Rocket Inertial Guidance vs Vision Lab

Flight Profile Preset

Orbital Ascent

Booster Return

Fog & Plume

Optical / Vision Obstacles

Rocket Plume & Soot Distortion 85%

Cloud & Re-entry Ionization Blackout 70%

Inertial & Sensor Fusion

Primary Guidance Algorithm

IMU Gyro Drift Rate 0.05°/hr

GNSS / Radar Altimeter Aiding Enabled

Flight Time (T+) 0.0 s

True Altitude 0.0 km

Velocity 0 m/s

Position Error (Δx) 0.0 m

System preflight ready. Triply-redundant ring-laser gyros online.

Sensor Health Matrix

INS / RLG: Nominal (500 Hz)

Kalman Filter Covariance: < 0.04m²

Optical Vision: Obscured / Blinded

TVC Gimbal: Active (±5.2°)

Inertial Navigation (INS)
0.42 m drift
Self-contained accelerometers & fiber-optic / ring laser gyroscopes. Zero dependence on exterior lighting, clouds, or rocket exhaust plumes.

Optical Computer Vision

78% Blinded

Severely degraded by 1,700°C supersonic exhaust plumes, soot accumulation on lenses, hypersonic shockwaves, and featureless dark water/space.

Landing & Orbit Outcome

Pad Touchdown

Convex optimization calculates thrust vector angles at 50 Hz from pure state estimation, achieving pinpoint landing.

Why Rockets Don't Need (and Often Can't Use) Vision

When Elon Musk noted "Rockets don’t need vision", it highlighted a fundamental difference in physics and control theory between ground autonomy (such as self-driving cars) and rocketry. Ground vehicles navigate dynamic, crowded environments filled with pedestrians, lane lines, and unpredictable drivers. Rockets operate in physical dynamic state space governed by strict Newtonian mechanics and aerospace guidance, navigation, and control (GNC).

1. The Physics Environment is Completely Hostile to Cameras

During liftoff and propulsive landing, rocket engines generate massive fireballs, high-temperature soot, and hypersonic vibration.

Plume Obscuration: Multi-engine kerosene (RP-1) or methane burns deposit thick soot and optical distortion around the base.
Dynamic Range: Sunlight in upper atmosphere or dark ocean barge landings create extreme exposure contrast where optical sensors wash out or under-expose.
Shockwaves & Fog: Transonic vapor cones and re-entry atmospheric ionization blind optical lenses.

2. Mathematical State Estimation (INS + Kalman) is Exact

In rocketry, the primary questions are: What is our exact position vector $\vec{r}$, velocity vector $\vec{v}$, and attitude quaternion $\vec{q}$?

High-Rate IMUs: Ring Laser Gyroscopes (RLGs) and quartz flexure accelerometers sample acceleration and angular rates at 500 Hz to 2,000 Hz with zero latency.
Sensor Fusion: Extended Kalman Filters (EKF) fuse inertial data with radar altimeters, GPS/GNSS, and star trackers for absolute calibration.
G-FOLD & Convex Guidance: The flight computer solves a real-time convex optimization problem every 20ms to gimbal the engine bell directly to the target coordinates without ever "seeing" the landing pad with eyes.