Jack Crossfire's blog View Details
Archive for March, 2008
Posted by Jack Crossfire | Mar 31, 2008 @ 03:25 AM | 7,898 Views
The answer is yes. U can mix a swashplate with a neural network. Most of the time, it works. Made really rough video & an LED tracking program to train the network. The head would always rotate during the tests. The linkages had a lot of play. The camera couldn't be exactly level. Video synchronization seemed to work.

Have a really rough conversion of cyclic to pitch phase in the disc. Since the goal is isolating collective from cyclic, phase alignment is secondary.

Video of training the neural network using machine vision followed by a strapdown test of neural network swashplate mixing.

Swashplate mixing with a neural network (1 min 51 sec)

Well, she certainly flies like this. Can't say it's any better than algebraic mixing. Swashplate phase is way off. Have some nasty glitches when the wrong combination of collective & cyclic is applied.

The credit card is nuked. All things considered, the EOS 5D is only rent inflation since last year....Continue Reading
Posted by Jack Crossfire | Mar 30, 2008 @ 01:27 AM | 7,832 Views
The problem with swashplates is they don't convert servo angle to lift very accurately. Every pushrod, reduction arm, & control horn from the servos to the blades is nonlinear & talks to each other. Collective lift for neutral cyclic is always different than collective lift for steep cyclic.

The ideal CCPM converts the sticks to an exact lift for every phase in the rotor disc, but we can't measure lift very accurately. Lacking direct lift measurements, the plan is to measure blade pitch for every phase. Mount an LED jig on a blade grip & videograph the pitch for the entire range of servo positions. Repeat with the phase in 120' increments.

Then the computer converts the sticks to pitch for 3 phases in the disc. A neural network converts the 3 pitch values into servo values. Hopefully that gives U a smoother ride.

The trick is synchronizing the video with the test program & keeping the head from turning during the test program.

Have our first daylight timelapse in HD.

Daylight timelapse from autonomous helicopter (0 min 32 sec)

Posted by Jack Crossfire | Mar 28, 2008 @ 03:36 AM | 7,853 Views
The RCE-BL35X has proven 2 B a real disaster. Firstly, spool up with too much positive collective & it locks in a hopelessly low RPM. Spool up with neutral collective to get it into a useful RPM. Secondly, it frequently loses its settings during reboots & goes to variable RPM mode.

Now, after 5 min flying the A560 in autopilot, disengaged the autopilot & everything was nominal.

Pulled collective to neutral to descend & the RCE-BL35X went into full throttle. It stayed locked wide open through landing & cutoff. Having pitch set extra high for low RPM made for exciting flying.

Suspect the 35X let its throttle integral increase very slowly in response to the high payload and it didn't drop when the collective dropped. The workaround is to use the autopilot to descend out of the shooting position.

Finally got our first 3k images from the computerized beast. 173 3k images taken during autopilot were scaled down to 1920x1440 & hand stabilized. Gootube can't play 1920x1440 of course. This segment includes operator directed climb in, 2 minutes of autonomous hover & operator directed descent out.

The HD master is on
Night timelapse from autonomous helicopter (0 min 17 sec)

Posted by Jack Crossfire | Mar 27, 2008 @ 07:02 AM | 8,881 Views
So U have 2 choices: replace the entire shutter assembly on the Canon for $400 or get a 5D system for $3000. The 5D requires a flash ($400) & normal lens ($310) but it would make those space station flyby's much easier. Don't think the 5D Mark II is valuable enough to justify $3500. The credit card won't reset until next week, so no Canon replacement for a long long long long, really long time.

Decided we had enough still photos from last year's shuttle/station flyby & photos like this R so rare that it justified making a movie. All 47 handheld photos were hand stabilized. The 200mm lens couldn't bring out any detail. The only novelty is the fact that U know what they R & U can see how low they R by the aspect ratio changes as they move across the sky. Next time, use a 180' wide angle lens.

Shuttle & Station flyby (0 min 10 sec)

Finally figured out a way to stack timelapse photos, so U get some UAV painting.

UAV painting (1 min 40 sec)
...Continue Reading
Posted by Jack Crossfire | Mar 26, 2008 @ 05:34 AM | 7,901 Views
The mane reason for this was to see if the flight controller was robust enough to do it. 3 simple letters is the most a single battery charge can handle. Strong winds R forming every night, making for sloppy movement. These missions R performed completely in hover mode with the safety pilot following her all the way downrange. Increased the velocities to 2m/s to try to extend the battery range. After 20 minutes of careful rotor manipulation, she finished CAT & DOG.

