Detail about radar data

Hello guys,

My question is

  1. In the radar format description, it says v_comp was compensated by ego-motion. What does the ego-motion mean here? Is it the ego-motion caculated by only radar? or ego-motion provided by ground truth GPS/INS.

  2. The technical documentation of Continental ARS40X radar shows that its output mode can be set by “RadarCfg_OutputType” parameter, so what is output mode you set in your dataset, cluster mode or object mode?

1 Like

Hi, sorry for the delayed response.

  1. Yes, the ego motion comes from our lidar-based localization algorithm (what’s available as ego_pose in nuScenes).

  2. We’re receiving the raw clusters.

Hello @holger-motional, thank you for the reply. I notice there are both radial and angular velocity information in v (before compensated) but I believe Continental radar can only provide radial velocity of clusters. Did you do any tracking process to caculate angular velocity of clusters?

Sorry, I missed the last question. I don’t think we provide radial velocity:
In any case, what we provide is directly the output of the radar.


What are the vx_rms and vy_rms values in the RADAR data? Can you elaborate.?

The definition according to the manual is “Standard deviation of longitudinal relative velocity” and “Standard deviation of lateral relative velocity”. I don’t know what it is the standard deviation of. Perhaps the radar is taking multiple measurements?

As I know, the radar receives measurements from multiple reflectors on the object and it clusters them to generate a Radar point measurement. Those vx_rms and vy_rms are the standard deviations of the velocity measurements in a cluster.

1 Like

The ideal scene is just like what your say, and i agree the meaning of “rms” which is the standard deviations of the velocity measurements in a cluster.

But i have a question here: maybe in most situation the radar can only get one reflectors. So i don’t know how the cluster it? And it makes no sense to calaulate the standard deviation.