Repository navigation
12. Robot Radar Map
The system models a perception stack using a robot-relative radar map representation (RobotRadarMap).
The RobotRadarMap is obtained by transforming the global RadarMap into the robot-centric frame using the robot’s current pose. The transformation applies a translation followed by a rotation aligned to the nearest cardinal direction, in order to reduce rotational variance and discretize orientation.
The system defines the following coordinate frames:
World frame (W) A global, stationary reference frame in which the RadarMap is initially defined. It represents a consistent environmental coordinate system independent of the robot’s motion.
Robot frame (B) A robot-centric coordinate frame centered at the robot’s current position and orientation. The RobotRadarMap is expressed in this frame after transforming from the world frame.
The transformation pipeline follows the sequence:
World frame (W) → Robot frame (B)
The transformation assumes that the robot pose is accurately estimated and available in real time. Errors or noise in pose estimation may directly affect the stability of the RobotRadarMap representation.
The alignment to the nearest cardinal direction introduces a discretization of orientation, which reduces rotational variance but also limits angular precision. This trade-off may impact applications requiring fine-grained orientation sensitivity.
The model assumes a consistent definition of the RadarMap in a global frame. Any inconsistency in the reference frame definition may lead to unstable transformations.
Release: 0.26.0