Types

Browse the API reference by type.
Kind Name Description
struct SceneInputs Selected input file paths.
struct PoseSample One sample from the reference pose stream.
struct ImuSample One sample from the input IMU stream.
struct WheelSpeedSample One wheel-speed CAN sample from zoe_veh_info.
struct LoadedScene Loaded reference-pose, IMU, and wheel-speed streams.
struct GpsSample One synthetic 2D GPS sample.
struct StartupInitialization Initialization data required to start the filter.
struct StartupTraceTruthState Truth values from the pose stream at one startup frame endpoint.
struct StartupTraceFrame One startup-prefix trace frame.
struct StartupTraceResult Full startup trace and final handoff result.
struct GpsUpdateResult Diagnostics from one 2D GPS update.
struct EskfLogRow One in-memory log row corresponding to one GPS update.
struct EskfRunResult In-memory result from a full post-startup ESKF run.
class Eskf Error-state Kalman filter over position, velocity, and attitude.

SceneInputs

Selected input file paths.

Signature

source 
struct SceneInputs

Fields

  • pose_path: Pose file path.
  • imu_path: IMU file path.
  • wheel_speed_path: Wheel-speed file path.

PoseSample

One sample from the reference pose stream.

Signature

source 
struct PoseSample

Fields

  • utime: Timestamp in microseconds as std::int64_t.
  • pos: Reference global position in meters.
  • orientation: Reference orientation quaternion in [w, x, y, z].
  • vel: Reference ego-frame velocity sample in m/s.

ImuSample

One sample from the input IMU stream.

Signature

source 
struct ImuSample

Fields

  • utime: Timestamp in microseconds.
  • specific_force: Measured specific force in frame I.
  • rotation_rate: Measured angular velocity in frame I.
  • q_AI: Quaternion from frame I to gravity-aligned frame A.

WheelSpeedSample

One wheel-speed CAN sample from zoe_veh_info.

Signature

source 
struct WheelSpeedSample

Fields

  • utime: Timestamp in microseconds.
  • fl_wheel_speed_rpm: Front-left wheel speed in rpm.
  • fr_wheel_speed_rpm: Front-right wheel speed in rpm.
  • rl_wheel_speed_rpm: Rear-left wheel speed in rpm.
  • rr_wheel_speed_rpm: Rear-right wheel speed in rpm.

LoadedScene

Loaded reference-pose, IMU, and wheel-speed streams.

Signature

source 
struct LoadedScene

Fields

  • pose_samples: Loaded pose samples.
  • imu_samples: Loaded IMU samples.
  • wheel_speed_samples: Loaded wheel-speed samples.

GpsSample

One synthetic 2D GPS sample.

Signature

source 
struct GpsSample

Fields

  • utime: Timestamp in microseconds as std::int64_t.
  • xy: Measured horizontal position in meters.

StartupInitialization

Initialization data required to start the filter.

Signature

source 
struct StartupInitialization

Fields

  • v0_G: Initial velocity in frame G at previous_imu_utime.
  • p0_G: Initial position in frame G at previous_imu_utime.
  • q0_GI: Initial attitude quaternion from the startup state’s IMU frame to G at previous_imu_utime.
  • first_unprocessed_imu_index: First IMU sample to pass to Eskf::Predict after initialization.
  • first_unprocessed_gps_index: First GPS sample not used by initialization.
  • previous_imu_utime: IMU timestamp of the stored state used as t_{k-1} by the first post-initialization Predict.

StartupTraceTruthState

Truth values from the pose stream at one startup frame endpoint.

Signature

source 
struct StartupTraceTruthState

Fields

  • position_xy: Truth position in meters.
  • yaw_rad: Truth yaw in radians.
  • speed_mps: Truth speed magnitude in m/s.

StartupTraceFrame

One startup-prefix trace frame.

