Navigation Solution for Marine Applications using MEMS-based Sensors and GPS

Student thesis: Master thesis (including HD thesis)

  • Bjørn Eskildsen
This thesis concerns the estimation of the position, velocity and
attitude using inertial sensors, magnetometer and GPS.
The purpose is to apply a model based estimation filter to fuse
these sensor inputs together to form estimations of the states of
This thesis has been made in collaboration with the company:
CDL - inertial engineering.
The main hardware platform is a MiniSense2 (MS2), produced
by CDL. The sensors on the MS2 platform consist of a 3-axis
gyroscope, a 3-axis accelerometer and a 3-axis magnetometer.
These are all low-price MEMS based sensors.
In this thesis, it is chosen to base the design of the navigation
solution on a loosely coupling between the sensors of the
MiniSense2 and GPS. The estimation filter utilized in this is
based on the structure of the extended Kalman filter.
Differential equations regarding motion on the rotating earth are
derived together with models of the sensors, which includes the
assumption of additive biases of the MS2 sensors. The thesis also
concerns models of the gravitational field and magnetic field of
the earth.
As the extended Kalman filter is chosen as basis for the filter
design, a linear small state perturbation model is also addressed.
Two tests have been conducted for evaluation of the designed
navigation solution.
The first test was based on a designed simulation environment,
as this has the advantage of having the true values of the system
states for reference. The simulation environment was based on a
point mass which is affected by forces and momentums in such
a way that it executes oscillations comparable to the ones of a
ship at sea. To simulate the behavior of sensors, noise and biases
were added to the state outputs.
The second test is based on real world data which originate from
a boat trip at sea. Along with the data from a MS2 and a GPS,
attitude estimates from a high-grade Attitude and Heading Determination
System (AHDS) are available for reference.
In both tests, simulated GPS outages have been included to evaluate
the handling of this scenario.
In the test on the simulated data, the navigation solution was
able to correctly estimate all the sensor biases. The performance
during GPS coverage and outage were both acceptable. This led
to the conclusion that the basic concepts of the designed filter
In the test on the real world data, the designed navigation solution
did not have an acceptable performance even though it was
seemingly able to estimate the biases of the gyroscope. The estimates
of the magnetometer bias indicated that a non-included
effect were present in the magnetometer measurements. It was
suggested that a non-included lever between the GPS and MS2
and a ”soft iron” magnetometer effect could be the causes of the
bad test results.
Publication date17 Feb 2012
Number of pages110


ID: 61436292