Skip to content

Auto-Discovery and Zero-Config Localization Roadmap #71

Description

@manankharwar

Summary

FusionCore has already removed much of the tuning burden traditionally associated with robot localization. A natural next step would be reducing setup and configuration requirements as well.

The long-term goal is simple:

ros2 launch fusioncore auto.launch.py

and have FusionCore automatically discover, classify, validate, and fuse available sensors with little to no user configuration.

Motivation

Most localization problems are not caused by the filter itself. They are caused by:

  • Incorrect TF trees
  • Sensor configuration mistakes
  • Poor covariance values
  • Timestamp issues
  • Misidentified or duplicate data sources

Users often spend more time configuring and debugging their localization stack than actually tuning the estimator.

FusionCore is already moving toward self-tuning and adaptive estimation. Extending that philosophy to setup and system integration could significantly improve the user experience.

Proposed Features

Sensor Auto-Discovery

Automatically detect and classify available sensors:

  • IMU
  • Wheel odometry
  • Visual odometry
  • LiDAR odometry
  • GPS / GNSS

Example:

Found:
✓ IMU: /imu/data
✓ Wheel Odometry: /wheel/odom
✓ GPS: /fix
✓ Visual Odometry: /vslam/odom

Automatic Noise Characterization

Instead of requiring users to provide sensor noise values, estimate them directly from incoming data.

Example workflow:

Collecting stationary IMU data...
Estimating gyro noise...
Estimating accelerometer noise...
Calibration complete.

This could further reduce the amount of sensor-specific configuration required.

TF Validation

Detect common frame and mounting issues before fusion begins.

Examples:

  • IMU mounted upside down
  • Suspicious frame rotations
  • Missing transforms
  • Invalid TF chains

Example warning:

Warning: imu_link appears rotated 180° relative to base_link.

Sensor Quality and Health Monitoring

Provide visibility into sensor reliability and confidence.

Example:

GPS Quality: Excellent
IMU Quality: Good
Wheel Odometry: Slip detected

This would help users diagnose issues much faster than manually inspecting topics and covariances.

Future Vision

The end goal is not just a better filter.

The goal is a localization system that understands the robot, configures itself, validates its inputs, and provides useful feedback when something is wrong.

FusionCore's biggest differentiator should not be that users spend less time tuning parameters.

It should be that users spend less time thinking about localization at all.

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions