tam_state_estimation_node¶

ROS 2 wrapper around the pure-C++ tam_state_estimation library. It subscribes to the sensor streams published by the vehicle interface, delay-compensates and converts them, feeds them into the (ROS-free) estimator, and republishes the fused vehicle state under /core/state.

Everything ROS-specific lives in this package — message conversion, delay compensation, the topic watchdog, TF broadcasting and the node monitor — while every line of estimation logic stays in the library. The node itself is a class template stateEstimationNode<TConfig, VehicleModel> (node.hpp); each shipped variant is one instantiation of it.

High level state estimation concept

Variants¶

The node is a composable rclcpp component. One variant is registered per (filter dimension × wheelspeed model); each is available both as a loadable component (plugin) and as a standalone executable. Pick the one matching your vehicle and sensor setup.

Component (plugin)	Executable	Filter	Wheelspeed model
`tam_state_estimation_node::StateEstimation2DEKFKin`	`tam_state_estimation_node_ekf_2d_kin`	2D EKF	Kinematic
`tam_state_estimation_node::StateEstimation2DEKFNh`	`tam_state_estimation_node_ekf_2d_nh`	2D EKF	Non-holonomic
`tam_state_estimation_node::StateEstimation3DEKFKin`	`tam_state_estimation_node_ekf_3d_kin`	3D EKF	Kinematic
`tam_state_estimation_node::StateEstimation3DEKFNh`	`tam_state_estimation_node_ekf_3d_nh`	3D EKF	Non-holonomic
`tam_state_estimation_node::StateEstimation3DEKFStm`	`tam_state_estimation_node_ekf_3d_stm`	3D EKF	Single-track

Kinematic fuses no wheelspeed odometry; non-holonomic fuses it as a linear-velocity measurement (vy = 0); single-track additionally estimates a side-slip angle from the steering angle.

Topics¶

All topics live under the /core/state namespace.

Published¶

Topic	Type	Description
`/core/state/odometry`	`nav_msgs/msg/Odometry`	Fused pose + velocity
`/core/state/acceleration`	`geometry_msgs/msg/AccelWithCovarianceStamped`	Fused acceleration
`/core/state/s`	`std_msgs/msg/Float64`	Arc length along the raceline
`/core/state/s/centerline`	`std_msgs/msg/Float64`	Arc length along the centerline
(TF)	`geometry_msgs/msg/TransformStamped`	`local_cartesian → vehicle_velocity_cg`, `vehicle_cg → vehicle_velocity_cg`

Subscribed¶

Localization / linear-velocity / IMU inputs — not hard-coded; the set of sensor topics is built at start-up from the ROS parameter overrides (measurements.* and inputs.imus.*) by parse_sensor_configs (helper.hpp).
/vehicle/sensor/wheelspeed_radps, /vehicle/sensor/wheelspeed_status — non-kinematic models only.
/vehicle/sensor/steering_report — single-track model only.

Configuration¶

Each sensor is described declaratively via parameters, e.g.:

measurements:
  <name>:
    message: { topic: ..., type: ..., timeout_ms: 500 }
    status:  { topic: ..., timeout_ms: 500 }          # optional
    position_m:          { R_init: [...], outlier_bounds: [...] }   # any subset of
    orientation_rad:     { R_init: [...], outlier_bounds: [...] }   # these fusion
    linear_velocity_mps: { R_init: [...], outlier_bounds: [...] }   # groups
inputs:
  imus:
    <name>:
      message: { topic: ..., type: ... }
      filter_coefficients: [...]   # FIR pre-filter (parameter key: imu<N>_filter_coefficients)
      backup: false

Running¶

Standalone executable:

ros2 run tam_state_estimation_node tam_state_estimation_node_ekf_3d_nh \
  --ros-args --params-file <your_config>.yml

Or load the component into a running container:

ros2 component load /<container> tam_state_estimation_node \
  tam_state_estimation_node::StateEstimation3DEKFNh \
  --ros-args --params-file <your_config>.yml