Save skipped sessions set to disk

Generate a LiDAR scan from the given pose.

Get a reference to the scan context database.

Scan sessions directory for .rrd files that haven't been extracted yet. Returns paths to .rrd files that need processing.

Scan sessions directory for all session directories

Background loop that evaluates cron-scheduled missions and auto-starts them.

Score all trajectories using 5 critics.

Send SLCAN error response

Helper to send e-stop. Uses full 8-byte header format required by parse_command.

Send a single u16 parameter and drain the ACK.

Send serial number response

Helper to send SetMode command. Uses full 8-byte header format required by parse_command.

Helper to send commands from operator station (for testing/CLI). Uses full 8-byte header format required by parse_command.

Send version response

Get the current session file path.

Set camera info (sent once per connection on the video channel). The camera_id, width, and height will be filled from the first video frame.

Set color order

Set the costmap receiver for streaming costmap data to the console.

Set the pre-integrated IMU delta for the next scan. Call this before `process_scan()` with the consumed delta from the preintegrator.

Set the keyframe requestor for PLI handling. Set the keyframe request flag for PLI handling. When set to `true` by the RTCP reader, the camera capture

Set the LED mode immediately without transition.

Set the obstacles receiver for streaming obstacle data to the console.

Set the path receiver for streaming navigation path data to the console.

Set position directly (for spawning/resetting).

Set role status bytes [16-23].

Set a new dispatch task.

Set LED timing profile

Check if we should add a new keyframe.

Check if status should be sent (~50Hz).

Check if status should be sent (~50Hz).

Simulate differential drive kinematics for one timestep Returns the new pose after applying velocities for duration dt.

Forward-simulate all trajectories using diff-drive kinematics.

Create an owned snapshot of the current costmap state for streaming.

Spawn the LiDAR reader as an async task. Returns a handle that can be used to abort the task.

Spawn the connection task

Spawn the depth processing thread. Returns: - `mpsc::Sender<TaggedFrame>` — send raw frames here (non-blocking, bounded) - `watch::Receiver<DepthOutp

Spawn an H.264 camera capture thread that sends frames to a channel. Returns a receiver for H.264 frames and a thread handle. Uses a bounded channel

Spawn robot at a random valid position.

Spawn a raw RGB camera capture thread. Returns a receiver for raw frames and a thread handle. Frames are uncompressed RGB suitable for neural network

Spawn with IMU and runtime control. The command receiver allows starting/stopping sampling at runtime. When stopped, the lidar motor stops spinning (

Spawn with IMU data channel.

Split a large point set into 96-point Livox packets.

Run startup animation (blocking)

Get all VESC states.

Sync GLB files needed by the web viewer to web/models/

Background loop that detects and fails orphaned tasks. Tasks are considered orphaned if: - Status is 'assigned' or 'active' AND - The rover is not cu

Get task progress for reporting.

Advance to next animation frame (call at ~30fps or so)

Serialize to wire format for TCP CAN bridge. Format: flags(1) + id(4 LE) + len(1) + data(0-8) = 6-14 bytes flags bit 0: extended ID

Format frame as SLCAN string (without trailing \r)

Convert to Twist with given velocity limits.

Tool type as protocol u8 value.

Get top lid for individual export

Get filtered total system current (smoothed to reduce noise).

Train Gaussian splats using gsplat. This is a simplified training loop. Production code would use nerfstudio or a more complete impl

Transform points from sensor frame to world frame using pose. Args: points: (N, 3) array of points in sensor frame pose:

Transform a pose from child frame to parent frame.

Try to claim operator status. Returns true if successful.