search

ros-simulation / gazebo_ros_pkgs

Wrappers, tools and additional API's for using ROS with Gazebo
http://wiki.ros.org/gazebo_ros_pkgs
774 stars 772 forks source link

Separate Gaussian noise parameters for accel/gyro/magnetometer #1255

Open jonatanolofsson opened 3 years ago

jonatanolofsson commented 3 years ago

Description

The Gaussian noise parameter means vastly different things for each sensor type, and they often differ in several orders of magnitude. It makes no physical sense to use the gaussianNoise parameter for all data types.

I suggest adding three new Vector3 parameters in the SDF (with the gaussianNoise left for backwards compatibility) to control the variance separately for each data type.

Expected behavior

Sensor variance is specified per sensor.

shonigmann commented 3 years ago

Adding to this, SDF 1.4 onward offers at least per axis, per sensor (except magnetometer/orientation, perplexingly) specifications for mean, standard deviation, bias mean, and bias standard deviation. SDF 1.5 onward adds noise precision. And SDF 1.6 onwards offers dynamic bias parameters for time correlated noise, which are certainly all relevant. But it doesn't seem like any of these are taken into account by the sensor plugin (unless I'm mistaken?). In any case, I think better noise modelling would be a very welcome addition to the community :)