Information Technology Reference
In-Depth Information
The physical nature of the sensors is irrelevant to the simulations. In a real case,
the distance can be obtained by computer vision, ultrasounds or a similar system,
while the speed can be extracted from the CAN bus, or can be provided by a sensor
fitted for that purpose.
14.2.3 Fuzzy Control System
Since the longitudinal dynamics of a car is very complex and highly nonlinear,
traditional controllers such as Proportional-Integral-Derivative (PID) are not well
suited to this task. For this reason it was decided to use a Takagi-Sugeno fuzzy
system with a constant value in the consequent term of each rule. The inputs, called
distance
and
speed
, are provided by the sensory system, and the output, called
accel
,
is a signal between
100 and 100 which is applied onto the accelerator pedal when
positive, or onto the brake pedal if negative, avoiding actuation on both pedals at the
same time.
Three Gaussian membership functions have been assumed for each input, evenly
spaced in their respective domains, as shown in Fig.
14.2
. With these membership
functions there is a total of nine rules such as:
−
IF
distance
IS
distance_mf_n
AND
speed
IS
speed_mf_m
THEN
accel
IS
v
[
n
,
m
]
where the product of the terms in the antecedent is taken as the AND function. Each
term
v
[
,
]
−
n
m
is a singleton in the range from
100 to 100. The output value is
obtained in the usual way:
2
0
μ
n
(
distance
)
·
μ
m
(
speed
)
·
v
[
n
,
m
]
n
,
m
=
accel
=
(14.1)
2
0
μ
n
(
distance
)
·
μ
m
(
speed
)
n
,
m
=
μ
i
(
being the degree of membership of the input
x
into the fuzzy set
i
in its domain.
The output value thus obtained is applied to the accelerator or the brake depending
on its sign:
x
)
accel
i f accel
>
0
ac
=
0
i f accel
≤
0
0
(14.2)
i f accel
>
0
br
=
−
accel
i f accel
≤
0
The nine values
v
[
n
,
m
]
are obtained bymeans of a genetic algorithm, as described
in Sect.
14.3
.
Search WWH ::
Custom Search