the slip ratio
force. is defined as Wang and He34
The main tire model established with LMS
Virtual.Lab Motion applies the Magic Tire module  ¼0:8 sinð1:5344 tan1 ð14:0326ÞÞ ð4Þ
based on classic Pacejka Magic Formula tire
model.33 Figure 3(b) illustrates the main tire force when the slip ratio is 0.15, and the frictional coeffi-
during the braking process. The frictional coefficient cient  reaches the maximum value, bringing about
Yin et al. 5

Figure 4. Static tests of NLG and MLG: (a) NLG; (b) MLG.

the highest braking efficiency. The braking torque and The results also show that the largest principal stress
the friction torque generated by the friction force both on the MLG is about 110.14 MPa, taking place when
act on the main wheel to control the braking level. the UAV is yawing to the right. However, this value is
Therefore, the angular acceleration !_ of the main much lower than the material strength 350 MPa.
wheel is given by Therefore, the NLG and the MLG both meet the
design requirements and can be used during the roll-
f m  R g  Mb out and braking processes.
!_ ¼ ð5Þ
Jw

where fm is the friction force, Mb is the braking Electric braking system design
torque, and Jw is the wheel moment of inertia.
Electric brake mechanism design
Figure 5 shows the disc-type electric brake mechanism
Landing gear model verification
of the small UAV. It is designed with two direct current
According to the design parameters of the small UAV (DC) motors on each main wheel. This redundancy
and the landing gears, the ground forces acting on the design ensures the reliability of the braking system.
designed landing gears under different landing and The DC motor governor system model built with
taxiing conditions are calculated. Then the finite elem- MATLAB/Simulink takes the voltage as the input
ent static simulations and the corresponding static and the motor rotational speed as the output.21 The
tests are carried out under heavier load conditions. lead screws are driven by the DC motors and transfer
The results of the working conditions related to this the rotary movement into a rectilinear motion. Then
paper are listed and compared in this section. The they press on the brake pads and the braking torque
conditions simulated and tested on the MLG are the is generated. The axial displacement of the brake pads is
right yaw, braking, and left turn conditions, while on based on the turns of the ball screw
the NLG is the right turn condition. If the landing R
gears can bear loads from every direction under !h dt
l¼  L0 ð6Þ
these conditions, they satisfy the design requirements 2
of the static strength.
During the static experiments, the actual load where l is the axial displacement, !h is the ball screw
acting on the landing gear increases gradually and angular velocity, and L0 is the screw lead. The pres-
slowly from 0 to 100% calculated load. Figure 4 dem- sure sensors are fixed at the top of the lead screws to
onstrates the static tests of the NLG and the MLG. collect the real-time braking pressures. The data col-
The strains of several key positions are measured by lected by the two pressure sensors and the rotational
strain gages, and then the stresses and displacements speed sensor are transmitted to the braking control
can be obtained under different loads. The experimen- system processed in a single chip microcomputer
tal results are compared to the simulation nepho- (SCM). The braking pressure on one main wheel is
grams. The maximum values of the MLG are shown calculated as
in Table 1, while the results of the NLG are shown in
Table 2. Pb ¼ 2  Kb  l ð7Þ
As illustrated in Tables 1 and 2, the errors of the
tests and simulations are all located within 2%, indi- where Pb is the braking pressure of each lead screw
cating the accuracy of the landing gear models. and Kb is the stiffness of the braking pads. The
6 Proc IMechE Part G: J Aerospace Engineering 0(0)

Table 1. Comparison of simulation and test results of MLG static experiment.a

Right yaw Braking Left turn

Maximum principal stress (MPa)

Test 110.14 þ0.52% 76.95 1.85% 46.39 0.06%
Simulation 109.57 78.40 46.42
Maximum lateral displacement (mm)
Test 14.50 þ0.42% 7.96 0.38% 4.04 0.24%
Simulation 14.44 7.99 4.05
a
% Test errors with respect to simulation results are given in the right column, similarly in Table 2.

Table 2. Comparison of simulation and test results of NLG

static experiment.

Right yaw

Maximum principal Test 110.14 þ0.52%

stress (MPa) Simulation 109.57
Maximum lateral Test 14.50 þ0.42%
displacement (mm) Simulation 14.44

braking torque is produced by the friction force

between the braking stators and rotors

ðRs þ rr Þ
Mb ¼ mc  Nmc  Pb  ð8Þ
2
Figure 5. Electric brake mechanism design.
where mc is the friction coefficient between the brake
pads, Nmc is the number of the friction surfaces, Rs is
the external radius of the stators, and rr is the internal the motor drive employ digital signals to send data
radius of the rotors, respectively. to the SCM.

