Forums

Acceleration, gyro, GPS (

Started by graemel December 4, 2009
Hi,
Im a fourth year student and am currently working on a project where I am
attempting to track the position of a racecar in the form of a strapdown
INS. I am trying to use a 2-axis accelerometer(x,y), z-axis gyroscope and
also have GPS coordinates available. I have been directed to look into
Kalman filtering however after much reading I am still unsure about how to
actually apply this to the data in order to improve things.

I would very much appreciate if someone could 'hold my hand' through my
particular application.

Thanks,
Graeme


On Fri, 04 Dec 2009 16:13:48 -0600, graemel wrote:

> Hi, > Im a fourth year student and am currently working on a project where I > am attempting to track the position of a racecar in the form of a > strapdown INS. I am trying to use a 2-axis accelerometer(x,y), z-axis > gyroscope and also have GPS coordinates available. I have been directed > to look into Kalman filtering however after much reading I am still > unsure about how to actually apply this to the data in order to improve > things. > > I would very much appreciate if someone could 'hold my hand' through my > particular application. > > Thanks, > Graeme
I very much doubt that you have enough information available to you. See if you can get your hands on another channel of acceleration and two of gyro. As for the filter itself -- do the Kalman filter thing: make a model of the vehicle motion, linearize it, and turn it into a Kalman filter. You'll need an extended Kalman filter, and the angular representation of the vehicle state will give you fits -- if you find yourself howling to yourself and mumbling at the moon that's normal for this sort of thing. Girls will flock after you -- at least all the ones that are attracted to guys who walk around campus waving their arms in the air and muttering gibberish to themselves. -- www.wescottdesign.com

graemel wrote:

> Hi, > Im a fourth year student and am currently working on a project where I am > attempting to track the position of a racecar in the form of a strapdown > INS. I am trying to use a 2-axis accelerometer(x,y), z-axis gyroscope and > also have GPS coordinates available.
My $.02 to the kindest comments from Tim: if you are planning on using MEMS sensors, this won't work. For INS, you need the accuracy better by several orders. Consider dead reconing by the steering wheel, compass and odometer instead of INS. Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
On Fri, 04 Dec 2009 17:57:48 -0600, Vladimir Vassilevsky wrote:

> graemel wrote: > >> Hi, >> Im a fourth year student and am currently working on a project where I >> am attempting to track the position of a racecar in the form of a >> strapdown INS. I am trying to use a 2-axis accelerometer(x,y), z-axis >> gyroscope and also have GPS coordinates available. > > My $.02 to the kindest comments from Tim: if you are planning on using > MEMS sensors, this won't work. For INS, you need the accuracy better by > several orders. Consider dead reconing by the steering wheel, compass > and odometer instead of INS.
I disagree. Certainly for a gyrocompass-like application where you're trying to deduce orientation by comparing the rotation axis of the earth with the gravity vector then even the best MEMS devices don't hack it. And _most_ certainly the sleazy parts that they put in cars to actuate airbags won't do for anything reasonable. But I've got mileage with Kalman filters using GPS input and some of the better MEMS IMUs; as long as the vehicle is undergoing accelerations distinct enough to overwhelm the combined GPS and accelerometer errors then there's sufficient excitation for the Kalman filter to orient itself. -- www.wescottdesign.com
On 5 Des, 06:59, Tim Wescott <t...@seemywebsite.com> wrote:

> as long as the vehicle is undergoing accelerations > distinct enough to overwhelm the combined GPS and accelerometer errors > then there's sufficient excitation for the Kalman filter to orient itself.
...so this stuff only works if the platform is a panicking spastic rabbit on speed...? Rune

Rune Allnor wrote:
> On 5 Des, 06:59, Tim Wescott <t...@seemywebsite.com> wrote: > > >> as long as the vehicle is undergoing accelerations >>distinct enough to overwhelm the combined GPS and accelerometer errors >>then there's sufficient excitation for the Kalman filter to orient itself. > > > ...so this stuff only works if the platform is a panicking > spastic rabbit on speed...?
From MEMS accelerometers that you can buy, with proper compensation, you can expect error at the order of ~1 cm/s^2. VLV
Vladimir Vassilevsky <nospam@nowhere.com> writes:

> Rune Allnor wrote: >> On 5 Des, 06:59, Tim Wescott <t...@seemywebsite.com> wrote: >> >> >>> as long as the vehicle is undergoing accelerations >>>distinct enough to overwhelm the combined GPS and accelerometer errors >>>then there's sufficient excitation for the Kalman filter to orient itself. >> >> >> ...so this stuff only works if the platform is a panicking >> spastic rabbit on speed...? > > From MEMS accelerometers that you can buy, with proper compensation, > you can expect error at the order of ~1 cm/s^2.
Isn't the required accuracy of the accelerometer dictated by how often the GPS readings are taken, how accurate they are, and the overall required accuracy? -- Randy Yates % "Bird, on the wing, Digital Signal Labs % goes floating by mailto://yates@ieee.org % but there's a teardrop in his eye..." http://www.digitalsignallabs.com % 'One Summer Dream', *Face The Music*, ELO

