Skip to content

dengebot v1.7.1 — Adaptive Sensor Fusion & Stability Improvements

Pre-release
Pre-release

Choose a tag to compare

@Gmby06 Gmby06 released this 28 Oct 23:26
· 30 commits to main since this release
4d8bcd6

dengebot v1.7.1 — Adaptive Sensor Fusion & Stability Improvements

Community Preview | Arduino Balancing Robot Library

We're pleased to introduce dengebot v1.7.1, featuring adaptive sensor fusion, gyroscope bias calibration, expanded telemetry, and improved reliability for two-wheel self-balancing robots.

This release builds on v1.6.1 while preserving the existing PID controller and motor command behavior.

What's New in v1.7.1

1. Adaptive Complementary Filter

  • Combines accelerometer and gyroscope measurements for pitch estimation.
  • Optional adaptive alpha switching between 0.980 and 0.998.
  • Acceleration-magnitude-based switching with hysteresis.
  • Updated default thresholds: 0.03 g activation / 0.02 g deactivation.
  • Preserves the original fixed-alpha filtering API.
  • Provides usedAlpha() for monitoring filter behavior.

Adaptive filtering remains disabled by default for backward compatibility.

2. Gyroscope Bias Calibration

  • New EDAGyroBiasCalibrator helper.
  • Supports stationary gyroscope bias measurement.
  • Rejects invalid or excessive-motion calibration samples.
  • Allows calibrated bias compensation in the complementary filter.
  • Includes a new GyroBiasCalibrationDemo example.

3. Extended Telemetry and Diagnostics

  • New lightweight telemetry support for embedded systems.
  • CSV output for sensor and controller analysis.
  • Optional reference pitch angle.
  • Event codes for recording test events.
  • Sensor timestamps and active alpha reporting.
  • Extended CSV output works without a reference angle.

4. Reliability and Safety Improvements

  • Improved adaptive-alpha configuration validation.
  • Protection against inverted normal/high-alpha settings.
  • Validation of sensor values and time intervals.
  • Shared portable floating-point validation utilities.
  • Existing PID fault handling and motor safety behavior retained.

5. Examples and Testing

Includes four Arduino examples:

  • BasicBalance
  • MotorCommandDemo
  • FilterTelemetryDemo
  • GyroBiasCalibrationDemo

Host regression tests cover the controller, motor conversion, filtering, telemetry, and gyro calibration functionality.

Simulation Results

In a 50-scenario inverted-pendulum simulation, the updated adaptive filter reduced average peak pitch-estimation error during the first second after a disturbance:

Configuration Average peak estimation error
Fixed alpha 0.980 4.95°
Adaptive alpha, previous thresholds 4.92°
Adaptive alpha, v1.7.1 thresholds 1.92°

This represents approximately a 61% reduction compared with fixed alpha 0.980 under the simulated conditions.

Gyroscope bias calibration also reduced simulated true upright-angle deviation from 0.48° to 0.25° for the adaptive configuration.

Important: These are model-dependent simulation results, not guaranteed improvements on physical robots. Sensor vibration, mechanical configuration, motor characteristics, and calibration can affect performance.

Compatibility

Designed for Arduino-compatible platforms including:

  • Arduino Nano / Uno / Mega 2560
  • ESP8266
  • ESP32
  • ESP32-S3

The existing PID core and motor command conversion remain unchanged. Physical compatibility and behavior depend on the selected board, sensors, drivers, and application code.

Development Status

This is a Community Preview release. Host tests and simulations have been performed, and initial user-reported robot testing is encouraging. Comprehensive physical validation of adaptive sensor fusion is still in progress.

We welcome community feedback, hardware testing, bug reports, and contributions.

Repository: https://github.com/Gmby06/dengebot

License: MIT