HobbyKing.com New Products Flash Sale
Jack Crossfire's blog
Posted by Jack Crossfire | Jul 13, 2008 @ 11:56 PM | 3,051 Views
VicaCopter now flies roughly once a week for 10 minutes to stretch the drive belt, redistribute the lubricants, & build confidence. She's basically the same as a Hornet UAS.

We have some ideas to put in landing & takeoff switches, set & retrieve waypoint switches. Maybe a numeric keypad. Maybe just pilot commands on the laptop. A battery indicator would be good too. Not really practical to carry a laptop wherever she flies.

To save on the weight of sonar & reduce the danger, they would rely on the pilot to sense ground level & switch in & out of landing & takeoff programs. This could make landing & takeoff in high wind much easier.
Posted by Jack Crossfire | Jul 13, 2008 @ 01:20 AM | 3,236 Views
So basically, magnet UAV is old news & the next big thing in magnet UAV's is levitation by repulsion. The first commercially feasable levitation by repulsion was patented in July 2007. Before then, you had diamagnetism, spin stabilization, & attraction.

The theory is a very large toroid permanent magnet (11) repels the opposite poles of a small disk magnet (16) if the disk magnet is in the center of the toroid field. It actually stabilizes it in the vertical axis & the rotation axes.

3 pairs of 2 electromagnets (A1,A2,B1,B2,C1,C2) in the base stabilize it in the horizontal axes. 3 hall effect sensors (A0,B0,C0) in the base sense horizontal position.

Not sure we would have had the budget to implement something this complicated just for cleaning contact lenses. Maybe U $5 million home owers can spend your mortgage stimulus checks on it.

Magical GRAVITRON Levitation Technology From Levitation Arts (5 min 6 sec)


Patent #20070170798

In other news, KQED began showing low definition movies on its HD feed to fill the airtime. Moreover, they crop all 4 sides of it, a practice which used to be illegal in the analog days. What a joke. Where's my TV stimulus check?
Posted by Jack Crossfire | Jul 10, 2008 @ 05:45 PM | 2,773 Views
FYI, July 9 was national geriatrics day, where we observe the process of getting old. U should all have eaten a cake or ice cream. It'll be 40% more expensive next July 9. Unfortunately some of us were trying to get Ubuntbunt to work on our laptops. After 40 years, someone's still finding things to redesign in UNIX.
Posted by Jack Crossfire | Jul 10, 2008 @ 12:40 AM | 2,958 Views
Hillary's prayers against Amazon finally overwhelmed us. In fact, this is the first laptop ever paid for out of the Jack Crossfire general fund. The days of corporate freebees R over.


Acer Extensa EX5620-4020 Intel Pentium dual-core T2370(1.73GHz) 15.4" Wide XGA 2GB Memory 250GB HDD DVD Super Multi Intel GMA X3100 NoteBook - Retail

$549.99 + $100 tax & shipping


Its screen is 3/4 the resolution of Amazon's 2005 resolution & much more faded. Its CPU is 670Mhz slower than Amazon 2005. It has no RS232 or parallel port. Its memory & hard drive R twice Amazon 2005. The combination of weight & inflation put historic screen & CPU speed out of reach. Only memory got cheaper, thanks to Micron's daily writedowns.

Fortunately, the dual cores make it very responsive & Linux actually supports some OpenGL features on it. Forget about sleep mode & OpenGL shaders of course.

In 2005, we needed a powerful heavy desktop replacement for tradeshows & demos. Now we need a disposable, light ground station for VicaCopter. Laptops fall apart too quickly to justify spending the money to recover what we had in 2005.
Posted by Jack Crossfire | Jul 08, 2008 @ 05:24 PM | 2,881 Views
With magnet UAV, every payload had a random preferred yaw orientation. Tried to line up the heater with the payload to transfer the most heat but preferred orientation was never certain.

Now we have an explanation for the yaw preferences. The payloads are off balance & each payload very slightly tilts the lift magnet a different way. The lift magnet feels no yaw force when it's perfectly vertical, but given the slight tilt, feels a yaw force from the Earth's magnetic field.

So magnet UAV's can have yaw control with no extra mass.
Posted by Jack Crossfire | Jul 06, 2008 @ 10:53 PM | 3,055 Views
Have found if U mention a seller & low price for a hot product on rcgroups, the seller immediately sells out, so no sellers 4 U.

Storm case M2950: $187
5 day UPS shipping: $60 each way
Airline shipping: $105 each way

Sending VicaCopter to Fl*rida for a shoot would cost over $310. It would probably be cheaper to pay someone to fly U over the property & the picture quality is better. On the other hand, U would own a really good case for other uses & only pay shipping costs of $120 + 40% annual inflation for every trip afterward.

Had $200 in clothing expenses in June because we expected 100% clothing inflation in July. The cost of driving to our day job has been doing a lot more than doubling.

July: $1178.44 so far
June: $457.95
May: $203.45
April: $142.21

Suddenly a Storm case isn't so expensive anymore. Would expect Storm cases to get cheaper if traveling decreases, but so far 100% travel inflation has been offset by free credit.