Electric anti-skid braking control system design Sensor model

The double closed-loop electric anti-skid braking con- Brake pressure sensor. In engineering application, the
trol system is designed to implement the anti-skid and dynamic property of a brake pressure sensor has
differential brake functions. Figure 6 illustrates the great influence on control system performance. A
schematic diagram of the braking control system pressure sensor modeled as a single DOF second-
with two PID controllers. The anti-skid PID braking order vibration linear system35 can simulate the
controller is employed to make the actual slip ratio  sensor dynamic performance accurately. Thus, a
approach to the optimal slip ratio  m to improve second-order transfer function H(s) is used to build
the braking efficiency. In addition, based on the the brake pressure sensor model
error of the actual and the target braking torques,
the braking torque PID controller is aimed at gener- Kbp !2nbp
ating an input signal into the DC motor drivers to HðsÞ ¼ ð9Þ
s2 þ 2bp !nbp s þ !2nbp
change the motor voltage U and trying to reduce the
error Mb to 0.
The double closed-loop electric braking control where Kbp is the proportionality coefficient of the pres-
system is built with MATLAB/Simulink, while the sure sensor, !nbp is the natural frequency of the sensor,
actual anti-skid PID braking control law is written and bp is the damping coefficient of the sensor.
as codes into the SCM in the experiment. The slip
ratio difference and the braking torque difference Aircraft velocity sensor. The aircraft velocity sensor col-
are calculated according to the data collected by the lects the aircraft velocity V during the rollout process.
speed sensors and the pressure sensors. The pressure The sensor can be represented by a gain coefficient in
sensor transfers the feedback data Mbm to the the model. However, the sensor measuring noise
SCM via analog signals, but the speed sensor and should be mainly taken into consideration. Since the
Yin et al. 7

Figure 6. Electric anti-skid braking control system diagram.

Figure 7. Measurement noise on the small UAV velocity used in the model.

aircraft velocity V is hard to measure, an accelerom- measure the yaw angle directly in lieu of an integral
eter is adopted to obtain the UAV longitudinal accel- operation. Napolitano et al.39 show that white
eration. Then the UAV velocity is estimated through Gaussian noise can also be adopted to simulate the
integration.36,37 This measurement of the UAV vel- angle sensor measuring noise. However, the variance
ocity will lead to relatively strong noise interference. of a yaw angle sensor noise is smaller than that gen-
Tanelli et al.38 show that white Gaussian noise can be erated by a velocity sensor owing to different measur-
used to simulate the velocity sensor noise of an air- ing principles. Figure 8 shows the measurement noise
craft or a car. Therefore, the measured variable V is nyaw affecting the small UAV yaw angle.
assumed to be affected by a digitized discrete
Gaussian white noise in the model, which is generated
Electric braking system model verification
by a MATLAB command wgn.
Figure 7 demonstrates the measurement noise nV The electric braking components are designed,
affecting the small UAV taxiing velocity. processed, and then installed on the MLG. The
electric braking mechanism experiment and the
Yaw angle sensor. Unlike the aircraft velocity sensor, corresponding simulation in MATLAB/Simulink
the yaw angle sensor used in the experiments can are carried out. Figure 9 shows the electric
8 Proc IMechE Part G: J Aerospace Engineering 0(0)

braking mechanism and the experiments on the quickly, and the lead screw stays still at the position.
small UAV. Owing to the environmental disturbance, sensor
Figure 10 illustrates the experimental and simula- errors, and mechanical device vibration, the max-
tion results. The braking pressure in MATLAB/ imum pressure error of the experiment and simulation
Simulink is calculated by the rotational speed of the is 8.33%, and the maximum rotating speed error is
DC step motor and the parameters of the electric 6.25%. The errors are both smaller than 10%, show-
actuators, while in the test, the pressure is obtained ing that this designed electric braking mechanism and
by the pressure sensors. From Figure 10(a), it can be control system are correct and effective.
seen that the step input signal of the braking pressure
on one braking actuator increases from 0 to 130 N at
1.4 s. The motor starts to work at the moment it Rectifying control law design
receives the braking pressure signal in Figure 10(b).
Electric differential braking control law
The motor rotating speed rises sharply to about
330 rad/s in 0.4 s. The motor rotor rotates in the for- One of the advantages of a differential brake is that no
ward direction, and the lead screw turns the rotational additional devices and space are needed to rectify the
motion into a linear movement. Then the lead screw direction during the taxiing process. The only thing to
pushes the braking stators and rotors to produce the change is to add a differential braking control law into
friction torque to brake the UAV. Figure 10 demon- the braking control system. The control schematic
strates that the braking pressure and the motor rotat- diagram of the electric differential brake is shown in
ing speed both fluctuate at 1–2 s, but the simulation Figure 11. The controller input signals are the actual
results are more stable and converge to the stable yaw angle and yaw rate r of the small UAV. The
values faster. When the braking pressure reaches the differential braking command signal b, calculated
target value, the motor rotating speed reduces to 0 by the rectifying control law, is transmitted to both
sides of the brake controller. The signal expression is
given by

b ¼ K þ Kr r ð10Þ

where K and Kr are the amplification coefficients of

the corresponding feedback variables.
While the differential brake system is rectifying the
taxiing direction, the anti-skid brake should be oper-
ated at the same time. Hence, the anti-skid and differ-
ential braking torques are allocated proportionally,
which is expressed as

Mbl ¼ 0:7Mblm
K þ Kr r5Kdb
Mbr ¼ 0:3Mbrm
 ð11Þ
Mbl ¼ 0:3Mblm
K þ Kr r4  Kdb
Figure 8. Measurement noise on the small UAV yaw angle Mbr ¼ 0:7Mbrm
used in the model.

