Hi, I am building a INS with three gyros and three acclerometer. The problem is that the gyro drift and the accelerometer is quite noisy. The accel noise is something I can deal with because I can filter it out, but the drift of the gyro is basically screwing me up. (the problems it that it is drifting in both direction) I am thinking of adding additional sensors and use a kalman filter to integrate these sensor values together and hopefully it will be better. Looking around, I cannot find any cheap sensor (< $100), except of another accelerometer and gyroscope, but I am not sure if it is going to help. I did see people building a INS with only six accelerometers mounted with different orientations, but I don't really want to have something oriented other than on the axis of xyz. Does another additional accelerometer on the xyz axis help? Using accelerometer to calculate the orientation will be nice, but the thing can be moving... Not sure if a low pass filter will help.. (I don't really think so) so .. any ideas? I tried using a 3d compass for other projects and it is a nightmare. Calibration is .... horrible, so I will avoid that as much as possible. Terence This message was sent using the Comp.DSP web interface on www.DSPRelated.com
INS gyro alternative or enhancement
Started by ●July 14, 2005
Reply by ●July 14, 20052005-07-14
blackmangoes wrote:> Hi, > I am building a INS with three gyros and three acclerometer. The problem > is that the gyro drift and the accelerometer is quite noisy. The accel > noise is something I can deal with because I can filter it out, but the > drift of the gyro is basically screwing me up. (the problems it that it is > drifting in both direction) > > I am thinking of adding additional sensors and use a kalman filter to > integrate these sensor values together and hopefully it will be better. > Looking around, I cannot find any cheap sensor (< $100), except of another > accelerometer and gyroscope, but I am not sure if it is going to help. > > I did see people building a INS with only six accelerometers mounted with > different orientations, but I don't really want to have something oriented > other than on the axis of xyz. Does another additional accelerometer on the > xyz axis help? Using accelerometer to calculate the orientation will be > nice, but the thing can be moving... Not sure if a low pass filter will > help.. (I don't really think so) so .. any ideas? > > I tried using a 3d compass for other projects and it is a nightmare. > Calibration is .... horrible, so I will avoid that as much as possible. >Folks are doing amazing things integrating inertial sensors with GPS. Basically the GPS helps to calibrate the sensors, and visa-versa, in one great big Kalman filter. I have absolutely no idea of what's bubbled down to being available on the web -- this is still a subject in which you can safely base a PhD thesis. -- ------------------------------------------- Tim Wescott Wescott Design Services http://www.wescottdesign.com
Reply by ●July 14, 20052005-07-14
Be nice to not depend on GPS because the thing will be mostly (not all the time) indoor This message was sent using the Comp.DSP web interface on www.DSPRelated.com
Reply by ●July 14, 20052005-07-14
blackmangoes wrote:> Be nice to not depend on GPS because the thing will be mostly (not all the > time) indoor >Well, you didn't mention that. You may consider using a 3D compass to keep the gyros straightened out, if the compass drift is better than the gyro drift. No matter what you do you can't get better than the worst sensor. If you could set out beacons of some sort for the thing you may be able to use them to enhance accuracy, but then you'll be restricted to only working where the beacons are set up, and you'll have to invent the beacon system. -- ------------------------------------------- Tim Wescott Wescott Design Services http://www.wescottdesign.com
Reply by ●July 14, 20052005-07-14
>> >Well, you didn't mention that.Sorry, I forgot :) Having a indoor "GPS" is nice. hmm.... Can you think of a way to use an extra accelerometer (maybe of different brand) to get the gravity vector? Or use gyroscope (maybe of different brand) to enhance accuracy? I am pretty stuck terence This message was sent using the Comp.DSP web interface on www.DSPRelated.com
Reply by ●July 14, 20052005-07-14
