Brainstorming Copter Launcher for a High Speed Plane

[Michail_Belov]

I am thinking of building a copter style launcher for a high speed plane. The plane weighs around 5.5 .. 6.6 kg depending on battery setup and has a stall speed of about 70 km/h and has a sluggish low speed acceleration due to very high pitched prop (it is version 2, the first one was succesfully flown, but after a mishap when launching from car roof, fell off, stalled, hit ground, caught fire and burnt).

A reasonable copter set-up with Lipo battery would weigh just below 3 kg

Initially, I thought about just hanging the plane from the tail, bring it to about 200 meters and release it. It would work, the battery has about 3 minutes of hover time. But it is very tight, both for battery and for the recovery from a dive.

The other idea is the following, very similar to the QUAD VTOL, and the copter would be attached to the plane (grabbing it, basically having one release pin and 4 legs pushing against the plane wing:

The copter would start in qhover mode. It would lift off, and a strong climb initiated immediately, at maybe 2 m/s. At the same time, the plane would start its motor and accelerate to a speed just below stall speed. In fact, since the total weight would be around 10 kg, the stall speed would be around 85..90 km/h. This should reduce the copter power consumption by about 80 %.

I am not sure whether my idea of keeping the flying speed below stall speed is important or not. I am worried that if allowed to accelerate beyond the stall speed, the plane would enter positive climb, that would reduce to a minimum the copter motors, which could make the whole system unstable, and also make the copter unstable after the release. One possible workaround could be setting MOT_SPIN_MIN to a very high value. Another idea would be to fly the copter in STABILIZED mode, but the transition from QHOVER to STABILIZED at some point could be delicate. Also, flying beyond stall speed would make the plane controls efective, and they could fight with the copter. In any case, during this phase, all controls would be in neutral, so most probably the plane would be in MANUAL mode.

Yet another possibility would be to setup the copter as VTOL QUAD with Q_ASSIST. Here I would love to get some opinions.

In this flight mode just below stall the plane with the copter could climb easily and efficiently to a safe altitude. In fact, the normal, full power hover time would be about 15 seconds, 2-3 seconds to get off the ground, and about 7..8 seconds to accelerate to 80 km/h.

For the release, the release pin would be pulled by a servo, and a strong downward elevator push would initiate a short shallow dive to accelerate safely beyond stall speed (may be 50 meters of altitude loss), with the conbtrols given to FBWA after maybe 3 seconds. At the same time, the copter PID parameters would be changed by Lua script for 3 kg flying mode, possibly a switch to STABILIZED mode and a short burst of full power (1 second?) applied to clear safely the plane. The copter would then enter AUTO mode or RTL mode and return and land automatically at the home position.

I would love to get some input as to how setup the system, which modes to use, whether to use VTOL QUAD or stick with pure copter modes, and which one would be the best, Qhover, or Stabilize (in that latter case, the throttle may be controlled by a lua script).

[MaxBuzz (Maxim Buzdalov)]

Some time ago I thought of a somewhat similar system, which didn’t get far because my experience with planes was exactly zero in the beginning, and I still don’t have anything resembling a test plane. My launcher would be a coax copter arranged as a tube with propellers and few passive aerodynamic surfaces that would assist with descent.

Some of the concerns I had, which are relevant to your case, is that the moment you detach from a plane, if it is not yet completely controlled by itself, can be very tricky, as you can impose a rotational moment on a stalling plane that would make its situation even worse.

Ideally, I think, you would accelerate the paired vehicle to a point of barely having lift > plane mass from the plane, and disconnect at this moment - which almost forces the launcher to be below the plane. The launcher (from the viewpoint of Copter firmware) would be flying nearly horizontally before this moment, which, in a sense, makes altitude holding considerations almost irrelevant, as its whole throttle and attitude control are fighting air resistance. (I remember AUTO commands of keeping a fixed attitude and throttle, which is probably how you command this part of the trajectory). Once detached, the launcher drops the throttle (nicer to the battery too), falls in an equivalent of airmode, and then initiates RTLS.

But take this with a grain of salt.

[Michail_Belov]

Very interesting.

