embedded-kf Library Overview

embedded-kf is a lightweight C library designed to implement Kalman filters on embedded systems. While similar in technical scope to kalman-clib, it introduces new capabilities while reusing its matrix utility functions. embedded-kf offers the following features:

• Asynchronous measurement updates: Supports systems with sensors that provide data at different rates.
• Automatic C code generation: Generates optimized .c/.h files from user-defined JSON configurations with user-friendly APIs.
• Control vector support: Direct integration of control-vector inputs during prediction steps
• Fully statically-allocated: No dynamic memory is required, which is ideal for resource-constrained environments.

Key Features:

• Optimized for Embedded System Use: Designed for real-time operation on embedded systems, with easy integration and API
• Customizable and Extensible: Easily configurable via JSON, with extensibility to add custom filters or measurement models.
• Typical Use Cases:
  • Sensor fusion for robotics, drones, or autonomous vehicles
  • Real-time signal processing for IoT devices
  • Navigation systems or state estimation in constrained environments

Usage

1. Define a filter .json file. See generator/tests/samples for example filters
2. Run python3 kf_generator.py {path/to/filter/json} {optional: output directory, default=kf_output}
3. Build and link the generated .c/.h files into the software application. A CMakeLists.txt file is generated for convenience
4. Call the filter API - see info/API.md

Theory and References

Kalman Filter Theory

Example Smoothing

From an IMU filtering example, the figure below shows the plot of pitch as estimated from raw accelerometer movements and the Kalman filter state estimate