blackmangoes wrote:> >> > >Well, you didn't mention that. > > Sorry, I forgot :) > > Having a indoor "GPS" is nice. hmm.... > > Can you think of a way to use an extra accelerometer (maybe of different > brand) to get the gravity vector? > Or use gyroscope (maybe of different brand) to enhance accuracy? > > I am pretty stuck > terence >I guess a ring laser gyro is out of the question.. you may want to reserch Honeywell (was Sperry) and Rockwell Collins aircraft INS systems. They do a lot of this stuff. Mark
Reply by ●July 14, 20052005-07-14
blackmangoes wrote:>>Well, you didn't mention that. > > > Sorry, I forgot :) > > Having a indoor "GPS" is nice. hmm.... > > Can you think of a way to use an extra accelerometer (maybe of different > brand) to get the gravity vector? > Or use gyroscope (maybe of different brand) to enhance accuracy? >Getting the gravity vector is pretty much what an accelerometer will do for you. Gyroscope accuracy correllates pretty directly with cost. You were expressing a disinclination to spend an extra $100 earlier, so gyros that cost $1000 or $10000 are probably out of your price range. If they aren't, here's some places to go: Northrop Grumman (already mentioned). Their 6-axis LTN-200 and 2-axis G2000 gyro products are quite nice, but will run $10000 or more for a 6-axis IMU. KVG. They have "open loop" fiber optic gyros that are quite nice, with 6-axis IMUs in the $4000 range. Systron Donner. They make quartz rate sensors, and have an IMU that's probably also in the $3000 or $4000 range. Cloud Cap Technologies makes 6-axis IMUs from Analog Devices chips. They use about $200 worth of chips, but then the calibrate the heck out of them and sell them for $1000. Crossbow also uses Analog Devices chips, as well as FOGs (Fiber-Optics gyros). You're probably using the Analog Devices gyro chips. I've been doing this stuff for the last 10 years. If you find a way to rival the accuracies of the units above at a fraction of the price please let me know -- I know a few companies that'll buy them. -- ------------------------------------------- Tim Wescott Wescott Design Services http://www.wescottdesign.com
Reply by ●July 15, 20052005-07-15
blackmangoes wrote:> I tried using a 3d compass for other projects and it is a nightmare. > Calibration is .... horrible, so I will avoid that as much as possible.Don't be so quick to discard the magnetometer. Although its absolute accuracy may be "horrible" (on the order of +/- a degree or two), it doesn't really have any inherent drift component, and it can be a valuable reference for estimating the bias of your rate gyros. -- Dave Tweed
Reply by ●July 15, 20052005-07-15
Are you using the Analog Devices gyros? Anyway, gyros drift for a reason, are you using any temperature compensation? The AD gyros are pretty much unusable for an INS unless you temperature compensation them (they have an internal temperature sensor for this) Using six accelrometers for an INS is not ideal and will lower the quality of your INS (assuming you using inertial sensors of a given accuracy)
Reply by ●July 15, 20052005-07-15
$1000! That�s too much for me. I need to sell my house to get this INS working! Yes, I am using ADXRS from Analog Device. Magnetometer? Hmm. Maybe I am having a horrible time because I bought a cheap ass, low quality compass. (It was a 3d analog compass from Honeywell) The magnetomer of the 3d axis does not have the same gain, linearity, so I have a horrible time calibrating it. Any suggestion? (<$100) will be nice. +/- a degree or two is VERY nice. What compass do you use? Can�t find anything that cheap. That�s right. Compass does not have drift, but it is hard to get the gravity vector out of it because anything perpendicular to it can be �gravity�. A 2D compass reading is off when the thing is tilted.It will surely be non linear, which will be hard to plug to a kalman filter. How exactly do you do it?> Accelerometer give me a gravity vectorHow do you remove the body acceleration. You use a low pass filter? May works on car and stuff (since acceleration for those is pretty much a spike). For a flying robot, that�s a little hard (Oh yah, I forget to mention that). Maybe using some sort of special filtering may help. Any ideas?> Use temperature to compensate the gyroscopeGood one! I was thinking of adding a temperature sensor. As you said ADXRS is really unusable without temperature sensor. But... Hmmm.... no matter it will eventually drift, after integration. Maybe compass is a good idea if I can find one cheap. Does using a high pass filter for the gyroscope helps getting rid of the drift? Oh there should be a better way? Terence This message was sent using the Comp.DSP web interface on www.DSPRelated.com