Figure 9. Experiments of the electric braking system on the small UAV: (a) electric brake mechanism; (b) electric braking system
experiment.
Yin et al. 9

Figure 10. Experiment and the simulation of the UAV electric braking control system: (a) braking pressure; (b) motor rotor
rotational speed.

Figure 11. Schematic diagram of the electric differential braking control.

where Mbl, Mbr are the braking torques on the two which can increase the reliability. The nose-wheel
main wheels and Mblm, Mbrm are the braking torques steering control chart is presented in Figure 12.
obtained by the anti-skid braking control system on The input signals of the nose-wheel steering con-
the two main wheels, respectively. Kdb is the threshold troller are the same as those of the differential brake.
value to decide whether to start the differential brake. The adding weight-sum s of the yaw angle and yaw
Kdb is a small positive value used to prevent the exces- rate is transmitted into the PID controller. And this
sively frequent rectification resulting from the sensor signal is then transformed into a control command to
errors and very small yaw angle. Here, Kdb is equal to steer the nose wheel. If the small UAV sways to the
0.05. If the differential braking signal is larger than left, the steering actuator will generate a rightwards
Kdb, the small UAV veers toward the right. Then the steering moment to the nose wheel. The control law is
braking torque and also the ground frictional torque shown as
on the left braking wheel are larger, which turns the 8
UAV to the left. On the contrary, when the UAV < K
> s þ Krs r K s þ Krs r 4 Ks
 
veers to the left, the braking torque on the right s ¼ 0 K s þ Krs r 4Ks
wheel is larger than that on the left one. >
:
ðK s þ Krs rÞ K s þ Krs r 5 Ks
ð12Þ
Nose-wheel steering control law
The small UAV still employs a steering actuator to where K s is the amplification coefficient of the yaw
drive the nose-wheel steering gear to rectify the taxiing angle in the nose-wheel controller, Krs is the amplifi-
direction. During the taxiing experiments, two inde- cation coefficient of the yaw rate, and Ks is the thresh-
pendent steering control systems are designed: the old value to decide whether to start the nose-wheel
SCM control and the remote control unit (RCU), steering control.
10 Proc IMechE Part G: J Aerospace Engineering 0(0)

Figure 12. Nose-wheel steering control chart.

Figure 13. Combined direction control schematic diagram.

The weight coefficient

b is obtained by an opti-
Combined direction control law design
mization method. The UAV rollout velocity is located
The maximum taxiing velocity of the small UAV in in the range of V 2 [0,20] m/s. Therefore, during the
this study is about 20 m/s, located in the middle-low optimization process, 10 single-objective optimiza-
speed phase. Therefore, the combination of the dif- tions are conducted under 10 different UAV rollout
ferential brake and nose-wheel steering is able to velocities V ¼ 2n(n 2 [1,10], n 2 N). Since the lateral
achieve good effects when rectifying the direction. offset y and the yaw angle are both the most import-
The combined direction control40 will not only ant evaluation parameters in a direction control
enhance the system reliability, but a well-propor- system, the two norms  of the maximum lateral–par-
tioned combination will also improve the rectifying ameter errors are defined as the optimization objective
efficiency. The combined direction control schematic function
diagram is shown in Figure 13. The differential qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
brake system and the nose-wheel steering system  ¼ y2max þ 2max ð13Þ
are connected parallelly. In order to promote the
rectifying effect, the proportions of these two sys- where ymax is the maximum lateral displacement from
tems vary under different taxiing velocities V. The the runway centerline and max is the maximum yaw
rectification weight coefficient of the differential angle. Besides, the UAV should not taxi backward
brake is
b (
b 2 [0,1]), while the weight coefficient during the rollout process. Thus, the yaw angle is
of the nose-wheel steering is 1 
b. The yaw angle restrained by
, the yaw rate r, and the lateral displacement y  
change with the action of the combined direction  max
490 ð14Þ
control system.
Yin et al. 11

Figure 14. Polynomial fitting curve of rectification weight coefficient.

