Replies to This Discussion

Permalink Reply by wayne garris on September 2, 2010 at 10:38am

Lance ,
you can use gps heading to bias the yaw gyro . one method that is rarely talked about here is to use a high end gyro like the ADXRS610 , thermally de-stress it , measure the bias & rate temperature coefficients , and use temp reading and the coefficients measured on a temperature controlled rate table to set the bias and rate conversion factors . this can decrees drift to <2 deg/hour on the ADXRS610 . this is how the big boys do it but is costly and time consuming but produces very accurate results.

• ▶ Reply

Permalink Reply by Lance Lefebure on September 2, 2010 at 7:53pm

The gyro I'm using now is a ADXRS401, which is 75 deg/sec. I tried doing temperature compensation with it when I first got it, but the formula didn't seem beneficial. I assume that you guys are using similar gyros on the drones. What kind of stability are you seeing?

-Lance

• ▶ Reply

DeveloperPermalink Reply by jasonshort on September 2, 2010 at 6:10pm

Can you place the magnetometer outside the chassis? Otherwise Optical Flow would work.

• ▶ Reply

Permalink Reply by Lance Lefebure on September 2, 2010 at 7:30pm

No, the sensors would need to be in a box somewhere on the machine, and I can't locate it outside the steel chassis on every machine.

I just googled Optical Flow. The environment this needs to run in is often too dusty for a camera to reliably detect points. Keeping a camera lens clean would be hard, and the dust clouds can obscure the view of the ground. Visibility of less than 20 feet is not uncommon.

-Lance

• ▶ Reply

Permalink Reply by Greg Fletcher on September 2, 2010 at 8:42pm

GPS correction works with the yaw axis on a moving craft. Faster is better. A farm tractor is moving slow and is not skidding around I would hope. Wheel sensors, naa. Maybe an old aircraft directional gyro with an analog output. They have autopilots with heading hold functions run of this gyro. It is an actual spinning gyro and drifts very little in 5 or 10 minutes. Then it would need to be calibrated it's self from GPS travel. Very reliable data. It would surely be analog, maybe needing amplification for the ADC input. You could get one cheap at an aircraft scrap yard. They either are electric or vacuum driven. Vacuum is more common. They seem to work pretty good in an airplane when I have used them.
Set them to the compass every 15 min. or so +-. Its goin on a tractor so I think you can spare the payload weight :)

• ▶ Reply

Permalink Reply by Lance Lefebure on September 2, 2010 at 10:25pm

I have found that using GPS to calibrate the yaw sensor works well at speeds above 2 MPH, and when the operator is driving mostly straight. At speeds less than 2 MPH, the machine can make a drastic change in heading without being picked up by the change in GPS position. If a machine is turning, even up to about 10 MPH, the location of the GPS antenna on the vehicle will make a difference as to heading reported from the GPS system.

Weight certainly isn't an issue. I might pursue this for testing, but using used airplane parts doesn't seem viable for a commercially available product.

-Lance

• ▶ Reply

Permalink Reply by Curt Olson on September 3, 2010 at 5:29am

Another idea would be to install some sort of steering wheel angle sensor and some sort of wheel rotation sensor on the tractor. I've heard of people using this to augment gps when you enter situations where it doesn't do well (tunnels, parking ramps, urban canyons, etc.) If you are interested in more details I could connect you up with a former coworker of mine at the U of MN who has installed some of these sorts of sensors on snow plows, buses, state patrol cars, ambulances ... I don't know about his time availability, but he might at least be able to bounce a few ideas back and forth.

Here's another thought. If you have differential gps available, can you install two receivers and do some math?

One more thought ... I know a professor at the U of MN that has been working on a 3 gps system that uses the phase differential data to very accurately compute the relative orientation of the 3 gps antennas (even though it may not be able to place their absolute position any better than any other gps.) This allows you to compute orientation pretty accurately. I believe there are commercial systems that can do this (I don't know any names) and I've heard they can range upwards of $20-50k.

From my personal experience, I think you will probably have trouble squeezing the performance you need out of mems gyros. The fancy 15-state kalman filter I run on my UAV requires substantial motion and accelerations to converge properly (but then it does a really nice job of converging to true heading and estimating the bias of the gyros.) But if you are puttering along at < 2mph I don't think the yaw estimator on our filter would hold together or be able to even converge initially.

Sounds like a fun challenge! :-)

• ▶ Reply

Permalink Reply by Lance Lefebure on September 3, 2010 at 6:59am

We use a steering wheel angle sensor too. In a rough field, the tractor can change heading by a few degrees without turning the steer axle. We're using the vehicle's heading and speed, distance from the desired path, and some dimensions like wheelbase length to calculate a desired steering wheel angle. The angle sensor then tells the system which way the wheels need to be turned to get to the desired angle.

We run either DGPS or RTK. A company called AutoFarm has patented a system with two receivers that will generate an absolute heading without any movement. Their systems are quite popular with vegetable growers where tractor speeds are below 1 MPH.

NovAtel has a similar method for finding absolute heading between two receivers called NovAtel ALIGN. The dual receiver concept works great, but I would like an IMU solution so I don't have to spend an additional $7000 on a second receiver.

-Lance

• ▶ Reply