Comparison
Robot-mower navigation: boundary wire vs RTK, vision, beacons and LiDAR for real yards
Direct answer: choose the navigation architecture that fits the difficult parts of your property, not the technology with the most impressive headline. Boundary wire, RTK/GNSS, cameras, LiDAR and beacons solve different positioning problems and have different failure points. Many current mowers combine two or more of them.
Start with five property facts: sky visibility, narrowest passage, number of disconnected lawn zones, maximum slope and the consequences of crossing a boundary. Then verify the exact model’s documented installation limits, signal-loss behaviour, connectivity requirements and recovery method.
This guide compares navigation approaches, not mower brands. It contains no hands-on reliability findings and does not rank individual products. Manufacturer claims are identified as such and must be checked against the current manual for the exact model and market.
The principal limitation
“RTK mower”, “vision mower” or “LiDAR mower” is rarely a complete technical description. A mower may use satellites for position, a camera for obstacle detection, wheel odometry between fixes and a cloud or local reference service for corrections. Two products using the same headline technology may respond differently to a tree canopy, a lost correction signal or a blocked passage.
If the manufacturer does not document a material behaviour and there is no controlled independent test, write UNKNOWN.
The four jobs inside navigation
Separate these jobs before comparing systems:
- Boundary definition: how the mower knows where it may and may not go.
- Position estimation: how it determines where it is within that boundary.
- Obstacle detection: how it reacts to people, animals, toys, trees and temporary objects.
- Route and recovery: how it covers the lawn, reaches another zone, returns to charge and behaves when a signal or sensor is lost.
A virtual map may define a boundary without guaranteeing obstacle detection. A camera may detect obstacles without being the only positioning sensor. A reference station can provide corrections without deciding the mowing route.
Essential terms
Boundary wire
A physical loop installed around the work area creates a boundary signal detected by the mower. A separate guide wire may help it return to the charger or traverse passages. Installation requires labour and later breaks or layout changes need repair, but the boundary does not depend on a view of the sky.
Husqvarna’s current guidance illustrates that exact passage limits are model- and installation-specific: in one Automower boundary-wire system, the recommended passage can narrow when a guide wire is used, and the guide wire routes the mower to remote areas and back to the charging station. These figures must not be transferred to another brand or model.
GNSS and RTK
GNSS is the family of satellite-navigation systems; GPS is one member. RTK uses correction data from a reference source to improve positioning. GPS.gov explains that satellite accuracy worsens near buildings and trees because signals can be blocked or reflected.
RTK corrections may come from a local reference station or a network/cloud service. Husqvarna documents both patterns for EPOS installations: its local reference-station option can navigate without internet, while its cloud option requires suitable Wi-Fi or cellular coverage for correction delivery. It also documents that tall buildings and dense tree crowns can block satellite signals. These are Husqvarna implementation facts, not universal specifications for every RTK mower.
Vision
Vision systems use one or more cameras and software to infer boundaries, position, obstacles or terrain features. Some mowers use vision as the primary wire-free system; others use it to support satellite navigation when satellite reception weakens. Husqvarna, for example, documents camera backup navigation on certain current vision-equipped EPOS models, while its non-vision NERA models can use a support wire in weak-satellite areas.
The word “vision” does not establish low-light, glare, rain, edge or obstacle performance. Those behaviours require exact-model documentation or controlled testing.
LiDAR
LiDAR estimates distance by measuring reflected light and can create a geometric map of the surroundings. Dreame describes its A1 as using a 3D LiDAR system for mapping and obstacle detection without boundary wire or an RTK station. Its precision, range and coverage statements are manufacturer laboratory claims, not independent yard reliability results.
LiDAR avoids the satellite-sky dependency of RTK, but its real performance around vegetation, reflective surfaces, dirt, changing geometry and edges remains model-specific.
Beacons and reference stations
A beacon system uses installed local anchors to help establish position or boundary geometry. An RTK reference station is a different kind of reference source: it supplies satellite-position corrections. Both add fixed infrastructure and placement requirements.
ECOVACS documentation for the GOAT G1 shows a navigation-beacon installation and states that mapping cannot proceed when none of its beacons communicate properly. That is an exact product-family dependency, not a rule for all beacon systems.
Hybrid
A hybrid combines approaches—for example, RTK plus vision, virtual boundaries plus a support wire, or mapping plus local obstacle sensing. A hybrid can cover a weakness in one sensor, but it also introduces more behaviours to verify: which sensor is authoritative, when fallback begins, whether the mower slows or stops, and how it recovers.
Navigation comparison matrix
The following is a Nerd Mango editorial decision matrix. It describes typical questions, not guaranteed product behaviour.
