point_cloud_filter_transform_nodes

Table of Contents

• Purpose / Use cases
• Design
  • Distance Filter
  • Angle filter
  • Static Transformer
  • Assumptions / Known limits
  • Inputs / Outputs / API
  • Error detection and handling
  • Security considerations
• Related issues

Purpose / Use cases

To properly separate concerns and make it easier to reuse data (i.e. prerecorded bag data, or from the simulator), we need to separate the transforming and filtering data from the drivers into separate nodes.

In particular, we require a boilerplate wrapper around our algorithm, which is implemented as a library called lidar_utils. This boilerplate node wrapper allows us to communicate with the rest of Autoware.Auto.

Design

There is one instance of these nodes for the PointCloud2 message type for visualization and open source support.

The following components are parts of the lidar_utils package but they are used by the point_cloud_filter_transform node to filter and transform the input point cloud data :

Distance Filter

Filter class to check if a point is contained in a range defined by a minimum and maximum radii from the origin. Operates in the square forms to avoid using sqrt().

Angle filter

Filter class to check if a point lies within a range defined by minimum and maximum angles. The inclusion zone is defined by the range [start angle , end angle] in counter-clockwise direction. To avoid explicitly computing the angle using inverse trigonometric functions, the following scheme is used:

• A vector called range_normal is defined to be the unit vector placed in the median of the range.
• The unit vector at the min or max angle is projected to this range_normal . The projection is now a factor of the cosine of the half of the accepted angle range \({\theta} \) which will be ( \(cos(\frac{\theta}{2})\)). This value is called the threshold.
• If a point is within the range, value of its projection on the range normal should be greater-equal than this threshold value. The filter operates in square form to avoid using sqrt()

Static Transformer

Applies a transform to given points. Uses the Eigen library during computations.

The point_cloud_filter_transform_node_exe is a wrapper around the DistanceFilter, AngleFilter, and StaticTransformer with the following executing order:

1. Check if the points are within the range of DistanceFilter and AngleFilter
2. Transform the points using StaticTransformer
3. Publish the transformed and filtered data in PointCloud2 format

Assumptions / Known limits

The implementation doesn't allow dynamic thresholds for distance and angle filters. It also doesn't support dynamic transformation during runtime. The thresholds and the transform parameters can only be loaded while constructing the nodes.

Inputs / Outputs / API

These nodes have the following basic structure:

Input:

• points_xyzi (Message of PointCloud2 directly from driver nodes or the point_type_adapter node)

Output:

• points_filtered (Message of PointCloud2)

On top of this, the nodes can be configured either programmatically or via parameter file on construction.

Error detection and handling

Most error handling occurs inside rclcpp.

The only explicit error handling that occurs in these nodes is catching errors in the runtime loop.

If such an error is caught, then the node will simply transition to a "failed state", log this fact, and attempt to continue running.

Security considerations

TBD by a security specialist.

Related issues

• #328: Port transform/filter node from Apex.Auto to Autoware.Auto; Remove filter/transform functionality from drivers