IEEE - Aerospace and Electronic Systems - July 2023 - 28

Flight Control Method Using Neural Network in Prediction for Suppressing Ship Airwake Impact in Carrier Landing
Let EvðkÞ represent the error between the reference
trajectory of the state quantity and the predicted trajectory
on the longitudinal channel. Then
EvðkÞ¼ YrefvðkÞ FvxvðkÞ Gvuvðk 1Þ LvDvðkÞ:
(21)
The optimal objective function Jkv can be rewritten as
Jkv ¼ GvDUvðkÞ EvðkÞkk2
v ðkÞGT
¼½DUT
þ DUT
¼ DUT
2ET
vET
v ðkÞ½GT
v ðkÞRvDUvðkÞ
v QvGyv þRvDUvðkÞ
v ðkÞQvGvDUvðkÞþ ET
Let Hv ¼ 2ðGT
1
2
DUT
v ðkÞQvEvðkÞ:
yvQvGyv þRvÞ, and fv ¼2GT
EvðkÞ. Equation (21) can be written in standard form
Jkv ¼
v ðkÞHvDUvðkÞþ fT
v DUvðkÞþ const:
yvQv
Qvþ DUvðkÞkk2
Rv
v ðkÞQv½GvDUvðkÞ EvðkÞ
(22)
It is also necessary to determine the optimization time
domain P and the control time domain M.The longer the
optimization time domain is to improve the stability of the
control, but the accumulation oferrors in the prediction model
will increase; the longer the control time domain, the greater
the degree offreedom provided to the control system, but the
more unstable the control system. And the increase ofP and
M will affect the real-time performance. Therefore, choosing
the length ofP andM requires comprehensive consideration
ofthe stability and rapidity ofthe control system.
In order to correct the system state error caused by the
prediction model error, it is necessary to add a feedback
link on the basis of the original open-loop control. Then
recalculate prediction and control based on the real state
at the current moment to realize feedback correction in
predictive control. The state value after feedback correction
at the (kþ1) time is written as
(23)
Considering the steering gear control has position and
speed limits, it is necessary to add corresponding constraints
to solve the constrained optimization problem of
the objective function
minDUðkÞ
1
2 DUT
v ðkÞHvDUvðkÞþ fT
jjDUvðkÞ DUvmax
v DUvðkÞ
:
(24)
In the same way, the objective function of the carrierbased
aircraft lateral controller Jkl is as follows:
minDUðkÞ
1
2 DUT
l ðkÞHlDUlðkÞþ fT
jjDUlðkÞ DUlmax
l DUlðkÞ
(25)
where Hl and fl have the same form as Hv and fv in the
longitudinal objective function, respectively.
All of the above-mentioned control rudder surfaces
such as elevator, rudder, and aileron have speed and position
limits, and the specific parameters are related to the
aircraft model.
Table 2.
Initial Conditions of Carrier Aircraft Simulation
Air speed (m/s)
AOA ()
Sideslip angle ()
Roll angle ()
Pitch angle ()
Relative heading ()
28
76.16
7.47
1.22
4.00
xvcorðkþ1Þ¼ xvðkþ1jkÞþNv½xvðkþ1Þ xvðkþ1jkÞ
xlcorðkþ1Þ¼ xlðkþ1jkÞþNl½xlðkþ1Þ xlðkþ1jkÞ
(26)
where Nv and Nl are both the weight matrix of error correction,
which are set according to the measurement accuracy
of the system, respectively.
SIMULATION
A simulation is performed to verify the effect of the proposed
landing control method that can suppress the interference
ofthe airwake. Suppose the carrier speed is 10 m/s and
the deck wind speed is 15 m/s. The carrier aircraft is in the
landing state and its initial parameters are in Table 2.And in
each simulation, randomvariables, namely the randomcomponents
ofthe ship airwake and the measurement error noise
of all sensors in the carrier and aircraft, will change. The
magnitude ofthe random components of the ship airwake is
set according to the American Standard MIL-F-8785C
(1980) [24]. The distance, azimuth, and elevation errors of
the aircraft relative to the runway, which is obtained from
shipborne radar, are 5 m, 0.12,and 0.12, respectively [19].
Roll angle rate (/s)
Pitch angle rate (/s)
Heading angle rate (/s)
Lateral distance (m)
Longitudinal distance (m)
Relative height (m)
IEEE A&E SYSTEMS MAGAZINE
1383
96.77
JULY 2023
IEEE - Aerospace and Electronic Systems - July 2023

Table of Contents for the Digital Edition of IEEE - Aerospace and Electronic Systems - July 2023