Unfortunately the American electronics worked & the Japanese electronics failed. The EOS 20D died halfway through DOG. It passed 70,000 exposures & now goes into infinite mirror flips whenever powered up. Suspect a broken sensor is reporting a stuck mirror & it's trying to unstick the mirror. It was rated for 100,000 exposures & only did 70,000.

Given our 35mm lens collection, the obsolete EOS 5D may be the next move. The only major purchases for 2008 so far R the T-Rex airframe ($1000) & the HP ($1300). Everything we buy works until it either crashes or desintegrates from old age.

Robotic sky writing (1 min 23 sec)

Posted by Jack Crossfire | Mar 25, 2008 @ 04:40 AM | 8,860 Views
Finally got around to making operator directed velocity actually work. U can't just couple the stick to X & Y velocities because wind just blows it off course & increases velocity in the undesirable direction.

The trick is keeping movement on 1 axis. In copter frame, just keep copter X position constant while changing copter Y velocity. In world frame, the axis U want to stay 0 is related to heading.

To simplify the math, just have a fake waypoint farther down the direction line & have her follow it.

Altitude is a different story. Vortex ring states occur after every movement which is long enough to build up translational lift. The workaround is storing the collective integral while hovering & recalling it after the movement is finished. Also adding a bit more to the movement when commanded to stop so the stop is more gradual.

This field is just too small. Lost tail rotor again & ran into a tree. Got a lucky crash right side up.
Posted by Jack Crossfire | Mar 24, 2008 @ 03:49 AM | 8,171 Views
The Gumstix has been rebooting itself a lot. Managed to get the Canon A560 in the air for the first time. On the first engine spoolup, the computer rebooted spontaneously & the engine stopped. On the next engine spoolup, it was stable. The T-Rex probably has the same vibration as the Corona. The A560 handles it better than the TX-1 because it doesn't have a wobbly stabilizing lens. The days of calm afternoons R over until October. Could only manage 2 minutes of autopilot before the wind took her away.

Download the raw version from http://vimeo.com/818267.

The camera flies again (3 min 50 sec)

Got some nighttime airborn footage. Unfortunately the Gumstix rebooted during the landing. Engine cutoff didn't work & she thrashed around for a while. Fortunately the battery was too dead to do any major damage.

Airborn Canon A560 at night (1 min 24 sec)
...Continue Reading
Posted by Jack Crossfire | Mar 21, 2008 @ 03:57 AM | 7,680 Views
Not exactly pumping collective but definitely changing altitude. In the deadest air in the universe, she did 2 ascents to 145m & 2 descents to 135m perfectly in 3.5 minutes. .5 m/s is definitely the descending limit before vortex ring state. Note since the satisfy radius is 5m, the waypoints are 150m & 130m and she never reaches them except after completion.

The strange thing with these tests is in real life the waypoints R nothing but empty space & she's dialing in positions in empty space which no human can quite grasp. To the satellite network the points exist as definitely as rock.

The hovers were running so well, decided to try to get the moon in the same shot again. The trick with the night shots is the moon blows away all artificial lighting. That means U need to hover as close to the camera as possible to get enough light on the copter & stop all the way down to begin to get some detail in the moon. It's autonomous guillotine flying at its best. Definitely need to try some daylight moon shots.
Posted by Jack Crossfire | Mar 20, 2008 @ 05:22 AM | 7,677 Views
Reduced the roll cyclic by 1/2 & that seemed to stop the oscillations. It makes sense since the CCPM gain was doubled to reeduce the collective crosstalk. Strong wind is back, making comparisons impossible. She's doing quite well in this wind, however.

Stepped up operator directed climb rate to 1m/s & autopilot climb rate to 0.5m/s. 1m/s is pretty unstable. U can get better results by adapting the velocity to the wind.

High wind: 0.25m/s
Calm wind: 0.5m/s

Combined with the manual velocities in the waypoint files, starting to evolve a gain schedule.
Posted by Jack Crossfire | Mar 19, 2008 @ 05:18 AM | 7,835 Views
Well, after banging on it, decided the easiest solution for the cyclic crosstalk was to make longer pushrods but a metric tap costs big bucks. We're trying to use lower head speeds than normal to increase flight time.

