2. Imu Sensor (Common Sensor)
• shape
• dead recking sensor
• roburst and high frequency
data
• error accumulation with
time
• typical sensors(DVL,
USBL) can correct
accumulated error
figure: imu sensor shape
figure: block diagram of imu sensor
3. Why to use cameras and Laser pointers
• typical sensors are good
enough for shallow water
• typical sensors has lack of
percision in underwater
environment
• cameras and Laser
pointers have higher
frequency than acoustic
but limited range with
accurate results.
figure: inspection of ship hull
figure: laser vision system
4. previous work with IMU and LVS
• integration of IMU and
GPS
• Integration of IMU and
DVL
• methodologies are made
using LVS to calibrate yaw
and correct parpendicular
distances.
figure: laser plane image to be seen by camera
figure: laser pointer image
5. Purpose of this research work
• propose velocity and
position vector
• LVS gives position vector
of vehicle refered to target
• Imu gives acceleration and
velocity of vehicle
• LVS and Imu are fused in
KF based system
figure: position, vilocity and acceleration vector
figure: Kalman Filter diagram
6. Benefits of using KF for fusion of Data
• MKF produces optimal
position vector as input for
closed loop position
control system
• MKF produces position
and velocity vector even if
there is no input from LVS
for 10 seconds
figure: closed loop position control system
figure: position estimation reference
7. Methodology
• xyz are for position and
others are for orientation in
n vector
• uvw are linear velocities
and pqr are angular
velocities
• angles and angular
velocities agains x and y
are stable because of
vehicle.
figure: position vector of vehicle
figure: velocity vector of vehicle
figure: x position and psi orientation of vehicle
8. Methodology - LVS
• 2 laser pointers and 1
charged copupled device
CCD camera
• position of x axis and
orientation of z axis is
achieved by L1 L2 and psi.
• mapping is achieved by
polynomial because
triangular mapping has
hardware constraints
polynomial plot
triangular mapping
9. Methodology - LVS - Tracking of Target
• active contour(snake)
vision (arround object)
• features are lines and
edges
• object is selected
fig: active contour
fig: snake curves
figL sx and sy are center of target, xo yo are center of
image ax, ay the camera focal lengths for xy image axis
10. Methodology - IMU used here
• 3 accelerometer and 3
gyroscopes
• strap down configuration
(fixed)
• practice of mathematical
model leads to unbound
position error of IMU
• external sensor is needed
for IMU
11. Methodology - LVS/IMU based MSKF
• in linear stationary model x
is state vector and w is
white noise and F is state
transition metrix
• measurement of N sensor
can be done by this
equation
• i sensors and k values of
each sensor
12. Methodology - LVS/IMU based MSKF
• Estimation stage of
kalman filter where K is
kalman gain for i sensor
while P is uncertainty
• Prediction stage is defined
as
13. Methodology - LVS/IMU based MSKF- Wiener
process acceleration model (stochastic)
• xk(state vector) with first 9
states of position velocity
and acceleration and last
two states tells angle and
angular velocity around z
axis
14. Methodology - LVS/IMU based MSKF- Wiener
process acceleration model (stochastic)
• data fusion decision
• target refer frame position
and yaw can find out from
LVS
•
15. Methodology - LVS/IMU based MSKF- Wiener
process acceleration model (stochastic)
• IMU is main sensor while
LVS is external sensor
• If LVS not available then
only IMU
• othervise
16. Experiments
• 1) teleoperation using a
joystick
• 2) closed loop position
control system
• Experiment is operated in
a pool
17. Exprements - system components
• module
• control system
• joystick
• laser pointer
• CCD camera
• IMU
• Aluminium object
18. Experiment - 1 teleoperation scenerio
• ROV is teleoperated by
joystick
• while user detects object
kalman filter fuse LVS and
IMU
19. Experiment - 2 Closed Loop Position Control
Scenario
• To derive the vehicle on
desired location
• A motion controller is
implemented by kinematic
controller on xy and PD
controller on z
• a is difference of principle
axis and distant vector e
20. conclusion
• fuse IMU and LVS as
position sensor
• LVS is external sensor of
IMU
• Proposed method can be
used for closed loop
position control as it
provide smooth
measurement at high
frequency