Posted by HugePanic on September 11, 2008 at 1:54am
I need some advice from people with fist hand experiences.How do GPS Errors occur??Is the just one single sataset send from the GPS-Sensor, or are there "many" datasets for a specific time that are wrong??What is the best way to find misleading datasets??Collect three sets and omit the "worst" one compared with the average???To be honest:I don't want to search in puplished code for that feature. I will search when i am done with my own programm to compare.I don't want to use too much copy and paste.ThxP.S.: of course will my code puplished when finished
You need to be a member of diydrones to add comments!
By no means am I a GPS expert but I can relate a few sources of errors you'll likely see at some point or another.
First, the set of satellites in view at any give point in time can change so the GPS receiver may transition from one set of satellites to another. Each satellite introduces a little different error in the signal, which is why you often see that most receivers are only accurate to 10/5/2.5 meters depending on the signal source. As the receiver transitions from one set to a slightly different set, you might see a "jump" in the position that is either singular or pops over to a different position and remains in that area.
Obviously, you don't want the autopilot responding drastically to a sudden, significant change but straight forward filters, such as averages, cannot handle pops well though they will fix one time errors. On the down side, an averaging filter adds latency to the position information because it is using multiple position estimates to produce its information. For slow vehicles, that isn't a problem but for something moving quickly it isn't optimal.
Another source of error is a sudden reduction in the number of satellites in view. This can occur due to natural terrain features obscuring line of site to satellites that are low on the horizon. If the count is reduced to less than 5 satellites, the altitude fix of the receiver is the first to deteriorate. One of the GPS sentences indicates the type of fix the receiver has either in 2D or 3D so you can utilize it to determine if the altitude value is meaningful. I cannot recall offhand which sentence it is in but its a standard NMEA output.
There are numerous websites and papers in the robotics literature that discuss GPS error and position estimating. You might want to search those. Many will be 2D in nature for ground vehicles but they are more likely to detail exactly what types of errors are present inherently in the data.
Most error reduction schemes use a Kalman filter using GPS along with IMU data to produce a location result. Kalman filtering is out of my area of knowledge but they are ideal for position calculations in unmanned system.
Most of my "GPS errors" were actually serial errors (lost characters) until we got our code right (and then ultimately switched to SiRF binary). So include that in your list of things that can go wrong.
Replies
First, the set of satellites in view at any give point in time can change so the GPS receiver may transition from one set of satellites to another. Each satellite introduces a little different error in the signal, which is why you often see that most receivers are only accurate to 10/5/2.5 meters depending on the signal source. As the receiver transitions from one set to a slightly different set, you might see a "jump" in the position that is either singular or pops over to a different position and remains in that area.
Obviously, you don't want the autopilot responding drastically to a sudden, significant change but straight forward filters, such as averages, cannot handle pops well though they will fix one time errors. On the down side, an averaging filter adds latency to the position information because it is using multiple position estimates to produce its information. For slow vehicles, that isn't a problem but for something moving quickly it isn't optimal.
Another source of error is a sudden reduction in the number of satellites in view. This can occur due to natural terrain features obscuring line of site to satellites that are low on the horizon. If the count is reduced to less than 5 satellites, the altitude fix of the receiver is the first to deteriorate. One of the GPS sentences indicates the type of fix the receiver has either in 2D or 3D so you can utilize it to determine if the altitude value is meaningful. I cannot recall offhand which sentence it is in but its a standard NMEA output.
There are numerous websites and papers in the robotics literature that discuss GPS error and position estimating. You might want to search those. Many will be 2D in nature for ground vehicles but they are more likely to detail exactly what types of errors are present inherently in the data.
Most error reduction schemes use a Kalman filter using GPS along with IMU data to produce a location result. Kalman filtering is out of my area of knowledge but they are ideal for position calculations in unmanned system.