Open frede791 opened 8 months ago
Is this different from the bias_mean and bias_stddev parameters?
Yes, as far as I understand, the noise model that is employed is slightly different to the old one that was used previously.
Could you elaborate more? Based on the lines of code you linked to, it seems to me that we can set bias_mean=0
and bias_stddev=<turn_on_bias_sigma>
to get the same results. cc @mjcarroll
Based on the lines of code you linked to, it seems to me that we can set bias_mean=0 and bias_stddev=1.0 to get the same results
Yes, this should accomplish the same results. With the IMU, we model both a static (turn-on) bias as well as a dynamic (time varying) bias.
The turn on bias is drawn from a gaussian distribution modelled with mean bias_mean
and stddev bias_stddev
when the sensor is loaded.
Desired behavior
The IMU plugin in old gazebo used a factor called
turn_on_bias_sigma
factor for modelling noise. https://github.com/PX4/PX4-SITL_gazebo-classic/blob/33ac87a37676fb597de110e926bbf0a080169ffe/src/gazebo_imu_plugin.cpp#L173C3-L180C4 As far as I can tell there is currently no way to model this in any way? Would it be possible to add this as a specifiable tag that can be used in the noise model?Implementation suggestion
Add an additional tag that can be read out in
GaussianNoiseModel.cc
and added as a factor in theLoad()
function