control_evaluator

ROS 2 Python package with a node named control_evaluator for control-performance evaluation.

The node reads global.adapter from config/global/global_config.yaml and switches topics per environment.

Subscriptions

• Always:
  • /path_planning/path (custom_interfaces/msg/PathPointArray)
  • /state_estimation/lap_counter (std_msgs/msg/Float64)
• If adapter is vehicle:
  • /state_estimation/pose (custom_interfaces/msg/Pose)
  • /state_estimation/velocities (custom_interfaces/msg/Velocities)
• If adapter is pacsim:
  • /pacsim/pose (geometry_msgs/msg/TwistWithCovarianceStamped)
  • /pacsim/velocity (geometry_msgs/msg/TwistWithCovarianceStamped)

Metrics start condition

Metrics start only after the first control command is observed.

• If adapter is vehicle: first message on /control/command (custom_interfaces/msg/ControlCommand)
• If adapter is pacsim: first message on either /pacsim/steering_setpoint (pacsim/msg/StampedScalar) or /pacsim/throttle_setpoint (pacsim/msg/Wheels)

Published metrics (prefix: /control_evaluator)

For each instant, average, mean squared error, and root mean squared error:

• distance_error/{instant,average,mse,rmse}
• velocity_error/{instant,average,signed_average,mse,rmse}
• heading_error/{instant,average,signed_average,mse,rmse}

Notes:

• Distance is computed to the interpolated line segment between the closest path point and its neighboring segment.
• Heading and reference velocity are computed from linear interpolation on that selected segment.
• average for error metrics is the mean absolute error (MAE).
• signed_average is the signed mean (bias).

Additional metrics:

• average_velocity
• average_lap_time

Run

colcon build --packages-select control_evaluator
source install/setup.bash
ros2 launch control_evaluator control_evaluator.launch.py