Skip to updatesThis is a community made site. To go to the original source click here.

MaticRelease notes

Stable 98

Smarter Edge Mopping

Improved mopping performance during edge cleaning, especially along straight walls.

Improved Wire Prediction Model

Introduced a new wire model to better predict wires and improve performance.

Room Completion Fix

Resolved a bug in the session summary where some rooms were incorrectly marked as completed. App Updates:

Delete Old Matics

Users can now remove previously connected Matic devices from the app.

Stable 97

Improved Path Planning Predictability

Enhanced Matic’s navigation logic to create more consistent and efficient cleaning paths, reducing randomness in movement.

Driver Timeout Handling

Updated the Matic's driver behavior to explicitly return an error when a timeout occurs. This prevents timeouts from being misclassified as anomalies, reducing instances of sudden stops.

Duplicate Room Coverage

Resolved an issue where Matic would clean the same room multiple times within a session, upon customization of room boundaries.

Stable 96

Transition Cleaning

Matic now attempts to clean transition areas—small spaces between carpets and hard floors—to optimize cleaning coverage.

New GUI Implementation

We’ve revamped Matic’s graphical interface using iced, a cross-platform Rust GUI library. The new design aligns closely with our app’s look and feel, maintaining consistent colors, fonts, and icons. Additionally, memory usage has been significantly reduced from ~30 MB to ~20 MB.

Improved Edge Cleaning Motion

Matic now maintains a straighter path during edge cleaning for more precise results.

Mopping Skipped Rooms

Fixed an issue where Matic would occasionally skip mopping certain rooms. Firmware Updates:

Pre-Mopping Check for Better Efficacy

Matic now performs an additional verification step before mopping to ensure optimal cleaning effectiveness.

Improved Sweeper & Duct Obstruction Detection

The system now more accurately differentiates between a detached sweeper and an obstruction in the duct.

Stable 95

DINO Model Enhancements

Introduced DINO (Distillation with No Labels), a transformer-based neural net model that significantly improves occupancy map predictions, particularly for steps. The system now processes Bayer images as input, eliminating the need for computationally expensive image processing. This not only reduces latency but also preserves richer signal data, leading to more accurate predictions.

Scheduled Queuing

When a robot is manually controlled via joysticking or user navigation, it will now automatically resume its scheduled session afterward.