Other than that, the other open issue is the new roll oscillations. They seemed 2 B a new artifact of the neural network, so reduced the neural network gain & the oscillations didn't go away. Back to attitude hold tests.

Altitude is still solid with the cyclic tuning. Have not had a single vortex ring state in 2 days. Amazing how invisible cyclic crosstalk is to a human yet the computer can't handle it.

Did the long mission with only 1 90' turn. 4 minutes with the turn. With the downgraded neural network gain & 4m/s velocity targets, she blew threw that mission like Hillary. The safety pilot got destroyed trying to keep up with her. The neural network didn't over correct as much & the route was sloppier as expected.
Posted by Jack Crossfire | Mar 18, 2008 @ 03:58 AM | 9,324 Views
The answer is yes. There was crosstalk between cyclic & collective. 1mm of down collective in maximum roll 2 B exact. While we can't eliminate the crosstalk, turned it into 1mm of up collective.

Despite the doubters, the proof was in the test flight. Maybe it was the weather, but no vortex ring states & the tightest altitude of any mission. CCPM crosstalk definitely seems to impact results.

Unfortunately, the sprinklers R coming on again, so we're watching copter & sprinklers simultaneously.

Eliminating all CCPM crosstalk mechanically seems 2 B impossible. The only way would be to measure the exact dimensions & mathematically model the pushrods, horns, & swashplate. Then there's replacing the swashplate with a solenoid. Then of course, U can have the computer set every possible servo position & photograph the swashplate, storing every input & position....Continue Reading
Posted by Jack Crossfire | Mar 17, 2008 @ 03:01 AM | 8,074 Views
Reduced the turn rate to 4.5 deg/sec & that basically ended drifting during turns. It's not as fast as the industry standard but this is an after-tax operation.

With such slow turns, the mission came to 5min. Previously it was around 4min30sec. The advantage is more chance of actually finishing. Pretty confident, given enough altitude & dead calm air, she could perform a mission beyond visual range....Continue Reading
Posted by Jack Crossfire | Mar 16, 2008 @ 02:51 AM | 8,001 Views
So the GPS satellite launched today does L1, L2 signals but no L5. There's now a choice between sonar, laser range finding, computer vision, & just waiting for better GPS accuracy.

1575.42 MHz
Satellites: Block I 1978, Block II
Completion: 1995

1227.60 MHz
Satellites: Block IIR-M 2003
Completion: 2013
Value: dual-frequency ionospheric delay correction with L1. Makes DGPS obsolete

1176.45 MHz
Satellites: IIR-20(M) June 2008, Block IIF, Block III 2013
Completion: When IIR-M satellites reach end of life and are replaced
Value: 6 dB higher power, more bandwidth, longer code, 20 dB higher power
spot beams for Calif*

3cm, fast velocity readings from carrier phase will never reach consumer level because it's not used anywhere.

Next, we have software revo mixing & software trimming. In the name of the tightest heading possible, added software revo mixing to the computer. Pumping collective is definitely a bit easier with revo mixing. It doesn't lock heading, but it greatly stabilizes the yaw. 3D masters have briefly dropped that they don't use heading hold gyros, but not often since they're paid to sell heading hold gyros.

Global equilibrium doesn't seem to work. In a very long 11min flight, she built up a positive roll to counteract strong wind. Beginning the counterclockwise turn, the positive roll should have turned into positive pitch but instead stayed a roll & moved her way way out west.

Now leaning towards a play between the IMU tilt & the equilibrium tilt.

> North-south surface gradients have been increasing after upper trough
> passage and gale force gusts are currently being reported at the
> offshore buoys.
> Winds have begun to pick up in the hills and should
> continue to strengthen at the higher elevations overnight

That's why this problem is taking a long time to track down.
Posted by Jack Crossfire | Mar 14, 2008 @ 05:43 AM | 7,997 Views
Well, decided the Maxbotix isn't robust enough, there aren't enough weather windows to play around, so it won't even be giving a sonar hover demo.

The idea of a custom sonar velocity sensor is still intriguing. The Maxbotic transducer goes up to 117dB using 20V. The strategy would be to buy another Maxbotic transducer & transmit a wave on it full time using a custom circuit. Then we would use the existing board in full time receive to detect the returning wave's frequency. The filters would be powered up but the output would go to an alternative PIC.