Randy Yates wrote:
> Vladimir Vassilevsky <nospam@nowhere.com> writes: > > >>Rune Allnor wrote: >> >>>On 5 Des, 06:59, Tim Wescott <t...@seemywebsite.com> wrote: >>> >>> >>> >>>>as long as the vehicle is undergoing accelerations >>>>distinct enough to overwhelm the combined GPS and accelerometer errors >>>>then there's sufficient excitation for the Kalman filter to orient itself. >>> >>> >>>...so this stuff only works if the platform is a panicking >>>spastic rabbit on speed...? >> >>From MEMS accelerometers that you can buy, with proper compensation, >>you can expect error at the order of ~1 cm/s^2. > > Isn't the required accuracy of the accelerometer dictated by how often > the GPS readings are taken, how accurate they are, and the overall > required accuracy?
I was thinking in context of the very practical problems of filling the gaps in GPS coverage when navigating in urban areas, and a "small scale" navigation in the areas of several m. Unfortunately, MEMS inertial doesn't seem to be adequate for that type of work. Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
On 12/5/2009 8:04 AM, Vladimir Vassilevsky wrote:
> > > Randy Yates wrote: >> Vladimir Vassilevsky <nospam@nowhere.com> writes: >> >> >>> Rune Allnor wrote: >>> >>>> On 5 Des, 06:59, Tim Wescott <t...@seemywebsite.com> wrote: >>>> >>>> >>>> >>>>> as long as the vehicle is undergoing accelerations >>>>> distinct enough to overwhelm the combined GPS and accelerometer errors >>>>> then there's sufficient excitation for the Kalman filter to orient >>>>> itself. >>>> >>>> >>>> ...so this stuff only works if the platform is a panicking >>>> spastic rabbit on speed...? >>> >>> From MEMS accelerometers that you can buy, with proper compensation, >>> you can expect error at the order of ~1 cm/s^2. >> >> Isn't the required accuracy of the accelerometer dictated by how often >> the GPS readings are taken, how accurate they are, and the overall >> required accuracy? > > I was thinking in context of the very practical problems of filling the > gaps in GPS coverage when navigating in urban areas, and a "small scale" > navigation in the areas of several m. Unfortunately, MEMS inertial > doesn't seem to be adequate for that type of work.
There are several applications that already exist where successful systems operate more or less as described, i.e., a MEMS sensor plus a GPS provides good tracking capability for a race car. Most of these apps provide accurate-enough traces of the line on the racetrack and acceleration data to determine lateral g-loads, speed, and forward acceleration. There's an app for iPhones/iTouch devices that works reasonably well, but not quite as well as the units with dedicated hardware. In fact, the group I race with is now considering using devices just like that for compliance testing of rules. The rule set we run with classes cars primarily by power/weight ratios. Weight is easy to determine by putting a car on a scale as it comes off the track. Currently power is tested with a portable dynomometer, which is expensive, not always available, and not hard to defeat by somebody really intent on cheating. Putting a MEMS/GPS device on a car that's on track, however, will readily reveal whether the car accelerates faster than it should for a given ptw ratio. If somebody sandbags when the device is on the car, the lap time will reveal it. So these things do already exist for this application and they work quite well. The only thing I'd mention is that a three-axis accelerometer is probably going to be needed over a two-axis, since race cars do roll in corners, and so do race tracks, some very significantly. The existing devices use 3-axis MEMS and GPS. Maybe the point of the project is to figure out how to use the gyro instead of the third axis, in which case, it does make it harder but it's certainly possible. Check these out: http://www.maxqdata.com/ http://store.traqmate.com/ Or you can peruse the on-line pages of Grassroots Motorsports for the ads on similar devices. I'll start you on page 16, where there's a pic and a paragraph about a very famous car: http://grassrootsmotorsports.epubxpress.com/wps/portal/grass/c0/04_SB8K8xLLM9MSSzPy8xBz9CP0os3iLkCAPEzcPIwODUFMjA0-TYM9QJ_9AQ39fM_1I_ShznPKWRvohIBMz9SNNTSxMQMxi_UgDEF2gH2loBhbILy1KTtWPTCvKTM1L0S_ITkyqSk2qcnRUVAQAdT9Hzw!!/ ;) -- Eric Jacobsen Minister of Algorithms Abineau Communications http://www.abineau.com
On Sat, 05 Dec 2009 04:39:09 -0800, Rune Allnor wrote:

> On 5 Des, 06:59, Tim Wescott <t...@seemywebsite.com> wrote: > >> as long as the vehicle is undergoing accelerations >> distinct enough to overwhelm the combined GPS and accelerometer errors >> then there's sufficient excitation for the Kalman filter to orient >> itself. > > > ...so this stuff only works if the platform is a panicking spastic > rabbit on speed...? > > Rune
Well, that would certainly allow for cheaper sensors. -- www.wescottdesign.com