1 |
Planetary rovers and data fusionMasuku, Anthony Dumisani 05 1900 (has links)
This research will investigate the problem of position estimation for planetary rovers.
Diverse algorithmic filters are available for collecting input data and transforming
that data to useful information for the purpose of position estimation process. The
terrain has sandy soil which might cause slipping of the robot, and small stones and
pebbles which can affect trajectory.
The Kalman Filter, a state estimation algorithm was used for fusing the sensor data
to improve the position measurement of the rover. For the rover application the
locomotion and errors accumulated by the rover is compensated by the Kalman
Filter. The movement of a rover in a rough terrain is challenging especially with
limited sensors to tackle the problem. Thus, an initiative was taken to test drive
the rover during the field trial and expose the mobile platform to hard ground and
soft ground(sand). It was found that the LSV system produced speckle image and
values which proved invaluable for further research and for the implementation of
data fusion.
During the field trial,It was also discovered that in a at hard surface the problem
of the steering rover is minimal. However, when the rover was under the influence
of soft sand the rover tended to drift away and struggled to navigate.
This research introduced the laser speckle velocimetry as an alternative for odometric
measurement. LSV data was gathered during the field trial to further simulate under
MATLAB, which is a computational/mathematical programming software used for
the simulation of the rover trajectory. The wheel encoders came with associated
errors during the position measurement process. This was observed during the
earlier field trials too. It was also discovered that the Laser Speckle Velocimetry
measurement was able to measure accurately the position measurement but at the
same time sensitivity of the optics produced noise which needed to be addressed as
error problem.
Though the rough terrain is found in Mars, this paper is applicable to a terrestrial
robot on Earth. There are regions in Earth which have rough terrains and regions
which are hard to measure with encoders. This is especially true concerning icy
places like Antarctica, Greenland and others.
The proposed implementation for the development of the locomotion system is to
model a system for the position estimation through the use of simulation and collecting data using the LSV. Two simulations are performed, one is the differential
drive of a two wheel robot and the second involves the fusion of the differential drive
robot data and the LSV data collected from the rover testbed. The results have
been positive. The expected contributions from the research work includes a design
of a LSV system to aid the locomotion measurement system.
Simulation results show the effect of different sensors and velocity of the robot. The
kalman filter improves the position estimation process.
|
Page generated in 0.0749 seconds