Also, owing to the fact that the runway width is experiments, the control law with the designed rectifi-
15 m, the lateral displacement should meet the condition cation weight coefficient is written in the SCM. The
  SCM then transmits the control signal to the braking
ymax 47:5 m ð15Þ control system on the main wheels and to the steering
system on the nose wheel.
For the optimizations carried out in this section, a
global optimization command patternsearch in
MATLAB is first used, and then another local opti- Co-simulation and experiment design
mization method fminsearch in MATLAB is adopted
Co-simulation implementation
for fine-tuning of the parameters to ensure that the
results obtained are optimum solutions. Here, the ini- The dynamic model of the small UAV in LMS
tial values of the local optimization are the optimiza- Virtual.Lab Motion is united with the control systems
tion results obtained by patternsearch. A similar in MATLAB/Simulink; therefore, the co-simulation
approach has been successfully used in Jiang et al.41 method is employed. The data in the two softwares
Then the 10 optimal values corresponding to 10 dif- are exchanged during the whole simulation process
ferent working conditions are obtained. A polynomial via the interfaces. Figure 15 demonstrates the struc-
curve is fitted based on these 10 optimal data points. ture diagram of the UAV ground motion co-simula-
The optimal points and a quadratic polynomial curve tion process. LMS Virtual.Lab Motion solves the
are illustrated in Figure 14. The red points are the mechanical system equations. The kinematics param-
optimal values, distributed uniformly on both sides eters exported from LMS Virtual.Lab Motion are the
of the black fitting curve. The R-squared value of forward velocity V, left and right wheel rotational
this quadratic polynomial fitting is 0.9778, very close speeds !l and !r, yaw angle , yaw rate r, and lateral
to 1, which indicates the high fitting degree. The poly- displacement y. The 3D simulation animation of the
nomial of this fitting curve is used to obtain the rec- UAV virtual prototype ground motion can also be
tification weight coefficients under different taxiing seen in LMS Virtual.Lab Motion. MATALB solves
velocities in the combined direction rectifying con- the equations of the control systems. The output
trol law. When the aircraft velocity V is smaller dynamic and control parameters of the ground con-
than 6 m/s, the weight coefficient is always equal to trol system in MATLAB are the left and right braking
0. Therefore, the specific formula of
b is given by torques Mbl, Mbr, and the steering angle l. The struc-
 ture and control parameters and their values used in
0:0032V2  0:0157V  0:0003 6 5 V420 the small UAV are shown in Table 3 in Appendix 1.

b ¼
0 0 5 V46
ð16Þ Design of experiments
This section shows the anti-skid braking and direction
The combined direction control system model is rectification experiments of the small UAV on the
built with MATLAB/Simulink, while in the ground. The schematic representation of the small
12 Proc IMechE Part G: J Aerospace Engineering 0(0)

Figure 15. Structure diagram of the UAV ground motion co-simulation process.

Figure 16. Schematic of the small UAV with test devices.

UAV with test devices is shown in Figure 16. The data from every sensor are controlled by the SCM,
engine is installed in front of the UAV to provide also called the vehicle control unit (VCU) here. The
power for its rollout on the ground. The anti-skid RCU is used to control the engine, the electric brake
braking system, the combined direction control system, and the nose-wheel steering system in case of
system, the gyroscopic compass measuring attitude an emergency, which ensures the security in the tests.
angles, and the data concentrator collecting all the Two battery packs are placed in the front and the
Yin et al. 13

back of the UAV fuselage, respectively, for both runway without crosswind. The initial yaw angle is 0
power supply and an adjustment of the gravity in the anti-skid braking experiment. The UAV is
center. Here, in the experiments, the sampling time accelerated to 20 m/s, and then the VCU sends
is 0.2 s, while the simulation step size only can be set instructions to the braking controller to actuate the
as variable-step during the co-simulation. As a result, electric braking mechanisms. The wheel speeds, the
the errors between the simulation and experimental braking pressures, and the main wheel slip ratios
results are unavoidable, and this technique problem can be obtained by the sensors and calculation. For
will form part of our future work. the direction stability experiment, the UAV is also
For the experiments (see Figure 17), the small accelerated to 20 m/s, and the UAV is deviated
UAV demonstrator is taxiing on a dry and smooth under the control of the RCU. When the yaw angle
is larger than 5 , the VCU starts the combined direc-
tion rectification control system. The obtained data of
the yaw angle and the steering angle are used to verify
this optimized combined direction system. The simu-
lation and experimental results are shown in the fol-
lowing section.