Approach | Boundary/setup burden | Trees and buildings | Narrow passages and zones | Connectivity dependencies | Signal-loss question | Best candidate when | Main evidence gap |
|---|---|---|---|---|---|---|---|
Boundary wire + optional guide wire | High initial physical installation; layout changes require wire work | Boundary itself does not require satellite sky view | Can be predictable when the exact model supports passage/guide geometry | Core boundary may be local; app, schedules and updates vary | What happens after a wire break or lost loop signal? | You value a physically defined boundary and can install/maintain wire | Exact spacing, guide-wire, secondary-area and recovery rules |
RTK/GNSS with local reference station | Map plus careful station placement and power | Satellite blockage and multipath can matter at mower and station | Virtual zones are flexible; tight corridors may have weak sky | Navigation may avoid internet if correction is local; app features vary | Does it stop, coast, use odometry/vision or require reacquisition? | Much of the lawn has open sky and virtual editing is valuable | Required sky view, base placement, radio range and fallback |
Network/cloud-corrected RTK | Less local reference hardware on some models; virtual mapping | Same satellite constraints plus correction-delivery coverage | Flexible zones; coverage must reach required areas | May require Wi-Fi, cellular and vendor correction service | What happens when internet or correction service fails? | Open sky and stable documented correction coverage exist | Subscription/service dependency and offline behaviour |
Vision-led | Usually no perimeter wire; map/training varies | No satellite dependency unless part of a hybrid | Depends on visual landmarks, route width and model logic | Core navigation may be local; app/cloud needs vary | How does it behave in darkness, glare, rain or obscured lenses? | Boundaries are visually distinct and exact-model evidence fits the yard | Edge cases and measured repeatability in comparable conditions |
LiDAR-led | Usually map by driving/walking the boundary; no satellite base | Buildings do not block satellite signals because satellites are not the positioning source | Geometry-rich areas may help mapping; exact minimum spaces vary | App/updates may need connectivity even if ranging is local | What happens when the sensor is obscured or the map changes? | Sky is obstructed and the documented geometric limits fit | Vegetation, reflective/absorptive surfaces, dirt and changing scenes |
Local beacons/anchors | Install, power or maintain several anchors | Avoids satellite dependency; anchor placement becomes critical | Coverage geometry can suit complex zones if documented | May be local; app/account needs vary | Can it continue if one anchor is lost? | You can place stable anchors with suitable coverage | Quantity, range, line-of-sight, batteries and failure handling |
Hybrid | Highest system complexity; may reduce a single-mode weakness | Potentially handles more conditions | Potentially flexible | Depends on all active paths and the fallback design | Which subsystem takes over, and is the handover documented? | The exact product documents a fallback that matches your hardest area | Whether “hybrid” is real failover or only feature coexistence |
The property-first decision framework
Step 1: Draw the non-negotiable boundary
Mark roads, footpaths, ponds, pools, drop-offs, play areas, pet areas, fragile garden beds and any place where crossing the boundary would be consequential. Read the exact mower’s safety manual. Do not rely on a virtual boundary beside a hazard when the manufacturer requires a physical barrier, setback or exclusion.
Husqvarna’s slope guidance, for example, requires suitable exclusion and describes a physical barrier beside a public road in specified installations. That is evidence for those models and illustrates why safety rules must come from the exact manual.
If you cannot create a safe installation that follows the manual, stop the purchase decision.
Step 2: Survey the sky
Walk the whole mowing area and the proposed charger/reference-station position. Mark:
- dense tree canopy;
- tall walls and buildings;
- covered side passages;
- metal structures;
- areas where the sky is visible only in one direction.
For an RTK candidate, compare those areas with the manufacturer’s current sky-view and correction-coverage requirements. A clear base-station view does not compensate for every mower position; Husqvarna explicitly requires adequate sky visibility across the lawn for its documented EPOS configurations.
Step 3: Measure passages and transitions
Record the narrowest usable width, not the fence-to-fence width. Subtract garden edges, roots, posts and the mower’s required clearance. Check turns at both ends. A passage that is wide enough for the chassis may still be too narrow for reliable guidance, turning or systematic mowing.
Use the exact model manual. Do not copy Husqvarna’s 60 cm guide-wire example, or any other vendor’s figure, into a different system.
Step 4: Map zones and transport paths
Identify:
- connected lawn zones;
- separated zones that require manual carrying;
- gates that may be closed;
- paths the mower may cross but not cut;
- slopes or obstacles between the lawn and charger.
Ask whether the model supports multiple work areas, transport paths and no-go zones; whether each can have its own schedule; and whether a lost map can be restored.
Step 5: Measure slopes where they matter
Calculate slope as vertical rise divided by horizontal run, multiplied by 100. Compare both the work-area slope and the boundary slope with the exact model specification; these may be different. Measure wet or soft transitions and the turning area, not just the easiest central strip.
Manufacturer maximum-slope figures are controlled-test specifications, not proof of dependable traction on every wet, uneven or obstructed yard.
Step 6: Separate boundary accuracy from edge cutting
A precise position estimate does not mean the blade reaches the physical edge. Chassis shape, blade offset, boundary setback, surface transition and mowing pattern affect the uncut strip. Require an exact-model diagram or comparable controlled test before promising edge results.
