Aaron Belinski

Filter demo

A small version of what I work on at RPL. A simulated rocket flies, a noisy barometer and accelerometer measure it, and a Kalman filter combines the two into one estimate of where it is.

Hit Gust to shove the rocket, or Baro glitch to feed the filter a few garbage readings and watch it throw them out. The slider makes the barometer noisier.

Time
0.0 s
Phase
Boost
Altitude
0 m
Error
+0.0 m
2σ bound
0.0 m
The filter tracks altitude, velocity, and acceleration. It reads the barometer at 20 Hz and the accelerometer at 100 Hz, and it ignores any barometer reading that lands more than 4σ from where it expects. The bottom plot shows how far off each raw barometer reading is next to how far off the estimate is. I wrote this for the page in TypeScript. It isn’t the team’s flight code, and the rocket is a generic one, not an RPL vehicle.