TELKOMNIKA ISSN: 1693-6930
Flight Reliability of Multi Rotor UAV Based on Genetic Algorithm Huang Ronghui 233
un-modeled error and put forward a final pseudo-control input form containing reference model, PD controller and neural network compensator to obtain the optimal estimation of motor attitude
and measure the drift error of rotor meter to realize the correction of drift error.
2.2. Rotating Speed Control Based on Genetic Algorithm
Genetic algorithm GA is a kind of calculation method simulating heredity and evolution process of biology, the principle of which is the evolution principle of “Natural selection and
survival of the fittest” in the biological universe [11,12]. PID controller is a kind of linear controller. R t is process quantity while Y t is the set value quantity and the difference value
between them is the deviation of the system [13-17]. The deviation is formed to be the controlled quantity u t of the controlled object through the controller after the linear
combination of three links
P, I and D, the regularity of which is:
dt t
de K
dt t
e K
t e
K t
u
d t
i p
6 Where,
is the proportionality coefficient, is integral coefficient and differential
coefficient [10]. In the engineering control, the absolute integral error is generally adopted as the
performance index of evaluation ITAE:
dt t
e t
J
t
7 To avoid overshooting, the function of penalty is adopted. In case of overshooting, the
overshoot will serve as an item of objective function. The optimal index at that time is:
dt t
e Ct
dt t
e t
ITAT J
t
8 It can be obtained after the discretization of equation 8:
1
1
k e
k e
T T
j e
T T
t e
K t
u
d k
j i
p
9 In equation 8,
C is penalty coefficient and when ,
C=0; in equation 9, , are respectively the integral and differential time coefficient.
Step 1: calculate PID parameter values in accordance with Z-N method:
,
,
,
and
,
; Step 2: population quantity of ant is m and each ant
k k=1~m has 15 attributes to store ordinate values of 15 nodes of paths of ants and crawling path.
Step 3: parameter initialization of mixed algorithm: set t=0 and ; assign
and initial time i=1~15, j=0~9; set ∆
,
,
and put all ants at the initial point O. Step 4: set variable i=1, if
, calculate the probability for the ants to transfer towards each node of the line segment in accordance with equation 6; otherwise, adopt
Roulette Wheel Selection to select the next node in accordance with equation 1 and store the value of the selected node in the tabu table
at the same time. Step 5: after each ant finishes a node, update local information element in accordance
with equation 7; make an adaptive adjustment to the local information volatilization coefficient in accordance with equation 8;
Step 6: set i=i+1, if i ≤15, go to step 3; otherwise, execute step 7;
Step 7: calculate the corresponding PID parameters ,
, of the path in accordance with array
and equation 9; carry out computer simulation to obtain the performance index
, steady-state regulation error and overshoot
of the system; calculate the corresponding objective function
of ant k in accordance with equation 9;
record the optimal path and optimal path performance index in this loop and store ,
, in
∗
,
∗
,
∗
;
ISSN: 1693-6930
TELKOMNIKA Vol. 14, No. 2A, June 2016 : 231 – 240
234 Step 8: if
and the whole ant colony has not converged to walk in the same path, set all ants at the initial point
O again and go to step 4: otherwise, the loop is ended and output the optimal path and its corresponding optimal PID parameters
∗
,
∗
,
∗
.
2.3. Direct Torque Control Current Servo