🧭 sensor_based_navigation #

Offline Position Tracking Using Sensor Fusion + AI

A Flutter package for estimating device position and direction without GPS, using sensor data and LSTM-based AI models.

Features #

• GPS-free indoor navigation
• AI-based predictions using LSTM neural networks
• Drift correction using footstep detection logic
• Sensor fusion (accelerometer, gyroscope, magnetometer)
• Optimized for CPU inference

📱 How It Works #

Inspired by research from the University of Utah and others:

• Detects steps and resets drift every ~100ms of foot-ground contact
• Uses direction (magnetometer), movement (accelerometer)
• LSTM neural network processes time-series sensor input to estimate movement

Why LSTM? #

LSTM (Long Short-Term Memory) networks are ideal for sequential sensor data because they:

• Handle long-term dependencies
• Filter noisy or irrelevant input
• Model complex motion patterns over time

Performance Notes #

• ONNX model not fully supported on NNAPI or CoreML
• Use CPU backend with XNNPACK for best performance

License #

This project is licensed under the MIT License.