Ublox Dead Reckoning GPS - Page 5

jmbickham · 02-26-2006, 09:52 PM

Kenwood has a device... KNA-PG100. I can't find the details now, but before it had said that it made a speed pulse by using the steel belts in car tires.

So ehiebert, do you think that UBLOX with DR is worth all of the extra money, based upon your expereience?

aoLhaTer · 02-26-2006, 09:52 PM

Hey ehiebert

Did you account at all for the variances in speed pulse protocol? (ie how some auto makers use 2000 ticks = 1 mile and others use 8000 ticks = 1 mile, etc)

Scouse Monkey · 02-27-2006, 10:10 AM

Very interested in the signal system you made bud. Been looking at this DR stuff with JCDillin and a few of the boys for about a year now. I was supposed to be advising on a DR system development by aerospace/avionics students at work but the kalman filtering was a littl bit heavy for them.

For the reverse switch we always intended using a relay on the reverse light and what you say about the sine wave is very interesting. I was wondering if it would be possible to pull the signal down into more of a binary signal.

Andy

Quote: Originally Posted by ehiebert

Hey all... newbie member to this particular forum, but not new to GPS.

Anyways, we used the ublox module with DR/Gyro input on a project we did for the local transit authority to provide them with an AVL system. For prototyping we tried to used the speed pulse signal from the transmission of a 1992 Ford Explorer. Problem with that was two fold..... no direction indication for forward/reverse, and the pulse was more of a sine wave with very little magnitude at slow speeds. Consequently, the speed tick input probably did more harm to position integrity than good.

So I deveised another solution which worked perfectly. I rigged an optical quadrature encoder (200 PPR) to one of the rear wheel hubs, and then used one of the phases for a speed signal, and the phase differential to provide a F/R input. This required some interfacing... that was done on the same carrier board we used for the UBLOX... which was a PIC 16F4550 based microcontroller board which I did.

After all of this, we could disconnect the antenna and drive a few circuits of an underground mall parking garage. At the end we finished with a position error of about 5m. That was after many turns and probably 2 km of tavel with no antenna. Repeatability with combined satellite lock and inertial signal was in the 1 m range. It was quite impressive...

I have the schematic for the carrier board if anyone is interested. I could also supply the firmware for the PIC. I had combined this board with some opto-isolated I/O and relays for random stuff. The whole thing then got wrapped into a custom USB protocol for reading NMEA sentences, reading inputs and setting outputs.

Now I'll compare this to a SiRF-III BU-353 GPS mouse..... the SiRF-III is an incredible value for its ~$100 price range. In drive testing it is just as accurate as the UBLOX without DR-enabled, with repeatability coming in at an average of 2-3m. Repeatability testing was done using software (i wrote the test software also) which would trigger a light at a trigger waypoint. Then a video camera is setup to watch the vehicle cross the marked waypoint and record when the light comes on. Since standard update rate of the UBLOX and SiRF is on 1 second boundary we used time division, linear extrapolation as a resolution enhhancement algorithm. This provided us with real-time estimated position updates at 100 Hz frequency. From this testing we noticed that the SiRF and the UBLOX were almost identical in absolute poisition accuracy. When the speed sensor input was added to the UBLOX it bacame a much more accurate interface. However, for the money the SiRF is an excellent device.

Eldon.
'94 Explorer
Intel MiniITX MB
8"TFT
Linux integrated application software (protoyping... and un-named as yet)

aoLhaTer · 02-27-2006, 11:20 AM

From what I have collected, a sine wave is fine so long as the rise is 3.5v or higher and the fall is 1.5v or lower

ehiebert · 02-27-2006, 12:16 PM

Quote: Originally Posted by jmbickham

So ehiebert, do you think that UBLOX with DR is worth all of the extra money, based upon your expereience?

In my opinion, the extra advantage of the DR is absolutely worthless in the framework within which we are using these devices. And here is why I say that.....

Typically when we are driving we know where we are... and if that happens to be an underground parkade, we know it.

Auto users want to use a GPS system for routing... not for AVL. If your reason to have positional accuracy all the time is so that you can log your position (through a wireless internet gateway) and provide a chart of where you are at all times, then the DR becomes essential. But how many of us out there want our wives/boss to know where we are et every single minute of the day? NOT ME!!!!