Results and analysis of simulations

and experiments
Anti-skid brake
Figure 17. The small UAV demonstrator during a ground The electric braking mechanism and the anti-skid
dynamic experiment. braking system introduced in ‘‘Electric breaking

Figure 18. Simulation and experiment results of anti-skid braking system: (a) UAV velocity; (b) wheel velocity; (c) slip ratio;
(d) braking pressure.
14 Proc IMechE Part G: J Aerospace Engineering 0(0)

system design’’ section are applied in the small UAV 7.2 s. The simulation curve fits well with the experi-
demonstrator and the virtual prototype. The simula- mental result. From Figure 18(c) and (d), it can be
tion and experimental results are given in Figure 18. seen that in the first 2 s, the braking pressure increases
During the simulation, the wheel speeds are acceler- faster in the experiment, and the experimental slip
ated in the first 0.5 s. In order to protect the wheels in ratio in 1–2 s is closer to the expected value. So, the
this process, the braking system starts to work from UAV demonstrator decelerates faster in this period.
0.5 s. Thus, the initial moment of the result curves in Moreover, the wheel slides seriously when the velocity
Figure 18 is 0.5 s in the simulation. Figure 18(a) and is lower than 2 m/s in the experiment. Therefore, the
(b) shows that the UAV decelerates constantly in slip ratio fluctuates sharply, and the velocity decreases

Figure 19. Simulation results of combined rectification control: (a) UAV velocity; (b) yaw angle; (c) steering angle of nose wheel;
(d) lateral displacement; (e) left braking pressure; (f) right braking pressure.
Yin et al. 15

slowly. Figure 18(c) and (d) also presents that the slip Figure 19(a) illustrates that the UAV is able to decel-
ratio and the braking pressure curves all vibrate in erate uniformly when rectifying a deviation. The brak-
both the simulation and experimental results due to ing time is longest while the initial yaw angle is 10 .
the sensor measuring noises. The overall trends of the From Figure 19(b), we can see the larger the initial
curves are very similar. The experimental braking yaw angle, the longer time the yaw angle converges to
pressure fluctuates more remarkably than that in the balanced state. But after 3 s, the yaw angles all remain
simulation by the reason that the sampling period is within 1 . As the UAV velocity decreases to 8 m/s, the
longer and the vibration exists in the experiment. nose-wheel steering plays a leading role in the rectifi-
However, the slip ratio is mostly close to the optimal cation process. At this moment, the yaw angle fluctu-
value, and the braking pressure remains stable at ates within a narrow range and tends to be stable
about 40—50 N. The results in Figure 18 reveal that gradually. Figure 19(c) shows the nose-wheel steering
this electric braking mechanism and the anti-skid angle curves, and the variation tendency is nearly the
braking system are of high braking efficiency. same with Figure 19(b). Equation (12) indicates that if
the yaw angle is very small, the nose-wheel steering
rectifying signal is 0. So, there exist platforms in the
Combined rectification control
steering curves. Figure 19(d) shows the lateral dis-
The combined rectification control law designed in placements in the rollout process. After 2 s, the lateral
‘‘Rectifying control law design’’ section is applied, displacements vary little due to that the UAV taxiing
and the initial taxiing speed is 20 m/s. The initial direction becomes stable gradually. Figure 19(e) and
yaw angles are 2 , 5 , and 10 . These three working (f) demonstrates the braking pressures on both sides
conditions are co-simulated, and the simulation under the control of the unity of the differential and
results are shown in Figure 19. The initial moment anti-skid brake. When the UAV veers toward the
of the result curves is 0.5 s in the simulation, the right, the left braking pressure is larger to turn the
same as the curves in ‘‘Anti-skid brake’’ section. UAV to the left. When the velocity is lower than

Figure 20. Simulation and experiment results of combined rectification control: (a) UAV velocity; (b) yaw angle; (c) steering angle of
nose wheel; (d) braking pressure.
16 Proc IMechE Part G: J Aerospace Engineering 0(0)

