[Documentation] [TitleIndex] [WordIndex


The self_see_filter inherits from the FilterBase class, and implements the functionality of removing or marking point cloud points that belong to a robot's body. There are two modes of operation - the first actually filters points out, discarding those points in the cloud that are determined to be in or near robot links or which are shadowed by the robot links. The second mode is not to remove points that lie on or near robot links but instead to mark them using an annotation channel in the point cloud. The self_see_filter additionally allows a sensor frame to be set to enable filtering of points that lie behind the robot's body from the view of the sensor. The code in this package works for links defined as spheres, boxes, cylinders or meshes.

API Stability

The filter does not directly subscribe to or publish any values other than tf transforms, but does use a number of parameters to control internal operation. It will attempt to set the parameters from a node handle passed into to the contructor, so all parameters will be drawn from the namespace <namespace> used to create the node handle.

~<namespace>/min_sensor_dist (double, default: ".02")

~<namespace>/self_see_default_padding (double, default: .01) ~<namespace>/self_see_default_scale (double, default: 1.0) ~<namespace>/annotate (string, default: "") ~<namespace>/self_see_links (XmlRpc::XmlRpcValue::TypeArray)

The entries for self_see_links are best explained in yaml format:

 - name: base_link
   padding: .03
   scale: 1.0

The name (required) specifies the name of the link to be considered in the self filter, the padding (optional) the distance (in meters) from the link to include in collision filtering, and the scale (optional) to consider the link. If the padding or scale are not set the defaults will be used.

Finally, C++ calls exist to set the sensor frame for shadow detection if desired.

Technical details

To determine whether a point is inside, outside or a shadow, the following computation is performed for every link we are testing against:

2020-04-04 13:13