5/10/10- Well, the semester has come to an end. We were able to complete the system and demostrate significantly improved position and velocity estimates by using the Unscented Kalman Filter. In addition, in the case of a GPS loss, the system is able to function for an extended period of time when compared to the unfiltered system.
We are currently finishing up our final report, which will be posted as soon as it is completed. We recently submitted a brief paper to the EIT Student Paper conference, to be held on 5/21/10. This paper has been posted in the Delieverables section. In addition, an updated version of our project poster has been uploaded.
4/20/10- The semester is almost over!
Keeping this site updated has proved to be quite a challenge.
On 4/9, we presented our project at the Bradley University Student
Expo. SINAPS earned second place in the engineering and
sciences division! The poster used in the presentation has
been placed in the deliverables section.
Both Dan and Luke have been working on completing the system.
Today, the final major piece was completed: the Unscented
Kalman Filter. We have our first set of actual results and
they look very promising. Results will be posted soon!
3/4/10- Dan is nearing completion
of the strapdown solution. For months the strapdown solution has
either been mechanized in a global frame such as ECEF or geodetic
(latitude, longitude, altitude), which makes sense for the project
because the IMU navigation data will have to be directly compared to
the GPS information. However, mechanizing in a global frame is
difficult because the attitude and force resolution are unintuitive
and complex, making errors tough to spot. An alternate mechanization
that is more natural to think about is a local plane tangent to the
navigator’s absolute initial position on the earth. This
mechanization is called either ENU (for east, north, up) or NED
(north, east, down), and they only differ by a series of rotations.
ENU will be used for this project. The only drawback of the local
tangent plane frame is that it is only an approximation for a
limited portion of the Earth’s surface, making it only good for
short term navigation.
Now that both IMU and GPS data can be taken simultaneously and the
reference frame issue has been resolved the strapdown computations
can be compared to the GPS. With an artificial initial attitude
inserted (to best match the trajectory of the GPS) and the means
subtracted from the IMU data, the strapdown solution closely matches
the GPS data for only a few seconds. After thirty seconds the
position solution has drifted thirty meters away.
Meanwhile, Luke has been focusing primarily on the sensor modeling
side. There were a few problems with the sensor error model
that had to be corrected. Luke has created a modeling
procedure to determine the autocorrelation time constant of each
sensor and a first order Gauss-Markov model for each sensor has been
developed. He has also completed the standard Kalman filter
and will be testing its performance in the near future.
2/21/10-
It's been a while since our last update. Things are progressing very well.
There was an issue with how we were getting the IMU data into Matlab- the system overhead
was causing non-uniform sampling. The sampling period would swing almost 50% depending on
the sampling frequency we chose. We spoke with VectorNav and found a solution to this problem,
and IMU data collection is now completed.
GPS data collection is almost complete as well- all that remains is to convert the hexadecimal values
from the UBX messages into the correct datatypes for Matlab.
With the IMU data coming in correctly, we have been able to continue work on
modeling the sensor error states.
A model for the X and Y axis gyroscopes has been developed and the rest of the sensors will soon be completed.
On the strapdown side of things, Dan has completed prototype versions of the navigation and attitude initialization algorithims.
1/21/10-
First day back after break. During the break, Dan worked on the Strapdown
system and Luke worked on Kalman filtering and taking IMU data sets. Also, we
chose and purchased a GPS unit- the uBlox5 EVK-5T evaluation kit. This unit
will give us access to the GPS pseudoranges, which will be necessary to implement
the tightly-coupled system.
Today Luke has been examining IMU data to model the error states for the IMU.
Dan spent the day continuing his work on the strapdown system.
12/07/09- The IMU has come in, and I (Luke) began working with it over the weekend. I established two-way communication within Matlab and am now familiar with the Sensor Explorer software.
Dan and I are finalizing the project proposal presentation, which will be given tomorrow. Afterwards, it will be uploaded to the site.
11/19/09- We now have a two-variable Kalman filter working in Matlab and are now working on converting into a function, which will greatly simplify our coding.
Also, the Functional Requirements document has been completed and is now posted in the Deliverables section.
11/12/09- It's been a slow few weeks. We ran into
a few complications with our choice of sensors, but those have been sorted out now.
    This project will be centered around the use of the Kalman filter- as such, Dan and I
have been becoming familiar with dynamic system theory, state space representations, and
the derivation of the Kalman filter. Dr. Ahn has been giving short lectures on the subjects
and we have been using a few textbooks as well. Today marked our first working filter- a simple,
one variable Kalman filter based on problem 5.1 from Optimal State Estimation by Simon.
10/22/09- Spent the day searching for an IMU for the project. We have found several viable canidates and will be presenting our findings to Dr. Ahn and Dr. Lu tomorrow afternoon. We hope to have both the IMU and GPS unit in hand by next week.
10/20/09 - Webpage completed. Functional Description and High-Level Block Diagram posted in Deliverables.