As far as the release of a plane below stall speed goes, it is very safe to do if you have the elevator pointing the plane nose downwards. Again, a lua script here could help. One could even do a script which would keep a certain g force by applying down elevator. For instance, at 0.5 g, the stall speed would be 0.75 of the stall speed during the straight flight. So, a g force between 0 (parabolic weightless flight) and 0.5 for 2 seconds to gain speed would be very safe.

The simpler option is just to have elevator pointing downwards, then in lua script wait until th e plane points some 20 degrees downwards, bring elevator to neutral, wait for 1 or 2 seconds, and then switch to FBWA mode which would pull the plane out of the shallow dive automatically.

There is one detail about the motor, it is very heavy (600 grams) and with a lot of torque and power (5 kW) so it should be operated carefully, without strong accelerations, especially during this flight stage at stall speed.

Launcher below the plane is very complicated due to props… But it is logical, especially if you want to detach at speeds higher than stall.

[LupusTheCanine]

You definitely don’t want to release under stall speed.

I would consider mounting the copter at an angle so plane wings start producing lift sooner, this way you will be able to accelerate more aggressively.

@MaxBuzz while falling off would be cleaner in terms of aerodynamic interference, punch out gives much larger margins and shouldn’t be an issue given rather high thrust of the copter. Given sudden decrease in hover
throttle I would expect unintended “punch out” when switching to altitude control modes. It will also avoid falling towards the ground with minimal attitude control.

If you want a fun idea you could consider making it a tailsitter

[Michail_Belov]

About the stall speed… I still insist that if you have down elevator it is OK. One might want to make the plane noise heavy just in case, but for ahigh speed plane that is not a problem (being noise heavy). A few days ago I had the issue with the normal VTOL where it did not want to do the forward transition due to ESC setting for battery protection. And I was hanging very very high, without the option to descend on VTOL however power (it was 500 metres and close to zero battery). In the end, I just switched to manual mode at zero speed, the plane dropped about 60 meters and I recovered without problems. Obviously, the stall speed of this one is 45 km/h but the principle is the same.

Mounting at an angle, that is actually very good idea and I thought about it. MH20 profile has best lift angle of about 10 degrees.

A tailsitter for the launcher is not a good idea at all, A tailsitter for high speed plane is a compromise. The plane I had reached a speed of 330 km/h before it burnt, but I was still setting it up and it was flying at 50 % of maximum motor power. I was using an 11 x 14 prop with 10S and I was aiming to go to 9 x 20 prop at 12S. The expected speed would be above 400 km/h and an endurance of about 20 minutes. It was extreme fun to fly it. And no tailsitter could do that, i.e. take off with such props.

[Michail_Belov]

The point which worries me is this: if I accelerate the plane-copter to a relatively high speed, the wing would be generating all the lift and the copter would either not work at all (close to zero power) or try to stabilize the plane in roll and pitch.

Ther are two questions here. Going from thrust of 10 kg to thrust close to zero and still maintaining roll and pitch control presents any issues or not.

The second issue is the release. Again, we have a copter at zero power, and suddenly the plane is released so the copter goes from zero thrust requirement to 3 kg thrust. Not sure if such a transition is a challenge or not… I know that you can throw a small 5 inch dron e by hand into the air and it will fly off, but I am not sure if it work the same way for a copter flying horizontally at 100 km/h.

That is the main reason why I think about limiting the speed of the plane-copter to below stall speed to generate partial lift from copter motors so as to make the transition after release much more limited

I am not sure if a arducopter is stable enough to go from zero power

[LupusTheCanine]

You recovered in manual mode, Ardupilot itself is very bad at stall recovery so unless you want to implement stall recovery yourself I would detach above stall speed.

I was thinking about using a copter configured as a motors only tailsitter plane as “tow” vehicle.

[Michail_Belov]

The release sequence would be done under lua script… both for copter and plane. Once arducopter made a gentle foamy spin… if I did not take control, it would have crashed.

[Michail_Belov]

Was thinking more about it, and I feel that the following setup will work.

Release below stall speed. Release with one pin at the forward central location and two TE wing retainer stops. This would guarantee that the nose would fall down first provided we are below the stall speed. Keeping the speed below stall speed is slightly tricky, because during the climb with copter, the total weight would be 10 kg, and the stall speed would be higher for that setup. So just before release, the speed should be really tightly controlled to be just substall for the plane alone.

