Precise navigation is essential for Yarbo to deliver reliable mowing, snow removal, and other autonomous yard tasks. Complex environments such as tree canopies, eaves, and reflective surfaces can cause GPS and RTK signals to degrade, leading to drift or work plan interruptions.
This update introduces a new feature that helps Yarbo continue working reliably even when RTK signal is briefly lost. With the support of an integrated Dead Reckoning (DR) algorithm, Yarbo can navigate smoothly through areas with weak GPS signal — such as under trees, near walls, or beneath eaves — until connection is restored.
The DR system now includes an app-level toggle, allowing users to set how far Yarbo can travel when RTK is unavailable or weak. This gives you control over how aggressively Yarbo operates in signal-challenged areas, balancing coverage with positional safety.
We’ve added a new quality/confidence status that evaluates the reliability of Yarbo’s current positioning data. Previously, an RTK “fixed” state (status = 4) could still contain errors, sometimes leading to sudden coordinate jumps — a key cause of boundary excursions. The new system raises the threshold for accepting RTK data, filtering out unreliable fixes before they cause navigation errors.
Even with DR enabled, poor RTK quality means higher positioning uncertainty. In obstructed environments (dense trees, buildings, reflective surfaces), the confidence value will drop. If 50% or more of a planned work area shows poor quality, it’s best to split that area into smaller zones or avoid it until conditions improve.
We are continuously refining our ability to detect multipath effects and obstructions through data collection in varied real-world environments. In the future, Yarbo’s visual navigation will help it intelligently seek out areas with stronger GPS signal before starting work, reducing the need for manual adjustments.
A built-in sensor that measures movement and rotation. It helps Yarbo understand its direction and speed even when GPS or RTK signals are temporarily blocked, keeping operations stable.
Odometry is the process of estimating how far and in what direction Yarbo has moved, based on its wheels, tracks, or internal motion sensors. It’s like a car’s odometer but also considers changes in direction.
A camera-based system that tracks movement by observing the ground and nearby surroundings. It detects slippage or drift and corrects navigation errors when physical conditions—like slick grass or uneven terrain—cause wheel spin.
To complement positioning upgrades, Yarbo is also advancing obstacle recognition and avoidance with next-generation perception models.
A deep learning model that improves Yarbo’s ability to estimate how far away an object is—helping it navigate around things like trees, poles, and fences with better spatial judgment.
An upgraded model that allows Yarbo to distinguish between different types of obstacles (e.g., pets, people, trash bins, garden beds), improving path planning and safety.
We’ve significantly expanded our training dataset with more real-world yard environments. This improves how well Yarbo recognizes a wider variety of obstacle types and terrain conditions.
Testing with PPP(Product Pioneer Program) users, with large-scale release scheduled for August.
Planned for September release to further enhance drift detection and correction.
Updates will be included in upc0ming firmware versions to improve obstacle handling and reduce uncut patches.