Log Analysis Tool

Good Day Everybody,

I’m currently developing with a partner an open source project with the university. We are working on a tool that does analysis and diagnosis of flight logs for Copters and Planes. We were working in the analysis and graphing of the log for a single flight as well as the analysis of the logs from several flights.

We would like to ask you as final users for help in what would you like to see in a tool like this. If you could help us with these questions and any ideas you can give us we would really appreciate your help.



Some Questions

  • Besides Mission Planner, do you know or use any other tool for log analysis?

  • Do you like the Mission Planner Log Analysis tool and why?
    What would you change, what would you keep?

  • What is the data that you find most useful reviewing after a normal flight?

  • What is the data that you find most useful reviewing after an unsatisfactory performance during flight?

  • We are working on a very visual tool. What graphs would you like to see by default when opening the tool?

  • Do you do any kind of work with the data table?

  • What are the main differences you see in the log analysis of copters vs planes?


In the part of analysing several log flights is where we are most lost, we have very little experience and we are not sure of what kind of data would be most useful to you.

  • Do you do any kind of maintenance based on le analysis of several logs?
    If you do, what do you look for in the logs?

  • What kind of analysis do you think could be useful in when looking at the data from various logs?

• APM Planner has very good graphing tools and also runs on Mac.
• Refer above, check out it’s data handling, filtering and double click options.
• Typically ATT and IMU as quick indicators of a)tune and b)hardware issues.
• We would look at everything after an unsuccessful flight.
• No graphs by default as each use is different.
• The data table is useful for filtering, checking individual readings and the like.
• In copters, vibration analysis is very important as it can indicate arising problems.

Besides Mission Planner, do you know or use any other tool for log
analysis?

Automated analysis using LogAnalyzer.py
(https://github.com/ardupilot/ardupilot/tree/master/Tools/LogAnalyzer)

MAVExplorer

dronekit-la

Hi @mboland, thank you for your reply and your very useful tips.

We will check out APM Planner as well to build up some more background.

We are going to work on a mostly visual tool, working with the D3 libraries to make those graphs as dynamic and interconnected as possible. If you could see the particular data of a point as you hover in the graphs, will you also have a need for the table?
We also wanted to incorporate zooming and panning in the graphs.

We were going to work on top a template similar to this LimitlessDemo.

We also already considered including some degree of vibration analysis as the feedback we have points strongly in that direction.

Thanks for your help. We’ll keep you updated with the progress.

Thank you @peterbarker for your help. If you could offer any more suggestions we would really appreciate them.

Great to hear you’re interested in this! Here’s some of my opinions:

  1. I want to do math on the logs in MATLAB, so I want some way to import the data there, and have it appear in an easily useable form.
    1A) I don’t run Windows, so using Mission Planner (Windows-only) to convert the logs from .bin to either .log or .mat files is an extra step.
    1B) Mission Planner doesn’t handle char-array (string) data from the logs correctly in MATLAB, it’s a known issue.
    1C) I have not investigated if other log analyzers can convert to MATLAB. I’ve heard APM Planner is perhaps better than Mission Planner? No experience.

  2. When opening a new log, I find the Lat-Lng overlay on a google-map useful for identifying the log, and the general flight mission (“Oh, that’s the one I flew big circles”). I usually identify the time-of-takeoff, number of flights on the log, and flight duration from the altitude plot, before looking at other things. (This is to figure out which parts of the log I care about, see next comment.)

  3. I want a useful system to select parts of my log (in time) for viewing. At a basic level, this can trim off log parts that are just sitting on the ground, or taxi-ing on the runway. At a more advanced level, maybe I can trim to “show parts of the log where the plane was gently maneuvering”

For what it’s worth, G. Zogopoulos Papaliakos and I are co-developing a MATLAB-based log importer. We plan to add some visualization capabilities, and then separately develop our MATLAB-based math tools to do the analyses that we like. (I’m interested in weather, he is looking at dynamics and control issues) While the code works today, we haven’t announced it for use… importing a .bin file is too slow at present. It’s a side-project for us, so we work on it as we find time, and aren’t promising any deadlines.

1 Like

Hi @hunt0r, sorry for the late reply. Thank you for all your insight.

  1. We want to make the application cross-platform, preferably web based, so it can be available on any device anywhere, well thats the final idea. We are currently working on the conversion from bin to log (we will also try to add the matlab conversion) and also trying out the current tools available for log analysis.

If you don’t mind my asking, what kind of math do you do on MATLAB with the log, could you give us a little insight into that?

  1. We also thought about adding a map to the tool, so that you can see the graphs with your flight next to each other, we are still thinking on what kind of map to add.

  2. We are going to try to make those graphs as dynamic and interactive as possible, when you say trim do you mean just make a zoom or something else? This kind of filtering is something we are really interested in.

Again thanks for you insight on the subject, we really appreciate the

For example: I’m going to use flight logs from the Pixhawk as system ID data for aerodynamic models. I’m also interested in wind estimation algorithms, specifically by observing the “unmeasured state” of wind velocity during the dynamic response.

I mean selecting subsets of the log for plotting, and/or analysis. Because I often have a 30min log surrounding 5min of the flight experiment data I actually want.

Hi piptin, good to hear that you are developing a new tool. What’s the open source project? May I join?