Vision-Powered Lawn Mowers How AI and 3D Mapping Eliminate the Need for Boundary Wires

For decades, autonomous lawn mowers have relied on buried boundary wires to define their working areas. This invention eliminates that requirement entirely by equipping mowers with vision-based navigation systems that build 3D maps of the landscape. The Toro Company‘s approach combines object detection, feature extraction, and spatial mapping to create mowers that understand their surroundings without external wiring infrastructure.

The Problem

Installing and maintaining boundary wires is expensive, time-consuming, and inflexible. Homeowners and landscape professionals face several challenges with traditional wire-based systems: the wires are costly to install initially, they can break and become inoperable, and they cannot be easily moved or reconfigured when landscaping needs change. The wires also create maintenance headaches as they degrade over time due to weather exposure and equipment damage. Additionally, traditional approaches to autonomous navigation have been computationally intensive, requiring significant processing power that portable mowers simply cannot afford given their battery constraints.

The fundamental challenge lies in creating robust navigation on machines with limited computing resources. Most autonomous vehicles rely on powerful computers and extensive sensor networks, but autonomous mowers must operate on batteries with minimal weight and power consumption. This creates an almost impossible balance between sophisticated navigation capabilities and practical operational constraints.

What This Invention Does

This patent describes a hybrid navigation system that combines the best of both visual and sensor-based navigation while minimizing computational overhead. The mower operates in multiple distinct modes, each optimized for different tasks.

During the training phase, an operator drives the mower around the perimeter of the area to be mowed. Cameras mounted on the mower capture images of the landscape, recording features like trees, buildings, fence posts, and natural terrain variations. The system extracts distinctive features from these images – points and patterns that the mower can recognize again and again – and stores them along with position information.

Once training is complete, the system processes all these images during an offline phase (typically overnight while the mower charges) to build a detailed three-dimensional point cloud map. This 3D representation captures the spatial layout of the entire work region, including terrain elevation and object positions. Critically, the system calculates camera positions for each training image, establishing precisely where the mower was located when each photo was taken.

During actual mowing operations, the mower uses a combination of low-frequency visual checks and high-frequency non-visual sensors. Inertial sensors and motor encoders provide constant position updates at high refresh rates, but they accumulate small errors over time. The vision system checks in at lower rates to correct these accumulated errors by recognizing previously mapped features. This approach keeps computational demands manageable while maintaining accurate navigation.

The system also incorporates object recognition to enhance performance. When the camera detects obstacles or previously unseen objects, it can prioritize searching for features that help distinguish those objects from the background, making navigation more robust even in changing conditions.

Key Features

Vision-Based Pose Determination. The system calculates the mower’s precise position and orientation by analyzing features captured in images. Rather than relying solely on inertial sensors that drift over time, it anchors position estimates to visually recognized landmarks in the environment.

3D Point Cloud Mapping. During offline processing, the system generates a complete three-dimensional map of the work region. This map serves as the reference against which all real-time navigation is compared, allowing the mower to understand whether it is moving toward or away from boundaries without any buried wires.

Hybrid Sensor Fusion. The invention combines high-frequency inertial and mechanical sensors with lower-frequency visual corrections. This optimization technique allows the system to maintain accuracy while operating within strict power and processing constraints.

Object Recognition Integration. When the camera detects objects in the environment, the system uses this information to prioritize which features to track visually. This adaptive approach makes the system more robust to seasonal changes and varying weather conditions.

Training Mode Flexibility. The system requires only a single tour of the mower around the perimeter during training. This simplicity means homeowners can easily redefine boundaries whenever they redesign their landscaping or expand their mowing areas.

Who Is Behind It?

The Toro Company, headquartered in Minnesota, USA, is one of the world’s leading manufacturers of grounds care and agricultural equipment. The company has a long history of innovation in the landscape equipment industry, including pioneering work in automation and robotics for turf maintenance. Inventors Trevor Myron Porter and Alexander Steven Frick are named on this patent. This application is a divisional of the parent patent application 2021218647, indicating that this work builds on several years of prior development and refinement.

Why It Matters

This invention addresses a practical problem that affects millions of homeowners and commercial landscapers. The elimination of boundary wires reduces installation costs, maintenance burdens, and environmental concerns. From a market perspective, this could accelerate the adoption of autonomous mowing technology by removing one of the significant barriers to entry.

The patent is also significant from a technical standpoint because it demonstrates how to achieve sophisticated autonomous navigation within severe computational constraints. The hybrid approach of combining high-speed inertial navigation with periodic visual corrections – a form of sensor fusion – has applications far beyond lawn mowing. This technique could be adapted for autonomous delivery robots, inspection drones, and other portable autonomous systems that must balance performance with power limitations. The systematic integration of object detection with feature-based localization represents a practical solution to the general problem of mobile robot navigation in dynamic environments.

The IPC classifications G05D 1/43 (automatic guidance systems), G06F 18/00 (machine learning), and A01D 34/00 (agricultural machines) reflect the invention’s interdisciplinary nature, spanning computer vision, robotics, and agricultural technology.

---

AU 2026201456 was published in the Australian Official Journal of Patents on 19 March 2026 and is open for public inspection. Patent applications represent inventions that are sought to be protected and do not necessarily reflect commercially available products.

Related Concepts

Robotic lawn mowers have traditionally used buried wire perimeters, but vision-based approaches now enable wire-free operation. Simultaneous localization and mapping (SLAM) algorithms build spatial models from camera data and are central to this approach. Combining high-frequency inertial measurement with periodic visual odometry corrections – a classic sensor fusion pattern – allows accurate navigation within tight power and compute budgets.