Get all online MCUs.

Open a CAN socket on the specified interface.

Optimize the pose graph using Gauss-Newton.

Outlet area (mm²) - rectangular

Check if this bounding box overlaps with another.

Parse a serial command and update state

Parse incoming data (auto-detects NMEA, RTCM3, or UBX)

Parse a GGA NMEA sentence

Parse a GSA NMEA sentence

Parse an IMU data packet. SDK2 IMU packet format (60 bytes total): Bytes 0: version Bytes 1-2: length (LE) Bytes 3-4: time_interval Bytes 5-6: data_n

Parse an NMEA sentence and update state if relevant.

Parse RMC sentence (Recommended Minimum Navigation Information). Format: $GPRMC,hhmmss.ss,A,llll.ll,a,yyyyy.yy,a,x.x,x.x,ddmmyy,x.x,a*hh

Parse spherical point data (type 3). Each point: depth(u32), theta(u16), phi(u16), reflectivity(u8), tag(u8) = 10 bytes

Get the peer count handle (for auto-sleep suppression in the control loop).

Get reference to physics.

Get the current planned path waypoints (from A* planner). Returns the path waypoints if a path is currently planned, along with the waypoint index be

Periodically sends COMM_GET_VALUES to the gateway VESC at ~20 Hz.

Predict step: propagate state with body-frame odometry delta (dx, dy, dθ).

Full preprocessing pipeline for point cloud. Args: points: Raw point cloud voxel_size: Voxel size for downsampling

Print a formatted evaluation report to stdout.

Get probability at grid cell (convenience wrapper).

Try to probe a CSI camera by sensor ID.

* Process any pending blob URL revocations that are ready.

Process a RecordBatch and extract relevant data.

Process pending dispatch events from the receiver

Process a newly discovered session

Process an RRD session file by extracting its contents. Extracts poses, LiDAR, camera, and GPS data from the .rrd file and writes them to an extracte

Process a single session

Progress percentage (0-100).

Calculate pulse brightness (0-255) based on elapsed time and period. Uses a sine-like curve for smooth breathing.

Convert quaternion [qx, qy, qz, qw] to 3x3 rotation matrix.

Send a parameter query and return parsed key-value pairs.

Internal raw RGB capture loop.

Reconstruct path from came_from map.

Record a successful CAN transmission

Refresh system metrics (call before reading).

Regenerate world with new random obstacles.

Register a blower tool from config.

Get a reference to the wide-tolerance relocalization scan matcher.

Remove statistical outliers based on mean distance to k nearest neighbors. Points with mean distance greater than (global_mean + std_rat

Reset the filter to uninitialized state.

Reset the estimator to origin.

Reset warm-start (call when switching to a new path).

Resolve camera mounts: use `mounts` array if present, otherwise build a single-entry list from legacy fields.

RLE decode: [count:u8][value:u8] pairs → flat Uint8Array.

Benchmark environment performance.

Run Gaussian splatting training. This is a simplified implementation. For production, you would: 1. Use COLMAP or similar for proper

Run episodes with a simple heuristic policy.

Run database migrations manually

Run episodes with random actions.

Internal reader loop.

Run the LiDAR reader (point cloud only).

Run the LiDAR reader with runtime start/stop control. This version keeps a control socket open and responds to Start/Stop commands to control the lid

Run a single rover simulation.

Run the serial reader loop

Run the Gaussian splatting pipeline. Queues a job for the splat-worker and optionally waits for completion.

Train using REINFORCE (simple policy gradient).