So much for that. Time for timelapse documentation of some autopilot missions. 1 used ISO 400, 2sec, F2.8. 2 used ISO 1600, 1/250 sec, F2.8. 1 was a sideways dolly, 180' turn, sideways dolly, & victory hover. 2 was a static hover. Note how the position changes as GPS drifts over time.

The batteries could not do more than a handful of flash photos before dying, & the shutter computer didn't work, so we were swapping batteries every few seconds, manually opening shutter, & standing by on the autopilot abort.

Flight 1 was 10min. Flight 2 was 11min & battery exhaustion.

Autopilot timelapse (0 min 45 sec)

Posted by Jack Crossfire | Mar 13, 2008 @ 03:41 AM | 7,734 Views
The answer is yes, if aggressively filtered in software, isolated from rotor wash, the Maxbotics EZ1 could probably hover 1-2m off the ground.

If it's on the ground, too high, or it doesn't receive a reflection, the Maxbotics outputs a maximum altitude. Most of the time it just outputs a random glitch & it's up 2 U to detect it. On 5V, it simply doesn't generate enough sound pressure to cut through engine noise & reflect off grass.

No-one wants to hover 1-2m off the ground because of decapitation risk, though it would make a nifty demo. It could enforce an altitude floor. It would require a heavy enclosure on the tail boom to isolate it from rotor wash & achieve the necessary calibration height. Mike Fouche's $17,000 beasts don't even use sonar.

The Maxbotics EZ4 might reject engine noise better, but we've played this game before. There's no such thing as a free lunch. It looks like it rejects noise but has shorter range.
Posted by Jack Crossfire | Mar 12, 2008 @ 02:46 AM | 8,025 Views
Well, got the Maxbotix EZ1 integrated & the fun began. First impressions R that it can't handle the vibrations or noise of flight. Got slight glitches in a handheld test too, but clearly flight is out of reach for the Maxbotix.

Sad, considering the software for sonar guidance was already written.
Posted by Jack Crossfire | Mar 11, 2008 @ 03:47 AM | 7,418 Views
Found the ideal target V 2 B 4m/s in the forward flight mode. That got real velocity between 0 & 2m/s, the fastest Americans can walk before they fall over.

After 5 tries, finally got her to finish the long distance mission. The mission took 8m50s & total flight time was 11min. Altitude hold with this mission was pure luck.

Set waypoint 3 2m lower than the others to keep her in visual range. Altitude hold is so bad with waypoint #3, suspect a weather change in that area. Had a huge drift with the turn. Also had a bug where target heading suddenly snapped to waypoint heading.

Got gas today even though we didn't need it. There definitely were more cars than usual but not the 70's lines. People R starting to figure out they can save 20% by keeping it over half full, despite the internet hypnosis machine.
Posted by Jack Crossfire | Mar 10, 2008 @ 03:45 AM | 7,611 Views
So now we hold target speed at a certain large number for forward flight & scale it for hovering. The mode is determined by the waypoint file & there's a complicated rule for turns.

If SPEED=0, the hover algorithm is used to track the waypoint. If either current waypoint SPEED or next waypoint speed SPEED is 0, the hover algorithm is used to make the turn. If SPEED is >0 for the next waypoint, the target speed after turn completion is held at SPEED.

With missions this long & only 2 batteries, only get 2 flights per commute.

Unfortunately, on flight 1 with SPEED=2 the battery died after finishing half the mission. Accidently had the final 1/3 of the mission in SPEED=0 & she crawled to the final waypoint. Flight time: 12m.

On flight 2 with SPEED=8, it was faster than the safety pilot could run. When she hit waypoint 2 she dove for the ground, entered altitude oscillation & couldn't recover. Looks like she barely missed it & tried to change direction suddenly. Without adapting to the new direction, she ended up stationary & couldn't handle the loss of translational lift.
Posted by Jack Crossfire | Mar 09, 2008 @ 04:41 PM | 7,315 Views
Before the next monster flu set in, got some autonomous hovering video in high wind. Given enough space, she can actually do it without sonar altimeter.

As Silicon Valley real estate inflation gets higher & higher & higher, the offices get smaller & smaller & the sick time gets worse & worse. Soon every company will share the same office & use it for 1 hour a day. Cisco will use it at 1am. Adobe will use it at 2am. Goog will use it at 3am.

Autonomous hover in high wind (1 min 34 sec)