The software plan is down to implementing neural controllers for target attitude & cyclic, but with 3 contenders for each.

1) Dynamic inversion using the last n seconds of navigation for training feedforward networks.

2) Predicting future state again & using traditional PID loops for control. Using the last n seconds of navigation for training feedforward networks.

3) PID emulation by recurrent networks & genetic training online based on observed success.

Based on The...Continue Reading
Posted by Jack Crossfire | Jul 06, 2008 @ 01:33 AM | 2,438 Views
Well, review Henrik's video & decide how stable VicaCopter is 4 yourself. Henrik has DGPS & a streamlined airframe which doesn't toss around the rotor wash as much, a fully GPS assisted Kalman filter, & that precious neural controller. Other than that, it's the same T-Rex, same GPS module, same computer, same IMU.

Henrik Copter (1 min 26 sec)

Posted by Jack Crossfire | Jul 03, 2008 @ 06:45 PM | 2,419 Views
Got foot power above 12mph for 9 seconds. Extending it to 20seconds brings it down to 9mph. It's duration vs. speed in the Hillary escape route.
Posted by Jack Crossfire | Jul 03, 2008 @ 02:23 AM | 2,368 Views
Got a video of some autonomous hovering in front of an area near Moffet field. Notice the hangers. The computer executed some turns & there was a light breeze. Light breezes sound a lot nastier on camera. Note tail wag.

Little Miss Moffet (2 min 41 sec)


Looks like log damping was a dud. Completely worthless oscillation returned to cyclic. Reduced throttle to get the RCE-BL35X to give us a good RPM again & the tail started wagging again. Log damping just makes U feel good. Looks like a proper heading hold gyro would do the job, but it's no fun.

So it's really a matter of balancing RPM, weather, & gain. U could have the computer take wind speed, desired stability & look up the right gains.
Posted by Jack Crossfire | Jul 02, 2008 @ 05:44 AM | 2,300 Views
Well, we'll probably never know why the GPS logger is so unreliable. It sometimes gets stuck in record mode when powered on. Sometimes it fails to record the state on EEPROM. Sometimes it can't write the MMC card. MMC cards probably need brown out protection. It was more about getting a mass storage driver on the PIC than logging GPS. It's probably good enough to do Hillary Clinton egress training. For GPS logging, you should obviously support a commercial product instead of doing it yourself.

Next, the reason we went with PIC for VicaCopter was PICs have the highest pin count in the smallest through hole packages. If Atmel just offered 44 & 68 pin PLCC's, it would have been Atmel. PIC's are horrible at C programs. Not sure about Atmels because a 16Mhz AT91USB we used in 2006 was also pretty slow. Alternatively, U could make a middle management salary & use surface mount parts.

June cashflow:

$38.05 - tail servo replacement. Airtronics 94761 replaced by Futaba S3102.
$36.58 - Gumstix hirose replacement. Theoretical cause of engine cutoff but it wasn't.
$21.21 - tail rotor hub replacement. Theoretical cause of high frequency vibration but it wasn't.
-------------
$95.84
Posted by Jack Crossfire | Jul 01, 2008 @ 07:36 PM | 2,240 Views
AP wrote:
> 12,000 workers, or 7 percent of Starbucks' global work force, will be affected by the closings

Whew. And U thought aerospace was bad. Looks like your fast food career is going to have to wait, too.

telegraph.co.uk wrote:
> Chinese astronauts were on schedule to get to the moon by 2017 or 2018

Something that's not a green tech startup nor a social networking startup. Now the Chinese might be on to something.

telegraph.co.uk wrote:
> "If we really wanted that to happen, we sure should have
> started putting more money into that programme sooner."

Buzz Aldrin did not say "programme".

telegraph.co.uk wrote:
>"Americans have been watching for over 25 years spacecraft
> coming back and landing on a runway," he said. "It is going to
> be a bitter disappointment to people here."

They sure spent a lot of time on those missions practicing runway landings. Water landings should free up more time for love, at least.

telegraph.co.uk wrote:
> America must invest now in the space agency Nasa, or
> surrender leadership of space exploration to Russia and China

But does China lead in mortgage entitlement exploration?
Posted by Jack Crossfire | Jul 01, 2008 @ 04:45 AM | 2,359 Views
Well, yesterday's log curve was a success. Tail wag isn't totally gone. Just enough that U can't see it in flight & the stick fights back more. With the T-Rex, U don't see tail wagging, U hear it. The 430XL motor makes an oscillating dying elephant sound when it wags. The dying elephant sound was notably more intermittant with the log curve.

Couldn't get any more gain than the linear curve. Maybe the limits could be extended.

Also reduced tail gain in the autopilot of course. Pointing downwind definitely was more stable. Unfortunately, now got exponential pitch oscillations & losses of pitch control in 20mph wind. Based on Henrik copter, U need a feedback rate of 100Hz & no latency of a serial port to stabilize the T-Rex.

Lost the RCE-BL35X settings again. Since ESC & computer don't have a common ground line & programming the ESC requires 2 batteries, U need some way of grounding the ESC for programming. With Dean on all the power connections, an ESC ground is safe again, but real men plug debugging serial ground into battery balancing tap.

