:: Video recorded during presentation using Julio camera ::

And to finish our blog publications... a short video with a different perspective. The video was recorded on Maria, using it's video capabilities, during the first flight of Julio.

Thank you everybody for welcoming this project and for the feedback given.

:: Video of the presentation ::

And here it is... the video of the presentation. The sound is not to good (and it's in Portuguese), but it was what we could record.

We take the opportunity to thank everybody who helped us on the project.

:: Report published ::

Here it is the final and complete version of the report.

Unfortunately we didn't had the time to translated to English, so the version here available is the original one in Portuguese. If you don't speak Portuguese, maybe Google will help on understanding it ;)

Link to final report

:: Presentation preparation ::

The expected moment is coming. It's today at 18h30.

We've been here all morning preparing the room:

• Drawing a 5x3meters rectangle as our "area" for Chasso;
• Testing the compass values of Chasso, so that we assemble the room with the least influences possible (turn out to be a maximum amount of 2 degrees - much better than any value we got anytime before);
• Testing the ambient light (to know how we can calibrate Maria and still have good results for facial recognition);
• Assemble a second projector, for closeups on the robots (since not everybody will be able to see them on the floor);
• Assemble everything else and to some dry runs.

At 17h00, with the heat, we started having problems. Until that time, everything was running smoothly and as planned. But, then, the USB hub inside Maria blew up, as well as the web cameras whom also stopped work. We had to rush thru the university to get 4 fans and try to cool everything up and substitute the hub.

We eventually manage to cool down everything, including the PC, which also, by then, was also having errors on the wi-fi board and because one of the web cameras was "dead", we had to recalibrate Maria and do the presentation with less sensitivity on Maria.

:: Global and final tests ::

Today, the day before the presentation to professors and fellow students, we are doing our final tests. It shouldn't be, but it is the first time we reunited all components on the same room, and use all together for the first time, as a global system. Unfortunately, because of the large scope of the project and the amount of time we had to develop it, only today we manage to test everything all at once. Everything went ok ;)

And the door with the "To do's" is clean! Everything done. Now, let's prepare the presentation.

:: Report delivered! Now, dry runs! ::

Today, we finished the final report and send it to evaluation. After the work is presented, we will publish it here.

After the report, we did our first assemble of everything has is going to be presented. The presentation will occur on the 28th of June, therefore tomorrow, will be spent preparing the presentation and do some dry runs.

:: The "Done" door is almost full and the "To do" almost empty ::)

:: Report, report, report ::

The last days are being rough. It's the final stretch. We have until the 25th to deliver the final report of the project, so, we've been working a lot of hours to complete it.

Today we had the visit of Prof. Tiago, who once more gave us some hours of his weekend to help us on the report. Yesterday, Prof. Pereira, also sent some information of what to change by email, so, the rest of the weekend, we will be working on it.

The deadline is near and now it's just, writing, revision, writing, revision...

:: Measuring results and compelling data ::

This weekend we did measured can compelled the values of all the results we identified during the last weeks as the main results we should present on the final report of our work.

It was a week and weekend all working on the final report.

We quantified some interesting results... the data confirmed some of the problems we already knew, like the Chasso compass problems near concrete walls, Julio GPS problems and mainly, face recognition problems in some angles and light conditions.

:: Thumbs up on Julio autonomous flight ::

Today we finished the tests (ok, maybe small adjustments can be made, but everything looks very good now) on Julio autonomous flight, according to the "locked" face. Julio can now respond very good to the movements of the face and follow it in a way to have it on center of the image.

Today we had our tutors at our "lab" and we had the opportunity to show all the functions implemented and discussed what we should adapt and do more. We came to the conclusion that because we can't have an acceptable accuracy on the GPS coordinates for Julio, on terms of implementation, we should wrap the project as it is. Everything is now at 95% (only some small adjustments and fine tuning it's needed), therefore, we should reserve the next two weeks to work on the report and on elaborating it's data

One public thank you to Prof. Melo and Prof. Tiago for taking part of their Saturday to come where we are working and help us on the project.

