Three labels, several very different systems
RTK, LiDAR and Vision are often presented as competing technologies, but modern robot mowers frequently combine them. The useful comparison is therefore not a simple winner-versus-loser table. It is a question of what information each sensor provides, what it depends on, and how the complete navigation stack behaves when conditions become difficult.
What RTK contributes
RTK is a correction technique used with satellite positioning. Standard GNSS positioning can drift by metres; RTK uses correction data to improve relative positioning substantially. In robot mowers, that can support virtual boundaries and repeatable routes without a physical perimeter wire.
The dependency is important. Satellite geometry, tree cover, buildings, antenna placement and the correction workflow can affect performance. Some systems use a local reference station, some use network-delivered corrections, and some combine RTK with cameras or other localization methods for difficult sections.
What LiDAR contributes
LiDAR measures distances by emitting light and analysing the return. It can produce a geometric representation of the surroundings and help a mower localize itself relative to visible structures. This can be attractive in properties where a purely satellite-led solution is challenged by obstruction.
LiDAR does not remove the need for good software. The mower still has to build or use a map, identify its position within it and react to changing scenes. Tall grass, seasonal changes, feature-poor spaces and unusual surfaces can still create challenges depending on the implementation.
What Vision contributes
Vision navigation uses cameras and computer vision to recognize environmental features, lawn boundaries or obstacles. Cameras provide rich information: colour, texture and object appearance can be interpreted in ways that a pure distance sensor cannot. Vision is therefore useful both for localization and for obstacle handling.
The trade-off is environmental dependence. Lighting, glare, shadows, dirt on the lens and visually ambiguous edges can affect what a camera sees. A system’s software training and fallback behaviour matter at least as much as the presence of a camera.
Why hybrid systems are common
Outdoor navigation is difficult because yards contain both open sky and obstructed areas, permanent structures and temporary objects, smooth lawns and irregular boundaries. Sensor fusion can combine RTK/GNSS, cameras, LiDAR, inertial sensors, wheel odometry or ultrasonic sensors so the mower is not dependent on one signal source.
That does not mean ‘more sensors’ automatically equals ‘better’. Integration is the hard part. The mower must know when one source is unreliable and how much to trust the others. Firmware maturity and mapping behaviour therefore remain important ownership factors.
Which architecture suits which yard?
An open property with good sky visibility may be an easy environment for an RTK-led mower. A heavily structured yard with walls, trees and visible geometric features may benefit from LiDAR or a strong hybrid stack. Vision can be valuable where scene understanding and obstacle recognition are priorities. Complex properties may favour systems that combine methods rather than depend on one.
None of these statements should be treated as a guarantee for a specific model. Two products using the same technology label can perform differently because hardware placement, algorithms, update quality and mechanical design differ.
What to compare beyond the label
Look at whether a reference station is required, whether network RTK is supported, how initial mapping works, whether separate zones and no-go areas are supported, what the published passage constraints are, and how the mower handles a temporary loss of its preferred positioning source. These details usually tell a buyer more than the headline acronym.
Base stations, network corrections and subscriptions
RTK products also differ in how correction data reaches the mower. A local reference station can provide corrections from the property, while Network RTK can obtain them through an online service. That distinction affects installation, connectivity and sometimes ongoing service terms. It is therefore useful to separate the presence of RTK from the way RTK is delivered.
A local antenna needs an installation position that meets the manufacturer’s requirements. Network corrections reduce local hardware in some systems, but they introduce dependence on data coverage and the provider’s service. Hybrid products may support more than one mode or add vision as a fallback.
Obstacle avoidance is a separate question
A mower can have precise localization and still have only basic object detection. RTK tells the machine where it is; LiDAR or cameras may also contribute to obstacle perception, but the functions should not be conflated. When comparing products, look separately at localization, boundary accuracy and obstacle handling.
Firmware matters because navigation is software
Outdoor autonomy is not a fixed mechanical capability. Mapping logic, obstacle classifications and recovery behaviour can change through firmware. This makes the manufacturer’s update history and support ecosystem relevant. A strong sensor package with immature software can be more frustrating than a simpler architecture with reliable behaviour.
Bottom line
RTK, LiDAR and Vision solve different parts of the same problem: knowing where the mower is, understanding where it may mow and avoiding things it should not hit. Yardnetic treats the navigation label as one field inside a wider property-fit model, not as a quality score. The right choice is the architecture whose dependencies fit the yard.
Separate positioning from obstacle detection
One reason specification sheets become confusing is that navigation and obstacle avoidance are often described in the same marketing sentence. They are related but different jobs. A mower needs to know where it is, where it is allowed to mow and what is temporarily in front of it. RTK can contribute precise outdoor positioning, LiDAR can contribute geometry and localization, and cameras can contribute both scene understanding and obstacle recognition. A product may therefore use one technology for its map and another as a safety or fallback layer.
RTK: excellent precision with environmental dependencies
RTK-led systems can be extremely effective in open gardens because corrected satellite positioning allows repeatable virtual boundaries without burying perimeter wire. The practical questions are how correction data reaches the mower, where a reference antenna must be installed if one is required, and what happens in places where satellite visibility becomes poor. Network RTK removes some local-base-station constraints but still depends on coverage, correction service and the mower’s own positioning stack.
Tree canopy, walls and narrow passages beside buildings do not automatically make RTK unusable, especially in hybrid products, but they are exactly the places where buyers should look for documented fallback behaviour rather than assume centimetre-level precision is available everywhere.
LiDAR: geometry is powerful, but mapping still needs context
LiDAR is attractive because it can measure surrounding structure directly rather than depend on satellite visibility. In a yard with walls, fences, buildings and other stable features, that geometric information can support localization and mapping. The limitation is that a sensor does not make navigation decisions by itself. The mower still has to recognize useful features, maintain a map over time and distinguish a permanent structure from temporary clutter.
Feature-poor open lawns can be a different challenge from structured gardens. Seasonal changes, moved furniture and vegetation growth also change what the environment looks like. The best LiDAR implementations therefore combine reliable mapping with sensible recovery behaviour when localization confidence drops.
Vision: information-rich but software-dependent
Cameras can identify objects and interpret visual boundaries in ways that pure ranging sensors cannot. That makes vision particularly useful for obstacle handling and semantic understanding. But camera performance depends strongly on the software model behind it. Low sun, deep shade, rain on a lens, similar-looking grass and paving, or unusual objects can all create edge cases.
When comparing vision systems, look beyond the number of cameras. Consider whether the product uses vision as its primary localization method, as an obstacle layer, or as part of sensor fusion. Those roles imply different dependencies.
A practical decision framework
For open sky and uncomplicated geometry, an RTK-led mower can be a natural fit. For heavily structured areas or places where satellite visibility is inconsistent, LiDAR or strong visual localization can become more attractive. For yards with many temporary obstacles, capable vision can add meaningful value. Complex properties often benefit most from a hybrid system that can continue operating when one source becomes weak.
There is no universal hierarchy. The better system is the one whose dependencies match the property. Yardnetic’s model profiles therefore store the actual navigation stack rather than collapsing every wire-free mower into a single category.