PIDS.

For the copter, I think one needs to get three sets of PIDs (three tunes).

Obviously, the no load PID after release, just the copter alone.

The full load PIDS, for static hover with plane

And partial load PIDs for hover with forward motion with unloaded copter. For this I would make a mock-up, if the plane weighs 7 kg, I would make a 2 kg. frame with weights at the extremes (nose, tail, wings) much further out than the real plane, so as to simulate rotational inertia, but keeping the weight down. So if I have a motor-prop-esc which weighs 800 grams and is 40 cm. in front of CG, I would put a 120 cm stick with 270 grams in front of CG for tuning. I am not sure whether this is strictly necessary, i.e. whether the PIDs will really diverge that much due to changing weight, but it definitively would not be damaging.

Once I have these PIDs, a lua script in copter will take the speed (probably just the GPS speed) as the basis for calculating the weight/hover thrust which is reduced to wing lift, and adjust proportionally the PIDs. Alternatively, the lua could use the current-voltage, i.e. power at the props to estimate the necessary PIDs, and this could be actually even simpler. There is a catch, if the the copter throttle is set for climb, it would be difficult to filter out hover power from climb power.

[Michail_Belov]

And another option for the release would be just before the release to put the copter briefly into guided mode, incline it forward to a certain angle, around 7..10 degrees forward, so as to set a very low angle of attack. In that release scenario one could be above stall speed and still safely release the plane so that gravitation will pull it down.

[ddegn (Duane W. Degn)]

I don’t have any suggestions myself but I’d sure like to see video of this project. I hope you document your efforts and share them with us.

Good luck.

[Michail_Belov]

Well, I have been dabling with this project during the last few days.

First, there is the issue of suspending correctly and releasing the plane. I made a release mechanism and ran some tests on it.

I made a release pin, stainless steel 4.7 mm which slides in three mild steel supports with a diameter of 5.0 mm. I suspended 8 kg and measured the force necessary to pull the pin out, it was 2.2 kg (22 N).

Next I built the release mechanism:

It is missing a compression spring which I ordered (it would be pushing against the right-most support and a stop to be placed on the pin itself. I expect to adjust its length so that in compressed state it would exert about 50 N (5 kg.) From the previous test I know that to push the pin out a force of 22 N for a plane weight of 8 kg. is sufficient, so that 50 N spring is more than double the force.

Next I ran another test to see how much force is needed to move the latch up to release the pin under tension. I again suspended 8 kg from the pin (that is equivalent to 80 N of spring force; I plan to use only 50 N spring), and tested the force necessary to fire the release. For the position as shown in the previous photo, about 2 kg were necessary.

However, in the position shown below, only 1 kg was necessary.

In next photo shows the position where the pin stays almost on verge of firing, where even the slightest force is enough to trigger (and vibration probably would be sufficient).

So the total movement for firing the pin is around 10 mm, and the maximum force is around 2 kg., which can be adjusted to be even lower, around 1 kg. In any case, this is way below the force exerrted by a stabdard servo, and most probably I could get away with using a small 12 gr. servo.

From the tests I concluded that two features are important: to have a safety latch which would prevent accidental firing (very unlikely) and a pull linkage which would not be rigid, because once the pin begins firing, it does so very rapidly and fixed linkage would probably damage the gears fo the servo. So either a string or an articulated link would be used for moving the latch.

The safety latch could simply be a piece of circular horn with a cut-out. When armed, the circle would prevent the movement of the main latch. When command is given, the servo would turn and would move the cut-out in front of the latch so that latch could move freely. At the same time the string or articulated linkage would begin to pull the latch out.

[Michail_Belov]

Second thing is how to safely carry and release the plane. I have MH 20 airfoil which has the following Cl for different angle of attack:

So the first thing one can see is that the maximum Cl is at around 10 degrees. That means that the plane should probably be suspended under the copter with the wing angled at about 10 degrees to maximize lift.

Next thing, is the release itself. I decided that it should be done with lift considerably below the weight of the plane to let the plane safely fall under the copter. That can be achieved either through a slow speed, so that even at maximum lift angle the lift would be below the weight of the plane, or through lowering angle of attack, basically inclining copter/plane forward, which would reduce the angle of attack, and Cl as a consequence, and total lift.

