As mentioned in the Machine Vision project description, I am supposed to work on a project during the second semester at Jacobs University Master’s programme. (Yes. I am using British english. 😛 ) Unlike the Machine Vision project, where we all were working on a single project, each student got a seperate project for the Robotics Project course, and used the Robotics seminar, to present the results of the project.
My project is “Motion Compensation in single beamed Sonar”. SONAR stands for SOund Navigation And Ranging. It uses sound waves and time of flight to detect where the object is located. Since sound wave is much slower than light waves, it suffered from distortion when the sonar sensor is actually moving while taking the scan.
These can best be illustrated in the presentation here:
As the presentation shows, I did a lot of research into techniques for solving motion compensation. Some of these use a sonar array, and others use multi-beam sonars; both of which are quite expensive. The real challenge is to use a single beam sonar, and have the motion compensated there.
The basic idea is very simple. Take each sonar scan line (one sonar reading in one direction), and the associated pose, and use the pose to transform the sonar scan.
Next problem: how to get the pose in an underwater robot? There are a number of techniques to solve this problem. Right now, I am focusing on the most straight forward one, which is using the acceleration readings from the innertial measurement unit (IMU), and double integrate it to get the distance travelled in the different axis. So far, the results have been pretty good.
I then used these pose estimates to plot the sonar. The results so far are best shown in the figures below:
The figure on the left shows the distorted image, whereas the image on the right shows the corrected image.
**UPDATE 16th May, 2011**
I have managed to do a lot more work on this, especially since now the semester is ending, and the deadline for the submission of the report is here. Actually it was 13th May, 2011. Suprisingly, I was able to finish the report well ahead of time, which is nice achievement if you consider it is somehow around 9000 words :S.
Anyhow… here it is in case you are interested: Motion Compensation For SONAR