Signature

source 
struct StartupTraceFrame

Fields

  • fit_end_index: Inclusive GPS end index for this frame.
  • gps_points_used: Number of GPS points in the prefix.
  • end_utime: Timestamp at the frame endpoint.
  • latest_gps_xy: Latest GPS point in the prefix.
  • fitted_endpoint_xy: Fitted path endpoint position.
  • fitted_velocity_xy: Fitted path endpoint velocity.
  • fitted_speed_mps: Speed from the fitted endpoint tangent.
  • fitted_yaw_rad: Yaw from the fitted endpoint tangent.
  • selected_yaw_rad: Final startup yaw choice.
  • global_yaw_rad: Global yaw candidate.
  • pca_yaw_rad: Principal-direction yaw.
  • projected_separation_m: PCA projected start-to-end span.
  • line_like: Whether the PCA cloud is line-like.
  • wheel_speed_mps: Smoothed robust wheel-speed magnitude.
  • wheel_supported_travel_m: Wheel-supported travel from startup begin to this frame.
  • required_wheel_supported_travel_m: Active wheel-supported travel gate.
  • path_xy: Sampled fitted path points.
  • truth: Truth at this frame endpoint.

StartupTraceResult

Full startup trace and final handoff result.

Signature

source 
struct StartupTraceResult

Fields

  • first_usable_gps_index: First GPS index startup may use.
  • frames: One frame per tested startup prefix.
  • ready_frame_index: Index of the first ready frame, if any.
  • startup_initialization: Final handoff state, if startup succeeds.

GpsUpdateResult

Diagnostics from one 2D GPS update.

Signature

source 
struct GpsUpdateResult

Fields

  • innovation_xy: Horizontal innovation in meters.
  • nis: Normalized innovation squared.

EskfLogRow

One in-memory log row corresponding to one GPS update.

Signature

source 
struct EskfLogRow

Fields

  • utime: GPS update timestamp in microseconds.
  • estimated_position_G: Estimated position after the GPS update.
  • gps_xy: GPS measurement used in the update.
  • true_xy: Pose-truth horizontal position at the same timestamp.
  • innovation_xy: Horizontal innovation returned by the GPS update.
  • nis: Normalized innovation squared for the update.
  • estimator_error_xy: Horizontal estimator error relative to pose truth.
  • estimator_error_norm: Norm of estimator_error_xy.

EskfRunResult

In-memory result from a full post-startup ESKF run.

Signature

source 
struct EskfRunResult

Fields

  • log_rows: One row per GPS update after startup.
  • num_gps_updates: Number of processed GPS updates.
  • raw_gps_rmse_xy: Raw GPS horizontal RMSE.
  • eskf_rmse_xy: ESKF horizontal RMSE.

Eskf

Error-state Kalman filter over position, velocity, and attitude.

Signature

source 
class Eskf

Methods

Initialize()

source 
void Initialize(const StartupInitialization& startup_initialization);

Initialize the filter from the startup initialization.

Non-obvious behavior

This seeds the nominal state and covariance for the first live IMU/GPS updates after startup.

Inputs
  • startup_initialization: Startup state used to begin filtering.

Predict()

source 
void Predict(const ImuSample& current_imu_sample);

Propagate the filter from the previous IMU sample to the current one.

Inputs
  • current_imu_sample: Current IMU sample k used for propagation from t_{k-1} to t_k.
Preconditions

Initialize() has been called.

Throws

std::runtime_error If the filter is not initialized, or if the state covariance becomes numerically invalid during propagation.

UpdateGps()

source 
GpsUpdateResult UpdateGps(const GpsSample& gps_sample);

Update the filter with one 2D GPS position measurement.

Inputs
  • gps_sample: GPS measurement at the current update time.
Output

Innovation and NIS from this GPS update.

Preconditions

Initialize() has been called.

Throws

std::runtime_error If the filter is not initialized, if the innovation covariance cannot be factorized, or if the state covariance becomes numerically invalid during the update.

Accessors

Signature Description Output Preconditions Throws
const Eigen::Vector3d& PositionG() const Get the current nominal position in frame G. Current nominal position in frame G. Initialize() has been called. std::runtime_error If the filter is not initialized.