8 m/s, the braking system is mainly used to carry out 2. The experiment and the simulation of the electric
the anti-skid brake function. Thus, in this phase, the braking control system reveal that the maximum
difference between the left and right braking pressures pressure error of the experiment and the simula-
is smaller. tion is 8.33%, and the maximum rotating speed
Figure 20 illustrates the simulation and experimen- error is 6.25%, showing that the designed electric
tal results of the combined rectification control with a brake mechanism and the control system are cor-
5 initial yaw angle. From Figure 20(a) and (b), it can rect and effective.
be seen that at the first 1.5 s, the difference between 3. The optimization method is used to design the
the simulation and the experiment is small. The sensor dynamic allocation for the rectifying weight coef-
measurement noises lead to fluctuation in both the ficients of the combined direction control law. The
simulation and experiment. Nevertheless, after 3 s in 10 optimal weight coefficients under different velo-
the experiment, the yaw angle still vibrates with cities are fitted in a quadratic polynomial curve,
larger amplitudes owing to the apparatus errors and which reveals that the differential brake occupies
external disturbance. Moreover, the wheel sliding phe- the most proportion under higher velocity, while
nomenon still takes place in the low-speed period. the nose steering gear plays an important role
Figure 20(c) shows the nose-wheel steering angle under lower velocity.
curves. On account of the steering actuator control 4. The co-simulation method and the UAV demon-
precision in the experiment, the steering angle signal strator ground experiments are introduced. The
is larger than that in the simulation. But the variation results comparison illustrates that the experimen-
trends are the same at the first 1.5 s, and the maximum tal results fit well with the simulation under the
steering angle in the experiment is always smaller than anti-skid braking control. The electric braking
1.5 . After 3 s, the steering angle varies greatly by the mechanism and the anti-skid braking system are
reason of the yaw angle fluctuation. In addition, after of high braking efficiency. The UAV decelerates
5 s, the nose-wheel steering control accounts for the constantly in 7.2 s, but the experimental braking
largest proportion to rectify the deviation, so the pressure fluctuates more remarkably than that in
steering angle is large at this phase. Figure 20(d) pre- the simulation by the reason that the sampling
sents the braking pressures on the two main wheels. period is longer and the vibration exists in the
The results before 1.5 s and after 5 s in the experiment experiment.
fit well with the results in the simulation. From 1.5 s to 5. The combined rectification control simulations in
5 s in the experiment, the braking pressures fluctuate three different conditions verify the stability and
greatly to correct the UAV yaw angle, which is in robustness. From the results, it can be obtained
accordance with the variation trends in Figure 20(b) that larger initial yaw angle will lead to lower
and (c). Although the measuring precision, the struc- braking efficiency, bigger variation of the nose-
tural vibration, and external disturbance in the experi- wheel steering angle, and larger lateral
ment all result in the errors between the simulation displacement.
and the experiment, the two results are consistent as a 6. The rectification experiment result demonstrates
whole. And the results still show that this designed that the UAV taxiing direction can be corrected
electric ground operating control system is stable effectively under the designed electric mechanism
and valid. and the new combined direction law. Although the
measuring precision, the structural vibration, and
the external disturbance in the experiment all
Conclusion result in the errors between the simulation and
The virtual prototype technique is employed to build a the experiment, the two results are consistent as
small UAV dynamic model with LMS Virtual.Lab a whole.
Motion, which is co-simulated with the electric
ground operating control system in MATLAB/
Declaration of Conflicting Interests
Simulink to study the UAV ground maneuver per-
formance. The landing gears, the electric anti-skid The author(s) declared no potential conflicts of interest with
respect to the research, authorship, and/or publication of
braking mechanism and control system, and the opti-
this article.
mized combined rectification control law are designed
and tested. The conclusions are drawn below: Funding
The author(s) disclosed receipt of the following financial
1. The errors of the static tests and simulations of the
support for the research, authorship, and/or publication
designed NLG and the MLG are all located within of this article: the National Natural Science Foundation
2%, indicating the accuracy of the landing gear of China (Grant No.11372129), the China Postdoctoral
models. And the test results show that the NLG Science Foundation Funded Project (Grant No.
and MLG both meet the design requirements and 2019M650115), Jiangsu Planned Projects for Postdoctoral
can be used in the rollout and braking processes. Research Funds, the National Defense Outstanding Youth
Yin et al. 17