For a 25 dm2 wing and 7 kg, the stall speed is around 76 km/h at sea level.

However, at speed of 100 km/h the total lift would be 12 kg, i.e. 2 kg beyond what is needed to keep the copter/plane tandem in the air. If using the Qhover mode for the copter that would most probably result in very destabilized situation.

The conclusion I would draw from this fact is that either the air speed must be strictly kept below the stall speed to have about 3 kg of downforce, or that a stabilize or acro mode for copter would have to be used where the thrust can basically be set, or that the angle of attack for suspended plane should be lower, for instance, with a Cl of 0.6, at 100 km/h the lift would be around 7.2 kg.

Cl of 0.6 corresponds to angle of attack of 4 degrees. However, if we have this angle of attack, at lower speeds during the initial acceleration and climb of the copter/plane tandem, the wing lift would contribute much less, only 4.2 kg…

From all this I conclude the following and the different stages of the flight I would consider:

Wing inclined at 10 degrees with respect to copter.

1. Take-off and initial climb, yaw pointing, initial acceleration up to around 70 km/h: QHOVER MODE
2. Further acceleration and Climb out: either ACRO MODE or GUIDED MODE. The moment when the switch from stage 1 to stage 2 will ocurr will be when the power on copter would be roughly equivalent to 3 kg of thrust, i.e. when the wing will carry the full weight of the plane, but not the copter. From that moment the copter must keep that same power fixed, meaning something like ACRO MODE with throttle set to something which will deliver the same power necessary for 3 kg thrust. The angle of the copter will be kept at 0, so that an excess of speed (extra lift) would simply go to faster climb.
3. Release. The copter would rotate very slowly forward until the vertical speed will become zero, i.e. when the angle of attack of the wing would generate lift equivalent to the plane weight. Once this is reached, the copter would rotate another 2..4 degrees forward which will significantly reduce the lift, and at that moment the plane could be released safely.

It is clear that the forward speed would be controlled by the plane motor and should be kept slightly above stall speed, but not very excessive either, so around 80.. 100 km/h is probably a good start. Higher speeds would have a much narrower angle margins and also could result in extreme forces should something go wrong…

[LupusTheCanine]

1. Guided supports using raw throttle with attitude setpoint.
2. You want way more than 10° as the copter will be pitched forward during acceleration.
3. As long as the copter is adding forward thrust at the time of release it will separate cleanly (if it is in a puller configuration), turbulence aside. If you did a pusher you could first release the latch and immediately after command low thrust for clean aerodynamic separation.
4. I strongly believe the release should be done well above stall speed to handle copter wake.6
5. Big thing will be making sure that copter FC and plane FC don’t fight each other.

[MaxBuzz (Maxim Buzdalov)]

I would advise considering a scenario of aborted separation, and fly the non-separated configuration first, so that you have real-world points of speed, AoA, time in flight and actually achieved lift.

[Michail_Belov]

That there is guided mode with raw throttle is exactly what I need most for the project. So that is very good news.

As far as second point is concerned, I am thinking of using only the plane motor for forward flight, i.e. in the qhover mode, the copter inclination would stay 0 degrees both for pitch and roll.

However, I agree that there will be a problem for assessing real angle of atack during climb. While one could take the copter inclination for obtaining angle of attack for a level flight (i.e. 0 degree wouild be equal to 10 degrees AOA, an inclination of 3 degrees backwards would be equal to an AOA of 7 degrees (for the wing mounted at 10 degrees with respect to the copter axis), during the climb the AOA would be significantly reduced. For instance, flying at 20 m/s and having a climb rate of 2 m/s means an angle of 5.7 degrees. So if the wing is at 10 degrees, and speed/climb is 20 and 2, the real angle of attack would be 4.3 degrees! That one I have to think out how to deal with.

For separation, the copter will defintiively produce thrust equal to about its weight, but not forward thrust.

If I can control well the angle of attack the release speed can easily be above the stall speed.

I will make a long post describing each stage of flight and approximate lua script algorythm. The fighting oif the FCs was always on my mind and I will tackle this issue in the post.

