Beruflich Dokumente
Kultur Dokumente
1. INTRODUCTION
Vision tracking technology is used in mobile robots to
enhance their visual sensing ability and enable them to track
the target continuously. Vision tracking systems using robot
motion information have been researched for number of years,
for instance, E. S. Shim et al. (2009), and A. Lenz et al.
(2008). However, in vision tracking systems accurate
localization of mobile robot for different robot motion
scenarios and environmental conditions remains an essential
task.
For accurate localization of mobile robot, various sensors and
techniques have been employed and are characterized as:
relative localization and absolute localization, J. Borenstein et
al. (1997). Relative localization or dead reckoning technique
uses kinematic model of the robot to compute the position of
the robot relative to its start position. It determines the
position and orientation using on-board sensors, such as
encoders, gyroscopes, accelerometers etc. However, the
conventional dead-reckoning method has the problem of
accumulating wheel slippage error which limits its
application. The absolute localization technique obtains the
absolute position of robot using beacons, landmarks or
satellite-based signals such as Global Positioning System
(GPS). The position of the robot is externally determined and
is independent from integration of noisy data or wheel
slippage of mobile robot.
Copyright by the
International Federation of Automatic Control (IFAC)
9379
2.1 Odometry
Odometric estimation is based on the data from robots
encoder. The encoder is a sensor attached to a rotating object
(such as a wheel or motor) to measure rotation. By measuring
rotation we can determine displacement, velocity,
acceleration, or the angle of a rotating sensor.
Robots position and orientation can be determined by using
the robot velocity and orientation obtained from the encoder
data.
pk
xk
yk
T k
rk lk
2
xk 1 vk 1.+t.cos(T k 1)
yk 1 vk 1.+t.sin(T k 1)
T k 1 wk 1.+t
rk lk
d
(1)
(2)
v k 1
vk a k ' t
x k 1
xk vk ' t
(4)
T k T k 1 w 't
(3)
9380
Voltage
(DC V)
356818
Tag
905817
Sonar
353515
Beacon
3.3 5 %
Tag
5~15
RF
transmission
Update
Time
Data
Channel
Detection
Range (m)
Accuracy
2.4~2.485
GHz
x
100 ms
where
128 channel
552.5
(6)
10 cm, 2
y T e T g
xk 1
f ( xk , wk )
rk lk
xk cos T e ,k 2
y sin T rk lk
e,k
k
k
2 w
T e ,k Ze, k 't
T g ,k Z g ,k 't
(7)
(5)
k n l 1
the process noise added by the sensor noise, and 't is the
sampling time. Partial matrix for estimating error covariance
and error covariance matrix is shown in (8) and (9)
respectively.
Fk
Pk 1
wf
wx
x xk
0
0
rk lk
2
r l
cos Tk k k
2
1
0 sin Tk
1
Fk Pk Fk T Qk
0
1
(8)
(9)
zk
9381
h( x k , vk ) T e,k T g , k vk
(10)
Hk
where
wh
wx k
>0
0 1 1@
(11)
wf
wx
Fk
Kk
Pk H k T ( H k Pk H k T M k Rk M k T ) 1
x xk
Fk Pk Fk Qk
zk
h( x k , vk )
Hk
x
>x
y T@
(13)
is the
xk 1
f ( xk , wk )
wh
>1 1 1@
wx k
Pk H k T ( H k Pk H k T M k Rk M k T ) 1
x k x K k [ zk hk ( x k , 0)]
Pk ( I K k H k ) Pk
(17)
(18)
(19)
measurement noise,
(14)
Kk
x k
y k vk
T k
xk vk cos Tk 't
y k vk sin Tk 't wk
T k Zk 't
(16)
(12)
where
measurement noise,
(15)
Pk 1
x k x K k [ zk hk ( x k , 0)]
Pk ( I K k H k ) Pk
0 1 vk cos Tk 't
0 0
+T
tan 1 (
X x
) T
Yy
(20)
9382
Experiment
Linear
Motion
Rectangular
Motion
Sensors
Final
position/Angle
Errors
Encoder &
Gyroscope
15 mm
EKF
5 mm
Encoder &
Gyroscope
EKF
Vision
Tracking
Success
Rate
Encoder
296 mm
EKF
25 mm
Slip test
98 %
x-axis: 95 mm,
y-axis: 30 mm,
heading: 1.89
x-axis: 16 mm,
y-axis: 20 mm,
heading: 1.02
90 %
96 %
6. CONCLUSION
This paper presents a vision tracking system by using dead
reckoning and absolute localization technique. The designed
system integrates the information received from the encoder,
9384