Science Foundation (Grant No. 2018-JCJQ-ZQ-053), the 14. Zhang M, Jiang RM and Nie H. Design and test of dual
Research Fund of State Key Laboratory of Mechanics actuator nose wheel steering system for large civil air-
and Control of Mechanical Structures (Nanjing University craft. Int J Aerosp Eng 2016; Article ID 1626015. DOI:
of Aeronautics and astronautics) (Grant No. MCMS- 10.1155/2016/1626015.
0217G01), the Fundamental Research Funds for the 15. Zhang M, Nie H and Zhu R. Design and dynamics
Central Universities (Grant No. NP2017401), the Priority analysis of antiskid braking system for aircraft with
Academic Program Development of Jiangsu Higher fourwheel bogie landing gears. Chin J Mech Eng 2011;
Education Institutions, Specialized Research Fund for the 24: 277–284.
Doctoral Program of Higher Education (Grant No. 16. Novellis DL, Sorniotti A and Gruber P. Driving modes
20123218120003). for designing the cornering response of fully electric
vehicles with multiple motors. Mech Syst Signal
ORCID iD Process 2015; 64–65: 1–15.
17. Sadraey M and Colgren R. Two DOF robust
Hong Nie https://orcid.org/0000-0002-1334-3649
nonlinear autopilot design for a small UAV using
Ming Zhang https://orcid.org/0000-0003-3151-9035
a combination of dynamic inversion and loop shaping.
In: AIAA guidance, navigation, and control conference
References and exhibit., San Francisco, California, USA, 2005,
1. Hou YX, Guan Y and Jia H. Research on motion char- pp.1–21.
acteristics for UAV ground maneuvers. In: 2015 IEEE 18. Danielson L. Electric braking debuts in military and
international conference on mechatronics and automa- commercial applications. SAE100 Future look, 2005.
tion, ICMA 2015, 2015, pp.22–26. 19. Anderson S, Brundrett LR, Corio FL, et al. Electric
2. Joshi K, Jeon S, Kwon H-J, et al. Braking avail- brake for aircraft. US 7,717,240 B2, Washington, DC:
ability tester for realistic assessment of aircraft landing U.S. Patent and Trademark Office, 2010.
distance on winter runways. J Aerosp Eng 2015; 28: 20. Hanlon C, Geck K and Quitmeyer NJ. Aircraft
1–10. electric brake actuator assembly with line replaceable
3. Krüger WR, Besselink IJM, Cowling D, et al. Aircraft actuator brake. U.S. Patent Application 12/704,971,
landing gear dynamics: simulation and control. Veh 2010.
Syst Dyn 1997; 28: 119–158. 21. Wei X, Yin Q, Nie H, et al. Aircraft electric anti-skid
4. Wheeler PW, Clare JC and Trentin A. An overview of braking system based on fuzzy-PID controller with par-
the more electrical aircraft. Proc Inst Mech Eng Part G ameter self-adjustment feature. Trans Nanjing Univ
J Aerosp Eng 2013; 227: 578–585. Aeronaut Astronaut 2014; 31: 111–118.
5. Wang B, Guo X, Zhang C, et al. Modeling and control 22. Zhang Z and Li Y. Modeling and simulation of
of an integrated electric parking brake system. switched reluctance machine based aircraft electric
J Franklin Inst 2014; 352: 626–644. brake system by BP neural network. In: Industrial elec-
6. Barnes AG and Yager TJ. Enhancement of aircraft tronics and applications (ICIEA), 2014 IEEE 9th con-
ground handling simulation capability. NASA ference on IEEE, 2014, pp.338–341.
Technical Report. Report no. AGARD-AG-333, Aug 23. Yan B and Wu C. Research on taxi modeling and
01 1998. taking-off control for UAV. In: 2014 seventh inter-
7. Song L, Yang H, Yan X, et al. A study of instability in national symposium on computational intelligence and
a miniature flying-wing aircraft in high-speed taxi. Chin design, 2014, pp.108–111.
J Aeronaut 2015; 28: 749–756. 24. Wang P and Zhou Z. Study of modeling and control
8. Abzug MJ. Directional stability and control during simulation for flying-wing UAV in ground motion
landing rollout. J Aircr 1999; 36: 584–590. during landing. J Syst Simul 2011; 23: 118–122.
9. Zhang H and Zhou Z. Modeling and direction-control- 25. Hu H. Study of control and simulation for
ling for flying-wing UAV in ground motion. J Syst UAV landing and ground taxiing. Chengdu,China:
Simul 2008; 20: 6759–6762. University of Electronic Science and Technology of
10. Ro K. A descriptive modeling and simulation of air- China, 2011.
craft-runway dynamics. In: 44th AIAA/ASME/ASCE/ 26. Dong S, Jiao Z, Sun X, et al. Dynamic allocation algo-
AHS/ASC structures, structural dynamics, and materials rithm for the gain of UAV nose wheel steering and dif-
conference, 7–10 April 2003, Norfolk,Virginia, ferential braking based on decomposition control. In:
American Institute of Aeronautics and Astronautics, Aircraft utility systems (AUS), IEEE international con-
Inc, 2003. Epub ahead of print 2003. DOI: 10.2514/ ference on IEEE Beijing, China, 2016, pp.831–835.
6.2003-1895. 27. Zhang H. Design and implementation of hardware-in-
11. Evans P, Perhinschi MG and Mullins S. Modeling and the-loop simulation test platform for wheel brake
simulation of a tricycle landing gear at normal and system. Changsha,China: Central South University,
abnormal conditions. In: AIAA modeling and simulation 2012.
technologies conference, 2–5 August 2010, Toronto, 28. Lin H, Wang Y, Zhang X, et al. Design and implemen-
Ontario, Canada, pp.1–20. tation of dual-redundancy electric braking system for
12. Wei X, Liu C, Liu X, et al. Improved model of landing- unmanned aerial vehicle based on DSP and CPLD.
gear drop dynamics. J Aircr 2014; 51: 695–700. Comput Meas 2014; 22: 1929–1931.
13. Khapane DP. Simulation of asymmetric landing and 29. Lu Q, Gentile P, Tota A, et al. Enhancing vehicle cor-
typical ground maneuvers for large transport aircraft. nering limit through sideslip and yaw rate control.
Aerosp Sci Technol 2003; 7: 611–619. Mech Syst Signal Process 2016; 75: 455–472.
18 Proc IMechE Part G: J Aerospace Engineering 0(0)