:: Bad news don't come only once ::

After some early morning tests with the Chasso robot to confirm that the Parallel search worked fine on a full room (as planned with 5 meters x 3 meters), we could verify that theoretically it worked fine, but the incorrect readings from compass, because of the concrete walls, just made us realize that the robot did some strange search patterns and not a parallel pattern. Nothing that we can avoid, at least with a compass not as accurate as this one from the Lego.

After this, we went to test the Ar Drone (aka Julio) on the exterior with the Android application for getting the GPS coordinates and send them over wi-fi to our console application, but we had several problems:

• With the weight of the mobile phone on top, the flight becomes very unstable, specially if the day is a little bit windy like today was and it's almost impossible to control the flight (at least in an autonomous way);
• Because of the high balance of the drone with the mobile phone, faces are hard to recognize on video (because of the maximum 30 degrees declination);
• The GPS of the phone we had available was very inaccurate. Readings with the phone placed on a fixed place, could give more than 7 meters differences. This just ruins our hope of having a way to know the location of Julio on space, therefore, all functions based on Julio needing to know it's location could go by the drain (returning home, quadratic search pattern, parallel search pattern and search based on waypoints). With only about 2 weeks to deliver the project, it should be difficult for us to get a solution for this problem :(

The rest of today, we spent implementing a new algorithm for autonomous flight of Julio, because of the changes we've been done to implement a better performance on Maria console application, the previous algorithm was not responding correctly. We have now a base that we can think will be final, but we still need to do more tests and do some fine tuning, because at this time, we only have the Yaw (rotation movement) and Pitch (back and forward movement)  working almost perfect. We still have to implement the Roll and Gaz (altitude).

:: Julio navigation functions are on track ::

With the GPS data recollected from the Android mobile phone on top of Julio, we are now able to get the location of Julio in exterior locations.

On the last days we've been working on getting the readings of the GPS and implementing all the functionalities on code, but, without having a proper space to test, until now, everything is theoretical. Tomorrow, because it's a national holiday, we will get together to test all this new code and see what adaptations may be needed.

Let's hope everything comes together and until next Sunday we will finish all Julio functions, otherwise, we will be behind our schedule and with only three weeks due to the deliver the project, we will be facing a problem.

:: Major milestone achieved ::

This weekend revealed itself as very good weekend in terms of closed items. It was our milestone to close all the operations/functionalities of the Chasso robot and also of Maria. We were having some problems during the last two weeks, but everything came together when we most needed it to be. We came just in time to close this milestone on the planned date.

Maria

• We finally reunited all the conditions required to test the console App on Maria, using the I7 PC. In terms of performance we got a small improvement just because of the new processor (the other was an I5), but after some changes on the code to use more cores and taking advantage of the multithreading capabilities of the machine, and also some changes on the Chasso timers to read the sensors (because they are slow) we manage to get the performance to the level we wanted initially.

• Having this problem solved, we then tested the usability of the application on Maria, because of the touchable surface. We encountered some problems on it's sensitivity, so we went to do some changes again on Maria. This time we added another camera, taking advantage of the new NUI Group (Natural User Interface Group) freeware driver's 1.5 version, that adds this multi camera functionality. Problem solved! With each camera covering half of the surface, we eliminated the light reflex and also increased the sensitive. We now have what we think is Maria 99% done! Everything is working smoothly and has expected on an amateur surface table, mounted with only this resources. There are small issues, like the need of calibration according to the light on the room, but nothing that can't be done every time we need to use it.

Chasso

• The quadratic spiral search, the parallel search and the track points navigation functions are now made and tested. We could say that everything went almost very well, if it wasn't for the differences on the compass readings near walls and metal sources. This misreadings make the odometer to not work very well, but we can say the results are very good, for a proof of concept.
• We just add one more little lego piece to the robot. With this piece on the back of the camera battery we managed to get the robot more stable when turning.
• Also final touches were made to the record video functionality. It's now fully working.

Julio

• The C# class to receive the GPS coordinates from the Android mobile phone to be mounted on top of Julio is now done and tested.

:: Big weekend is near ::

Next weekend we're going to do the tests on the rest of the Chasso functions developed during this week: quadratic search, paralel search and track points navigation. It's going to be hard work.

If everything goes well, next week we will implement the "same" functions on Julio, this time based on GPS coordinates and not on on an estimate position.

Fingers crossed :)