Real-time state estimation for amateur rocket avionics
This repository implements a Kalman Filter–based state estimation pipeline for amateur rocketry avionics, developed as part of the Bishop Strachan Rocketry Team (BSRT).
The goal of this project is twofold:
- Educational: provide a clear, beginner-friendly introduction to Kalman filtering for team members with little to no background in control systems.
- Practical: implement a real-time Extended Kalman Filter (EKF) suitable for flight use, operating under noisy sensors, vibration, and tight timing constraints.
To support this, the repository is intentionally structured in two stages:
- a 1D Kalman Filter for learning and intuition
- a full EKF used for flight-relevant state estimation
Launch Canada 2025 simulation testing and debugging flight avionics hardware during sensor integration.
In high-power rocketry, raw sensor readings are:
- noisy
- biased
- sampled at different rates
- occasionally unreliable (especially during boost)
A Kalman Filter is a control algorithm which provides a principled way to:
- fuse multiple sensors
- track hidden states (e.g., velocity)
- quantify uncertainty
- remain computationally lightweight enough for embedded hardware
The 1D Kalman Filter serves as an introductory stepping stone for new team members.
It models a simplified vertical system with:
- state: altitude
- measurement: barometric altitude
- assumption: constant vertical velocity over short intervals
This implementation is intentionally minimal and heavily commented, focusing on:
- prediction vs update steps
- covariance intuition
- tuning process noise vs measurement noise
- understanding filter behavior through plots
This section exists so that future team members can learn what a Kalman Filter is doing before encountering Jacobians or nonlinear dynamics.
The EKF implementation extends the 1D filter to a nonlinear, multi-state system appropriate for real flight data.
Typical states include:
- altitude
- vertical velocity
- barometric pressure sensor (altitude)
- IMU accelerometer (vertical acceleration)
- nonlinear state transition model
- sensor fusion with different noise profiles
- covariance propagation and update
- designed with embedded constraints in mind (runtime, memory)
This EKF is intended to run on a microcontroller-based flight computer and serve as the foundation for:
- apogee detection
- event timing (e.g., deployment logic)
- post-flight analysis
This implementation was designed with real-world constraints in mind:
- Sensor noise: accelerometer vibration during boost, barometer lag
- Timing: predictable execution time per update step
- Robustness: reasonable behavior during brief sensor dropouts
- Simplicity: no matrix libraries that would be infeasible on embedded hardware
The filter prioritizes reliable state trends over aggressive responsiveness.
The filter has been tested on:
- simulated flight profiles
- logged sensor data
- injected noise scenarios to observe stability and convergence
Plots and logs demonstrate:
- reduced noise compared to raw measurements
- smooth velocity estimates
- stable covariance evolution
*All results produced in simulation, not through genuine test-flights. The complete, SRAD-based avionics system is currently being engineered for payload-testing for Launch Canada 2026.
This repository is designed to be read, not just run.
It is actively used as a learning reference for:
- high school rocketry students
- first-time exposure to estimation theory
- understanding how theory translates into embedded systems
Code clarity and progression were prioritized over abstraction.
Planned extensions include:
- tighter integration with full flight software
- sensor bias estimation
- adaptive noise tuning
- validation on flight hardware
- complete integration of srad systems (removal of all COTS pyrotechnics, and otherwise)
Developed as part of the Bishop Strachan Rocketry Team (BSRT) avionics effort for the Tripoli Competition '25, Launch Canada 2025 and thereafter.