9 Marzo, 1:22am

spreadsheets sucks!

One of my hypothesis of my master's work is that you can do sensorial fusion between several distance mesaurement sensors, specially sonar and laser. So I started a simple experiment: log the data in a typical travel of both sensors and graph their mesaurements in order to compare their diference.

I was expecting that the sonar beam match with the shortest laser beam in it's correspondient beams set.

I set up a player configuration logging all the gathered data. In 5 minutes of operation the log fetched 3.9M of raw data.

Well, I though, lets process this bunch of data with OO Calc and graph each instance of time. But I've to preprocess the data in a form that could be imported by Calc, like csv, and facilitate the charting operations. Perl was the path to follow.

The whole thing become slow, mouse tedious operations, and hard to alineate all the data. Waste almost all the day trying to achive that. Very frustrating without results.

Late at night, when I was in bed, a glorious idea came to me: let's use gnuplot!

In the next day I recode the pre-processing script and write a new one which open a pipe to the gnuplot process, send the charting data, et voila! and also it display the distance at every moment in time, resembling and animation. Just beautiful.

It was when I realize: Spreadsheets just sucks! they're slow, bloat, rigid... at least when you want deal with big chunks of raw data.

Well, the bad news is that my expectations weren't true always, there are many times where the beams don't match at all. Explanations: a) we're sensing two different planes (the laser is above the sonars); b) both sensors don't sense synchronously, the sonar is far more slow than laser; and c) a sonar is damage.

Now I've to report this and see the adviser's feedback.