[Michail_Belov]

Definitvely this is going to be a long, step by step process, and most probably I will start with some old wing and maybe even some old motor to simulate the real plane. I defintively do not want to start fooling around with the finished plane because the motor is rather expensive (500 USD). There is no room for error or second chance…

[Michail_Belov]

So, in this post I will consider the five stages of flight, both from the pilot perspective and lua script algorythm.

The initial setup would be that one radio controls both the plane and the copter. The plane will have all the four basic controls mapped directly to sticks (Aileron, Elevator, Rudder, Throttle). The copter will have the first three maped directly to sticks, and throttle will be on a different side pot. Initial modes will be MANUAL for the plane and QHOVER for the copter. There would be one channel for the release command, activated by a switch. And there would be another switch and channel to force QHOVER MODE on the copter at any moment.

STAGE 1. Take off, initial vertical climb (unassisted by the wing), pointing the nose, initial acceleration up to a speed of 30 km/h. Here the pilot would control fully the copter in traditional way, using pitch, roll, and yaw controls and seecondary throttle to climb vertically. The pilot can also apply forward throttle for the plane motor. The usual procedure for the pilot would be climb vertically to clear obstacles, point the nose in the direction where the release would happen, and start forward acceleration.

STAGE 2. Once the speed goes beyond 30 km/h, the rudder becomes effective, so the directional steering would be through rudder input to the plane rudder. In order to prevent fighting between copter yaw control and plane rudder the yaw PIDs will be relaxed or even completely set to zero, i.e. the copter will definitively stop correcting for yaw at a speed of 50 km/h or more. That can be easily achieved by lua script, and I have done so for a trimotor VTOL for the transition, and it worked very well. The vertical climb will be controlled by the copter throttle. Up to here the modes remain QHOVER for copter and MANUAL for plane, which would prevent any fighting in roll and pitch, and the pilot should not really touch the roll and pitch stick here.

[LupusTheCanine]

I think you may gain a bit of efficiency by having the copter tilted forward.

I would have the copter pitched forward enough to maintain speed after release if it was in horizontal flight without the plane, this way it will accelerate away from the plane even if it flies straight and level.

[Michail_Belov]

STAGE 3: Climb to release altitude with wing assisted lift. Here the pilot allows the plane to climb to a safe release altitude, continuing to control the direction using rudder, forward speed using the throttle.

The mode for this stage would go from QHOVER to GUIDED MODE and would be invoked automatically based on the following criteria: the speed must be at least 50 km/h (i.e. slightly less than the stall speed of the plane) and the copter power must have fallen. Ideally, one would have to transition to this mode when level hover thrust would be equivalent to the copter weight, but since the copter will be also in vertical climb, the total power will be the sum of hover thrust and climb thrust. In order not to complicate too much some empirical threshold could be used. For example, if static hover requires about 2600 W for 10 kg, and 2 m/s climb about 3100 W we could take 50 % of that power as the criteria for the transition to stage 3.

So, when that is met, i.e. power falls to 1500 W and speed is above 50 km/h, the lua script automatically will switch the mode to Guided mode, with roll and pitch set to 0, and throttle adjusted to the same 1500 W. The throttle setting to a certain power is not that straightforward. It is possible that one would have to make a table for the lua script where lua script will select the throttle setting based on the battery voltage and the required power.

For the pilot, there would be no real change, only the roll/pitch stick will become totally inoperative, but the idea is that the pilot would not touch it at all. The copter throttle would also stop working. The pilot would continue to steer with rudder and could change the forward thrust of the plane motor.

Here the lua script will begin to control the angle of attack of the wing indirectly. First it will wait until speed goes beyond the stall speed of the plane, i.e. something around 80 km/h in our case. Up to that point the pitch would still be at 0 degrees. Once the speed is above 80 km/h the lua script will very gradually input rear pitch, i.e. like pulling back on the stick. Maybe something like 1 degree per 2 seconds until the speed stops growing. Remember, the pilot controls the throttle of the plane. So if pilot would push more throttle, the plane would initially accelerate and increase its speed, lua will detect it, and increase backward inclination. If the pilot reduces the throttle, the speed would begin to fall, and the copter would compensate by changing the pitch inclination to forward. Also the maximum pitch limits would be rather limited, maybe something like 6 degrees backwards and 6 degrees forward, no more.