30. Stevens BL, Lewis FL and Johnson EN. Aircraft control Ks threshold value to decide whether to
and simulation: dynamics,controls design, and autono- start the nose-wheel steering control
mous systems. New York: Wiley, 2015. K cornering stiffness of the nose tire
31. Coetzee E. Shimmy in aircraft landing gear. Technical (N/rad)
Report No.56/2006, Airbus, 2006.
K s amplification coefficient of the yaw
32. Sura KN and Suryanarayan S. Lateral response of
angle in the nose-wheel controller
nose-wheel landing gear system to ground-induced exci-
tation. J Aircr 2007; 44: 1998–2005.
l axial displacement of the brake
33. Pacejka BH and Bakker E. The magic formula tyre pads (m)
model. Veh Syst Dyn 1992; 21: 1–18. L0 screw lead (m)
34. Wang J and He C. Determination of tire-runway fric- Mb braking torque (N m)
tion coefficient. J Northwest Polytech Univ 2000; 18: Mbl, Mbr braking torques on the two main
569–571. wheels (N m)
35. Chang Z and Tang B. A mathematical modeling Mblm, Mbrm braking torques obtained by the
method of strain gage pressure transducers with inlet anti-skid braking control system on
pipe or pressure chamber. Meas Contr Technol 1999; the two main wheels (N m)
18: 6–8. Mbm actual braking torque (N m)
36. Savaresi SM, Silani E and Bittanti S. Acceleration- Nmc number of the friction surfaces
driven-damper (ADD): an optimal control algorithm Pb braking pressure of each lead
for comfort-oriented semiactive suspensions. J Dyn
screw (N)
Syst Meas Contr 2005; 127: 218–229.
r yaw rate (rad/s)
37. Bittanti S and Savaresi SM. On the parameterization
and design of an extended Kalman filter frequency rr internal radius of the rotors (m)
tracker. IEEE Trans Autom Contr 2000; 45: 1718–1724. Rs external radius of the stators (m)
38. Tanelli M, Piroddi L and Savaresi SM. Real-time iden- U motor voltage (V)
tification of tire-road friction conditions. IET Control vx,vy x-direction and y-direction aircraft
Theory Appl 2009; 3: 891–906. velocities in the body coordinate
39. Napolitano MR, An Y and Seanor BA. A fault tolerant system (m/s)
flight control system for sensor and actuator failures V aircraft velocity (m/s)
using neural networks. Aircr Des 2000; 3: 103–128. y lateral displacement (m)
40. Hao X, Yang Y, Jia Z, et al. Mathematic model and ymax maximum lateral displacement (m)
deviation-correction control for UAV taxiing. Electr
Mach Control 2014; 18: 85–92.  attack angle (rad)
41. Jiang JZ, Matamoros-Sanchez AZ, Zolotas A, et al. c nose-wheel critical sideslip
Passive suspensions for ride quality improvement of angle (rad)
two-axle railway vehicles. Proc IMechE, Part F: J n nose-wheel sideslip angle (rad)
Rail Rapid Transition 2013; 229: 315–329. b differential braking command signal
s Nose-wheel steering command signal

b rectification weight coefficient of the
Appendix differential brake
 frictional coefficient between the tire
Notation and the ground
an distance between the nose wheel and mc friction coefficient between the brake
the gravity center in x-direction (m) pads
fm friction force of the main tire (N) l nose-wheel steering angle (rad)
Fy lateral force of the nose tire (N)  slip ratio
(Fy)max maximum lateral force of the nose m optimal slip ratio
tire (N) ! main tire rotational speed (rad/s)
H(s) transfer function of the brake pres- !h ball screw angular velocity (rad/s)
sure sensor !nbp natural frequency of the pressure
Jw main wheel moment of inertia sensor (Hz)
(kgm2)  optimization objective function
Kb stiffness of the braking pads (N/m) yaw angle (rad)
Kbp proportionality coefficient of the max maximum yaw angle (rad)
pressure sensor bp damping coefficient of the pressure
Kdb threshold value to decide whether to sensor (N/(m/s))
start the differential brake Mb error of the target braking torque
Krs amplification coefficient of the yaw and the actual braking torque (Nm)
rate in the nose-wheel controller
Yin et al. 19

Appendix 1

Table 3. Parameters and their values used in the small UAV model.

Parameter Value (default)

Geometry and structure parameters

UAV mass, M (kg) 24.69
Distance between NLG and MLG, a (m) 0.49
Distance between two main wheels, bw (m) 0.77
Height of mass center, h (m) 0.36
Distance between center of gravity and MLG, am (m) 0.085
Original radius of main wheel, Rw (m) 0.09
Rolling radius of main wheel, Rg (m) 0.075

Control parameters
Control coefficients of speed feedback loop in anti-skid braking system Proportional, Kpv 102
Integral, Kiv 9.4
Differential, Kdv 5
Control coefficients of brake torque feedback loop in anti-skid braking system Proportional, Kpt 11.1
Integral, Kit 1.3
Differential, Kdt 0.5
Control coefficients of nose-wheel steering system Proportional, Kpn 0.4
Integral, Kin 0.01
Differential, Kdn 0.003
Amplification coefficients in differential braking system Yaw angle, K 2.5
Yaw rate, Kr 0.95