Differences

This shows you the differences between two versions of the page.

--- projects:mihab:glider [2007/03/24 16:51] – laurenceb
+++ projects:mihab:glider [2008/07/19 23:33] (current) – external edit 127.0.0.1
@@ Line 3: / Line 3: @@
  Following MiHAB 1 and 2, I've decided to try something new, hence the glider project. This project may merge with the
 UKHAS glider project, and I'm designing it to be applicable to other gliders.
+===== Mark #2 Glider ====
+With the glider used in flight 1 completely trashed, something new will be needed ;-) Having considered a 2 axis stabilised rigid glider (probably using thermopiles and an ARTF off ebay) and also a parafoil, I'll probably go for a semirigid rogallo. These are very stable and forgiving on control inputs, and also easy to make, and foldable for transport to the launch site. Parafoils also meet these criteria, but readily avaliable parafoil kites are harder to adjust and test unless you're good at sewing and have a nice high platform to test from.
+ Judging from the previous test flights, a "PIDR" loop with SIRF2 would almost certainly be adequate for a rogallo, but I'll try and be on the safe side (money and time allowing) and go for a rate gyro ( off spark fun) and kalman loop. This should be able to give an extremely accurate and responsive heading value for the PID loop. Temperature compensation of the gyro will be essential, but a peltier cooler system has been constructed for that purpose, and should be capable of -30 centigrade (-14 reached so far). One problem will be small steps in the Kalman filter output with each new GPS heading, which would mess up the D term in the PID controller unless it is turned off each time this happens. (i.e. turned off for one iteration of the Kalman filter)
+ A second, and more serious problem is the GPS lag, approx 1.1 seconds for a sirf2. Solving this may mean that the correct error treatment cannot be used in the kalman filter, but more work is required. Running the filter back and forwards to pick up each GPS heading is not an option as there will be insufficient recources (ie SRAM and clock cycles).
+ Here is an overview of the planned system, it would appear that there is a lot to do, but much of the work has already been done for flight #1, so it may not be as hard as it appears :-)
+{{projects:mihab:evil_plan_to_take_over_the_world.png|}}
+===== Rogallo experiments =====
+This RC rogallo is approx 1 meter long. The pitch control is not being used at present, but the "payload" is currently being rebuilt to include pitch control as well. The vertical and horizontal cross spars were added to improve the stability and remove the possible problem of the wing folding up or diving. It now seems extremely stable, and should fly in virtually any weather conditions.
+ {{projects:mihab:p4130055.jpg?700|}}
+It seems a rogallo can be made to descend vertically by moving the c of g well towards the back. This feature would be brilliant for landing. A small micro servo with pot removed for continuous rotation could be used to adjust the guy lines when the rogallo reaches 100 meters or so.
+[[http://video.google.co.uk/videoplay?docid=-247659611589004030|Demonstration (unfortunatly a bit dark)]]
+===== Flight 1 =====
+ This appears to have gone into a spiral dive soon after release, descending at 15m/s into an industrial estate at the edge of St-Neots, it was found the next day and will be recovered shortly. [[Launch photos (at EARS)]]
+== Lessons for future ==
+  * Although the glider appreared quite stable in tests, a drop from a balloon is more violent, and there is a greater probability of the glider spiral diving.
+  * 9V lithium batteries cannot supply more than 100ma on a continuous basis
+  * Dividing up power supplies increases reliability i.e. flight computer and gps off a seperate battery from servo(s) and cutdown
+  * Only fly with 100% reliable code
+  * Use a high quality compiler
+  * Unless the aerodynamics is thoroughly tested, rigid gliders are only reliable with thermopile stabilization. Rogallos and parafoils are c of g stabilized, so more stable in that respect
+  * A wing trailing edge dipole antenna works very well
+  * Tethered drops are useful as an intermediate stage between "hill" tests and drop tests.
+  * It's best to test in the calmest possible weather
 ===== Simulation =====
- Using just a basic approximation for the yaw behaviour of the glider, the oscillatory behaviour can be analised. It would appear that approximating the current GPS heading from the current one and the previous makes the system more responsive and less prone to oscillation. A and B show the heading offset and servo position using the approximation technique, whilst C and D show the same data for uncorrected PID loop using the same constants (P=0.15,I=0.03,D=0.25) in E and F the value of D has been adjusted so that the overshoot is the same as in the corrected case, and it can be seen that for the same P and I, ie the same resistance to trim errors(I) and tendancy to head towards the target(P) we have far superior behaviour. Next step will be automated PID optimisation! Source code is [[here]]
+Using just a basic approximation for the yaw behaviour of the glider, the oscillatory behaviour can be analised.
+ A more advanced extropolation technique has now been tried out, taking the aerodynamic damping effects and current rudder position into consideration. It can be seen that the results are very encouraging. The +-15 degree range is reached within 2 seconds! and the glider stays there. This is very important as turbulence occurs on the 10 second timeframe, so the glider can recover from one disturbance in time for the next.
+{{projects:mihab:forward_extropolation.png|}}
+ It would appear that approximating the current GPS heading from the current one and the previous makes the system more responsive and less prone to oscillation. A and B show the heading offset and servo position using the approximation technique, whilst C and D show the same data for uncorrected PID loop using the same constants (P=0.15,I=0.03,D=0.25) in E and F the value of D has been adjusted so that the overshoot is the same as in the corrected case, and it can be seen that for the same P and I, ie the same resistance to trim errors(I) and tendancy to head towards the target(P) we have far superior behaviour. Next step will be automated PID optimisation! [[Source code]]
 {{projects:mihab:simulation.png|}}