Using the Accelerometer

So I can connect the test vehicle and its accelerometer to the PC, now to do something with the data. Accelerometers are devices for measuring the strength of the Earth’s gravitational field in different axes, which means that with some arithmetic you can extract the orientation on the accelerometer and whatever it is attached to. Ever wondered how an iPhone knows when it is turned on its side, well it uses an accelerometer to sense its orientation. I am using the LIS302DL triple-axis accelerometer; there is a tutorial on how to use it here. 

Measuring tilt angle
Taken from: Tilt Sensing Using Linear Accelerometers by: Kimberly Tuck

"Measuring Tilt using a Three Axis Solution
In order to define the angles of the accelerometer in three dimensions the pitch, roll and theta are sensed using all three outputs of the accelerometer. Pitch (?) is defined as the angle of the X-axis relative to ground. Roll (f) is defined as the angle of the Y-axis relative to the ground. Theta (?) is the angle of the Z axis relative to gravity."

Once the equations were established for calculating the orientation I could extract the angles and display the result. I wanted to show the orientation being tracked in real time with a simple intuitive animation. I chose a basic artificial horizon type animated display; it’s simple and clear to understand.

The raw measurements from the accelerometer were already being inputted into MATLAB; all that was needed was some more code to initialise the figure and to update the figure each time new data was received. I took the roll angle value and used that to alter the “angle of bank” on the figure. The pitch angle is also available but that visualisation will be included at a later point.

 Video

All the movements I am making need to be reasonably slow, this is because an accelerometer can also sense dynamic acceleration. Image that an accelerometer is held perfectly flat, but it is accelerating in a straight line, the accelerometer will detect the dynamic acceleration and will give a false reading of its orientation. At this point you need to bring in other sensors taking other measurements and fuse the data together to get a more accurate picture.

Usually gyros are used with accelerometers; gyros are devices which measure the angular velocity about an axis. The gyro will read non zero during a rotation and will be zero once it stops rotating, you can then obtain a value for the angle by integrating the angular velocity over time. However gyros have “drift” which is that they are constantly accumulating errors, so eventually you have no idea which way you are pointing. This is where you use the accelerometer to correct the gyro drift, but we know that the accelerometer can be corrupted by sensing dynamic acceleration too…. So for that reason the average value from the accelerometer over a period of time is used to correct the gyros.

That is one method to fuse sensors others include using a Kalman filter which estimates the state of a system from noisy/inaccurate measurements. It’s a recursive filter using statistical modelling to remove the noise from the signal/output.  The Kalman filter is not a light topic; it took a good couple of weeks and a lot of reading to get to a point where I had a very general understanding of how and why it worked. But once you reach a certain level the actual algorithms you implement are not too difficult. If you are interested in the filter for your own work then here are some documents that I found useful during my research.

• An excellent introduction to the filter. Including MATLAB code for you to observe for yourself how the filter operates.
• Its Wikipedia entry
• An academic paper covering some of the mathematics behind the filter.

What I leaned about Kalman filtering so far should be useful, because the work from ATB1 will lead onto experimentation with stability and control.

You can see that this can be a very complex topic, and we are beginning to venture into the world of Autopilots and inertial measurement systems (IMS). Again if you would like to know more about autopilots or gyros and accelerometers then DIYdrones.com is an excellent place to do so. As well as general browsing of the site, here are some specific links to discussions on this topic.

• Why  we need accelerometers and gyros.
• A more technical  expansion on "Why we need accelerometers and gyros."
• Kalman filter for idiots (not my words, but I still think it isn’t exactly crystal clear!)
• Another method to combine multiple sensor readings.
• What getting it right looks like. Notice how it's not affected by dynamic acceleration.