How to get velocity vector(NED) from GPS NMEA data?

[scott198510]

Mr Doc,
The GNSS speed in your code is obtained directly. As we all know, the original GNSS speed value is only one，such as NMEA format.
Then how to use the original GNSS data which is NMEA format to get the velocity in three directions?

Anyway, at present, the published papers or books directly obtain three values for GNSS velocity, but there is only one value in original GNSS data. Why not explain this? Can we say that the default is to use the data which is dfferential data(such as RTK)?

https://stackoverflow.com/questions/55554306/how-to-get-ned-velocity-from-gps

http://aprs.gids.nl/nmea/

[wangfeng35]

IN NEMA xxVTG and xxRMC, the GNSS speed is the 2D Speed Over Ground ignoring the altitude. In order to get the 3D velocity components, you need to use the GNSS receiver that supports to output the 3D velocity in its own protocol. For example, here: https://portal.u-blox.com/s/question/0D52p00008HKDe8CAH/2d-or-3d-gps-speed .

[scott198510]

Anyway, in fact, there are NMEA format data that are either solved or used, but most of the literatures/papers/books do not mention it. By default, the receiver is used to directly obtain three effective velocity values. Right?

[rodralez]

In fact, I have considered this method earlier. For the two velocities in the horizontal plane, we adopt the decomposition method, and the the vertical one, we adopt the altitude data difference. But in fact, I want to reproduce the three velocities of GNSS receiver, any other methods?

The GNSS sensor estimates positions and velocities using different methods. Velocities are obtained by Doppler shifts or carrier phase measurements [1]. This is beneficial since GNSS positions and velocities tend to be uncorrelated, one of the core assumptions in Kalman filter implementation.

When Vn and Ve are estimated from GNSS speed-over-ground velocity and GNSS course-over-ground angle, I am not sure if these estimations are uncorrelated from positions. On the other hand, Vd obtained from the derivative of altitude will be correlated to altitude for sure. So, this is a sub-optimal approach.

In summary, the best approach is to get a GNSS sensor that directly delivers Vn, Ve and Vd.

[1] https://insidegnss.com/wp-content/uploads/2018/01/marapr15-SOLUTIONS.pdf

[rodralez]

Besides, you can also use the position-only update with GPS position vector (lat/lon/alt) in NaveGo. In this way, you need to modify the measurement vector to GPS position vector and also H matrix. However, the performance may be degraded.

Yes, according to my experience, a loosely-coupled INS/GNSS system with position-only updates is useless, at least using low-cost IMU sensors.

[wangfeng35]

yway, in fact, there are NMEA format data that are either solved or used, but most of the literatures/papers/books do not mention it. By default, the receiver is used to directly obtain three effective velocity values. Right?

Yes, if the velocity update is used, the receiver they used can directly output the three velocity components.

[wangfeng35]

Besides, you can also use the position-only update with GPS position vector (lat/lon/alt) in NaveGo. In this way, you need to modify the measurement vector to GPS position vector and also H matrix. However, the performance may be degraded.

Yes, according to my experience, a loosely-coupled INS/GNSS system with position-only updates is useless, at least using low-cost IMU sensors.

You're right. The best way is to use both position and velocity update when a low-cost IMU is used.