Everything is between HeroineCar repairs nowadays.
Posted by Jack Crossfire | Jun 30, 2008 @ 05:42 PM | 3,463 Views
Now some quotes from The Georgia Tech Unmanned Aerial Research Vehicle by Eric N. Johnson* and Daniel P. Schrage. It's basically the insides of the mighty Henrik Copter.



Does Henrik Copter use a kalman filter?

He uses a 17 state Extended Kalman Filter. The states include: vehicle position, velocity, attitude (quaternion), accelerometer biases, gyro biases, and terrain height error. The system is all-attitude capable and updates at 100 Hz.




Does Henrik Copter use a neural network?

An adaptive neural network trajectory following controller with 18 neural network inputs, 5 hidden layer neurons, and 7 outputs for each of the 7 degrees of freedom. The 7 degrees of freedom include the usual 6 rigid-body degrees of freedom plus a degree of freedom for rotor RPM.





How does he train the neural network?

He uses a simple hover simulation offline, then corrects model errors using online training. U need to pay to find out anything specific.





How did he get automatic takeoff & landing to work?

The navigation system determines if the helicopter is on the ground by comparing the altitude above ground level [sonar] with a preselected [hard coded] value. That is, if the helicopter is within a few inches of the ground it acts as though it is on the ground. Commanded position is slaved to the current position when the helicopter is declared on the ground. Also, internal states of the flight control system are frozen.

Takeoffs are performed by ramping rotor RPM and collective until the helicopter is detected airborne, at which point the trajectory generator produces a smooth climb trajectory. Landings end with a slow vertical descent command until ground contact is detected and rotor RPM and collective pitch are reduced to an idle state.


What language is his software in?

ANSI C/C++ (and the OpenGL API for graphics used in simulations and graphical user interfaces).

Why did GATech do all this work when the R-Max already has an autopilot?

More jobs, more money.
Posted by Jack Crossfire | Jun 30, 2008 @ 01:18 AM | 3,618 Views
Log yaw damping is in, but without an exhaust manifold & with the falling home equity around the golf course, there's no way to test it. Need to get those condominiums back over $4 million so flying isn't a problem again. There was a time when yaw damping was a black art. Now it's routine graphing & theory.

Big surprise in the GPS logging department, the 32MB MMC card died. It recorded half of a walk, then suddenly stopped recording. It would read but not write even when used as a raw device on Linux. Maybe there's a hardware write protect in the protocol. Maybe it was static electricity. An MMC interface without enough RAM to store a complete block is a mess.
Posted by Jack Crossfire | Jun 28, 2008 @ 11:11 PM | 4,349 Views
So much for exhaust manifold 1. Got 8 years & 144,000 miles before discovering it. The exhaust manifold + new oxygen sensors should be from $1000-$2000. It hits 600F so no J-B weld. No aftermarket exhaust manifolds in Calif* are legal for HeroineCar. Amazing that your entire economy is based on rusted, brittle cast iron technology from the 1800's. Without it, your jobs would all be done in India over internet connections.
Posted by Jack Crossfire | Jun 27, 2008 @ 05:10 PM | 3,773 Views
Running as fast as possible is going to be a very important skill under Hillary & what better way to test your chances of survival under Hillary than VicaCopter's spare GPS module.

It's up to where it can record very consistently in shaky conditions once it initializes. Saving & recovering state from the MMC card is still flaky. May just revert that part to the PIC EEPROM. Running speed is slightly lower because we have to carry the GPS board.
Posted by Jack Crossfire | Jun 26, 2008 @ 01:25 AM | 3,723 Views
Getting GPS logger to work in the dumpy apartment was easy. Getting it to work in a hand held, portable device in the real world is a major operation. Going to need a lot more error detection, recovery, & redundancy with that one.

The next step with VicaCopter will be logarithmic yaw damping on the PIC. This should reduce oscillation to increase tail servo life & increase the amount of rate damping available.

http://www.rcgroups.com/forums/attac...mentid=1702634

The equation was pwm = yaw rate * gain / sqrt(yaw rate * falloff).
gain = 0.2
falloff = 4


Found some more showstoppers with the neural network plan. Mainly observing the real world performance of many sets of weights to determine the optimum set would probably take eternity to yield useful results. Each set would need several seconds in the loop to determine its fitness & it takes thousands of sets to converge on a solution. E-Coli evolves faster.

Looking through a few thousand more copies of copies of a few papers on The Goog, correlating observed outputs & inputs in a neural network & inverting it to predict inputs was probably the right direction. The pros call it Dynamic Inversion & it's only possible with enough computing power to crunch huge data tables in flight.

It probably didn't work with acceleration because the data was too noisy & we didn't have enough computing power to gather enough points. There literally was no correlation between attitude & acceleration in 30 seconds of data.

Attitude changes from cyclic could be a different story. A lot more data points in a lot less time & they could be more accurate. Record attitude changes & cyclic inputs so you always have data points in all the ranges of magnitude. Either genetically adapt the recurrent network to get integral behaviour or go back to back propagating the feedforward network from your EM406 days & hope it adapts faster than an integral.