Ground truthEstimated poseInflated obstaclesGoal
Follow a C++20 autonomy stack from noisy position measurements to closed-loop navigation. Compare nominal and sparse position updates on the same cluttered course.
Switch between position update intervals of 0.5 and 10 seconds. Scrub the mission to compare the true trajectory with the EKF estimate; watch the position error as the vehicle turns through the map.
The original C++ implementation runs EKF fusion, A* global planning, local lattice rollouts and pure pursuit. The browser visualizes exported traces; it does not substitute a JavaScript planner.
Both runs use deterministic synthetic sensor noise and the same fixed map. Position updates are less frequent in the sparse scenario, not completely disabled. The dashed route is the final replan, not the route active at every frame. LiDAR is raycast by the simulator but does not update the EKF. This is playback, not a live simulation.