Sign of g seems wrong

[pandisv]

I've recently been playing with this fantastic library, and all is well, except for the direction of gravity which seems to be off by a sign. Consider the following motion model:

ini lat (deg),ini lon (deg),ini alt (m),ini vx_body (m/s),ini vy_body (m/s),ini vz_body (m/s),ini yaw (deg),ini pitch (deg),ini roll (deg)
31.9965,120.004,0,0,0,0,90,0,0
command type,yaw (deg),pitch (deg),roll (deg),vx_body (m/s),vy_body (m/s),vz_body (m/s),command duration (s),GPS visibility
1,0,0,0,0,0,0,20,1
5,0,0,0,4,0,0,20,1
3,90,0,0,0,0,0,8,1
5,0,0,0,0,0,0,20,1

The object is initially stationary, then moves in a straight line towards the east and after a while makes a 90 degree turn south, keeping its altitude steady throughout the motion. While it moves in a straight line, the x_body axis points east, y_body points south, and z_body points down. When I run the simulation, I get the following results:

• In ref_gyro.csv, I see that the gyro readings are initially zero (for all intents and purposes), and then gyro_z becomes positive, as expected from the prescribed trajectory.
• In ref_accel.csv, I initially see a positive acceleration along x_body, which is then followed by a positive acceleration along y_body, both of which are exactly as expected. However, the acceleration in the z direction is negative (-9.79ms^-2). Maybe my understanding is wrong, but with the z_body axis pointing downwards, I would expect the acceleration from gravity to be positive.

This is with the latest version in master. I'm using demo_no_algo.py with ref_frame=0 in the call to Sim().

Any help with this would be greatly appreciated. I'd also like to take this opportunity to thank you for your work on this very useful library.

[dxg-aceinna]

Hi pandisv, your question regarding the sign of the Z acceleration is pretty common.

As we know, an accelerometer measures non-gravitational acceleration.
f = a - g
f is the output of the accelerometer;
a is the linear acceleration;
g is the gravity.

When an accelerometer is on a horizontal plan and stationary, a is zero: [0; 0; 0].
In the NED (north-east-down) frame, g = [0; 0; 9.8]. (It is positive here.)
Considering only the Z accelerometer, you can see f_z = -g_z = -9.8m/s^2.

[pandisv]

Thanks for your answer. I agree with your comment -- the equation is really f=R(a-g) in your notation, where R is some rotation matrix aligning the body frame to the NED plane, but I still agree with your conclusion that f_z = -9.8m/s^2. That's the point of my bug report -- if you run a simulation of the above turn, the accelerometer reports f_z = +9.8m/s^2, which is wrong, it should be negative. This is what ref_accel.csv looks like:

0.000000000000000000e+00,-1.199521838941635691e-15,9.794839130570458963e+00

[dxg-aceinna]

Hi, I tried your motion profile. I saw the reproted f_z is around -9.8, which is as expected.

[pandisv]

Thanks for the help and feedback. You are indeed correct, it was I who got the signs wrong. Thanks again for all the work you've put into this very useful library!