There reason to provide AVL on a private vehicle would be advantageous to locate the car if it were stolen. But that is assuming the thieves enable the wirless internet connection. Maybe they don't know they have to plug their phone into the CarPC?

I went with the BU-353 rooftop mounted. No reason for anymore accuracy than that.

Eldon.

ehiebert · 02-27-2006, 12:27 PM

Quote: Originally Posted by aoLhaTer

From what I have collected, a sine wave is fine so long as the rise is 3.5v or higher and the fall is 1.5v or lower

The signal we used was unconditioned directly from the transmission tailshaft. Voltages ranged fro 30 mV to 18V from the speed "generator" signal.

Quote: Originally Posted by Scouse Monkey

Very interested in the signal system you made bud. Been looking at this DR stuff with JCDillin and a few of the boys for about a year now. I was supposed to be advising on a DR system development by aerospace/avionics students at work but the kalman filtering was a littl bit heavy for them.

For the reverse switch we always intended using a relay on the reverse light and what you say about the sine wave is very interesting. I was wondering if it would be possible to pull the signal down into more of a binary signal.

Andy

You could use a simple resistor voltage divider to adjust the input voltage divider to obtain +5V. The drive the signal though a TTL buffer or Schmidt Trigger to clean it up.

voltage divider:

+12V ----- R=660-------+--------R=470--------------GND
|
V
+5V

Eldon.

Scouse Monkey · 02-27-2006, 06:10 PM

yeah it was something like that I was thinking.

As for the DR...I wanrt to use it as whenever I am in London or Birmingham or Manchester my BU-303 loses all signal. True a new GPS may be better the DR adds that little extra bling.

aoLhaTer · 02-28-2006, 02:05 PM

Well I tested mine again last night with the new revision of JCs board and it seems like good news: Im getting positional data as well as all the UBX protocol data including values from the gyro. This would seem to indicate everythings working

Mounted it up to the top of my case last night

Going to install it now and go for a test drive. Will record a u-center log file so I can analyze the data. Wish me luck

jcdillin · 02-28-2006, 02:19 PM

woo hoo! 0l33l

but that sucks because yet again mine is got a problem, grrr

aoLhaTer · 02-28-2006, 03:41 PM

Done with round 1 of test driving

Heres the log if anyone wants to look at it http://www.dwoloz.com/COM5_060228_202208.ubx

Everything looks good; the map plot is perfect. However, Id like to now test with a complete signal loss. Will go out again later and try to find a parking garage to test in

As a note to any future testers, if you want to log with all the data fields visible you need to be in the UBX protocol. The board is set by default to output UBX from the second serial port transmitting at 57.6k. NMEA output is on the first serial port transmitting at 9600

spaceskater · 02-28-2006, 04:36 PM

please JC or aoLhaTer send me a schematic or just a crud connection diagram of the circuit!!! PLEASEEE

aoLhaTer · 02-28-2006, 04:39 PM

All relevant information is in the Ublox spec pdf (have to register on their site to see it)
THey have a pinout and a rundown of operation

jcdillin · 02-28-2006, 04:48 PM

Yeah the pinout is listed on the datasheet, i'll post the board layout, but I dont have an original schematic because I didnt create one.

jcdillin · 02-28-2006, 04:48 PM

Quote: Originally Posted by aoLhaTer

Done with round 1 of test driving

Heres the log if anyone wants to look at it http://www.dwoloz.com/COM5_060228_202208.ubx

Everything looks good; the map plot is perfect. However, Id like to now test with a complete signal loss. Will go out again later and try to find a parking garage to test in

As a note to any future testers, if you want to log with all the data fields visible you need to be in the UBX protocol. The board is set by default to output UBX from the second serial port transmitting at 57.6k. NMEA output is on the first serial port transmitting at 9600

Great to hear.

aoLhaTer · 03-01-2006, 01:39 PM

Well Im getting great data from the u-center log tool but when I tried to use CoPilot to plot me it just didnt work. It would move me on the map very infrequently (about 10 seconds per move).

I can match up my u-center map plot with with a MapPoint map and its a perfect plot but theres some problem when it comes to 3rd party applications. Im going to try out iGuidance and Destinator and see if it is the same