POSITION AND ATTITUDE MEASURING SYSTEMS FOR AUTONOMOUS VEHICLES
- Introduction to robotic sensors
- Incremental sensors
- Environment referred localization
- Introduction to Sensor Fusion
SENSOR FUSION IN FREQUENCY
Some material that I used to give the lesson (part is in Italian, sorry in the Doebelin version in English the complementary filtering seems not to be present):
- Doebelin material
- Notes on the filter order needed to compensate for the different kind of drifts (constant, linear, parabolic, etc)
- How to choose the cutting frequency?
- Laplace transform relations
The Simulink file to simulate the Baro-inertial altimeter:
- Simulink File (this link was cancelled this year)
During the lesson I composed the following notes in English:
- Baro-inertial altimeter
- General considerations about a measurement system: wide band noise and its effects (the low frequency and high frequency components). How the integration operation change its behavior?
- Here a note on how the two sensors output are fed to the two complementary filters given their intrinsic noise behaviors
- How to choose the cutting frequency?
- Comparative table between the two sensors
- How to design the two complementary filters in Matlab
- How to model in time-frequency and compensate a constant Bias
- How to model in time-frequency and compensate a linear Drift
- How to reduce the transitory in complementary filtering: First Step (increase cutting frequency at the expense of high frequency noise introduction) and Second Step (increase filters order to increase the low-pass filtering efficiency)
How to use Simulink to simulate whichever differential equation, either linear or non-linear:
- Using integrators
- or Transfer Function Blocks
- Using derivative blocks (not to use)
- How to avoid derivative blocks (for example in higher order filters where you have high order input derivatives in the basic differential equations)
- Another way to avoid derivative blocks with Direct Form n1, pdf doc 1 and doc 2
Work in class to simulate Complementary filtering in Baro-inertial altitude estimation:
- Download this file that contains instructions that you shall follow step to step
Gyro-inclinometer fusion to estimate angle:
- Gyroscope + capacitive inclinometer (capacitances enclosed with a dielectric fluid): the gyro has the problem of a linear drift, the inclinometer slow response. So the two instruments are fused with 2° order complementary filters.
- In the above link there is also the method to implement in real time the designed complementary filters via recursive equations that can be implemented in a WHILE LOOP.
- Simulink scheme to run the simulation via a matlab script
SENSOR FUSION IN THE STATISTICAL DOMAIN
- Sensor fusion in Statistical Domain
—-> Example shown in class about Joint PDF, Marginal PDF, Conditional PDF - Uncertainty expression in the multivariate (vectorial) domain
- Bayes Theorem
- Exercitation on sensor fusion with multiple images
- Kalman Filtering
- SLAM
—-> TOF in Unity - Students’ project on SLAM