My autonomous rover - GPSBot - Discussions

Here are a couple pics of my GPSBot

The processor is an ATMEGA32.

The problem is the GPS sucks! The processor calculates the angle for the first waypoint, and constantly reclaculates the angle as it is moving, but when it finally thinks it has arrived at the waypoint, it may be 60 or 70 feet away! The way point data is collected when I walk to the location and press the red button. When it's time to retrace the waypoints I press the black button. My GPS sensor from parallax is probably a year and a half old (or more) and has been discontinued. I am thinking about getting another one. Any recommendations? Can someone tell me why the ArduRover seems to have such accurate positioning? Can you get a GPS sensor with DGPS built in?

Russ

Relying on GPS alone is dicey. A good GPS can probably pull off 5m accuracy or so. 60-70 feet is about 20m which is pretty bad, yes.

You can do several things to improve performance.

One thing you can do to improve reception is to throw a ground plane under the module. Just a square of tin or something, about 4-5" on a side. If you are only locking into 3 or 4 sats, your position error will be awful. What is the HDOP values you are getting? Anything above about 1.2 means very large position error. Even at HDOP=1.2 or lower still has some pretty big error (5m is what I'm usually seeing on the two gps units I use)

Another tip: don't measure waypoints with the GPS itself because you've got error in your waypoint AND error in your current position. The error could cancel itself out or it could double your error or anything in between. Instead, if you use something like Google Earth to pick your waypoints you reduce one source of error. To address Google Earth error, calibrate your vehicle to a known landmark / known lat/lon pair in Google Earth.

I'm extremely pleased with the Venus638FLPX that I put on my rover. With the ground plane I get really good signals even in difficult areas. Position is still dicey. I've also used a SiRFIII (Paros iGPS-500) that seemed to do pretty good, but not as good as the Venus.

The downside with the latter is massive amounts of filtering, even with the high dynamic range firmware. So heading signal bandwidth is very limited; you turn sharp, the gps takes a couple seconds to catch up. I'm no sure all the filtering helped position estimation all that much either.

On the Venus GPS (and on the SiRF III) There's also a 1second lag between position being estimated and reported. That may also be biting you.

I feel you have to augment with other sensors to get improved position estimation accuracy/precision.

Replies

Barrett Ames July 1, 2012 at 8:53pm

Think you could post the code somewhere so we could take a look at it?
Bot Thoughts June 25, 2012 at 10:13am

Relying on GPS alone is dicey. A good GPS can probably pull off 5m accuracy or so. 60-70 feet is about 20m which is pretty bad, yes.

You can do several things to improve performance.

One thing you can do to improve reception is to throw a ground plane under the module. Just a square of tin or something, about 4-5" on a side. If you are only locking into 3 or 4 sats, your position error will be awful. What is the HDOP values you are getting? Anything above about 1.2 means very large position error. Even at HDOP=1.2 or lower still has some pretty big error (5m is what I'm usually seeing on the two gps units I use)

Another tip: don't measure waypoints with the GPS itself because you've got error in your waypoint AND error in your current position. The error could cancel itself out or it could double your error or anything in between. Instead, if you use something like Google Earth to pick your waypoints you reduce one source of error. To address Google Earth error, calibrate your vehicle to a known landmark / known lat/lon pair in Google Earth.

I'm extremely pleased with the Venus638FLPX that I put on my rover. With the ground plane I get really good signals even in difficult areas. Position is still dicey. I've also used a SiRFIII (Paros iGPS-500) that seemed to do pretty good, but not as good as the Venus.

The downside with the latter is massive amounts of filtering, even with the high dynamic range firmware. So heading signal bandwidth is very limited; you turn sharp, the gps takes a couple seconds to catch up. I'm no sure all the filtering helped position estimation all that much either.

On the Venus GPS (and on the SiRF III) There's also a 1second lag between position being estimated and reported. That may also be biting you.

I feel you have to augment with other sensors to get improved position estimation accuracy/precision.

This reply was deleted.