Step 7: Audit connectivity and service dependencies
For each candidate, record:
- whether core navigation needs internet;
- whether RTK corrections are local, cellular, Wi-Fi or vendor-cloud delivered;
- whether mapping and schedule changes require an account;
- whether the mower can start and return to charge during an outage;
- what functionality remains at the mower itself;
- update and support commitments;
- any subscription or data-service dependency.
Do not equate “wire-free” with “infrastructure-free”. An RTK station, beacons, Wi-Fi coverage or cloud correction service may replace the perimeter wire.
Step 8: Demand a recovery answer
The critical question is not only “How does it navigate normally?” Ask what it does when:
- satellite accuracy or correction data is lost;
- a camera or LiDAR window is obscured;
- one beacon fails;
- the local network or internet fails;
- the charger moves;
- the map is corrupted or the app account is unavailable.
The answer should say stop, continue within a physical boundary, use a named fallback, return to charge, or unknown. Husqvarna documents different signal-loss behaviour for systematic and irregular mowing when its EPOS Support by Wire feature is used, demonstrating why mode-specific answers matter.
Three example yards
Yard A: open sky, simple boundary, frequent garden changes
An RTK/virtual-boundary system may deserve a trial because the sky view is favourable and zones can be edited without moving wire. Verify correction delivery, charger placement and behaviour if connectivity fails.
Yard B: mature trees, tall walls and a covered side passage
A satellite-only decision is risky without exact evidence for the obstructed sections. Compare boundary wire, a documented vision/LiDAR system, local beacons or a genuine hybrid with a named fallback. The headline “centimetre accuracy” under open sky does not answer the covered-passage problem.
Yard C: multiple zones, steep transition and a public path
Start with safe boundary and slope compliance. Determine whether zones connect through an allowed transport path and whether the mower can return without crossing the hazard. If the manual requires a barrier or excludes the slope, no navigation feature overrides that requirement.
These examples are decision demonstrations, not product recommendations.
The robot-mower property scorecard
Complete this original asset before comparing prices:
Property constraint | Your measurement/evidence | Boundary wire | RTK local | RTK cloud | Vision | LiDAR | Beacons | Hybrid | Unresolved blocker |
|---|---|---|---|---|---|---|---|---|---|
Consequential boundaries/hazards | |||||||||
Sky obstruction across lawn | |||||||||
Reference-station location | |||||||||
Narrowest passage and turns | |||||||||
Connected and separated zones | |||||||||
Maximum work-area slope | |||||||||
Maximum boundary slope | |||||||||
Edge-cut expectation | |||||||||
Wi-Fi/cellular coverage | |||||||||
Internet/cloud dependency | |||||||||
Lost-signal behaviour | |||||||||
Charger return/recovery | |||||||||
Installation and repair burden | |||||||||
Support/parts/update route |
Use SUPPORTED, UNSUITABLE, TEST REQUIRED or UNKNOWN in each technology cell. Eliminate any candidate with an unresolved safety or installation blocker before scoring convenience features.
What to verify in a demonstration
If a dealer or controlled trial is available, use your property’s hardest conditions:
- the narrowest real passage;
- the most obstructed sky area;
- representative slope and edge transitions;
- travel between zones and return to charge;
- a safe, documented signal-loss or connectivity-loss scenario;
- recovery after the obstruction or connection is restored.
Record the exact model, firmware, map settings, weather, surface and result. One successful pass is not proof of repeatability; ask for repeated runs if reliability is material. Do not create a hazard to test failure.
Stop conditions
Do not proceed when:
- the safety manual’s boundary, barrier or slope requirement cannot be met;
- the vendor cannot state what happens after loss of the required positioning signal;
- a critical passage is narrower than the exact documented minimum;
- the charger or reference station cannot meet placement requirements;
- a cloud or connectivity dependency is unacceptable and no local alternative is documented;
- the return, recovery or manual-rescue path is impractical;
- a model-specific claim is supported only by generic technology marketing.
The strongest choice is not the newest sensor. It is the system whose documented dependencies and failure behaviour fit the hardest parts of your yard.
Sources
- RET-S01 — How to create a passage in an Automower installation — Husqvarna; retrieved 2026-08-07; market US.
- RET-S02 — How to lay a guide wire in an Automower installation — Husqvarna; retrieved 2026-08-07; market US.
- RET-S03 — GPS Accuracy — GPS.gov / US Government; retrieved 2026-08-07; market GLOBAL.
- RET-S04 — Install virtual boundaries with Husqvarna EPOS — two options — Husqvarna; retrieved 2026-08-07; market US.
- RET-S05 — Why does Automower keep stopping? EPOS wire-free signal issues — Husqvarna; retrieved 2026-08-07; market GB.
- RET-S06 — EPOS Support by Wire for weak satellite areas — Husqvarna; retrieved 2026-08-07; market US.
- RET-S07 — Roboticmower A1 product page — Dreame; retrieved 2026-08-07; market ES-locale global site.
- RET-S08 — How to properly install the Dreame A1 robotic mower — Dreame Support; retrieved 2026-08-07; market US.
- RET-S09 — GOAT G1 support hub — ECOVACS Australia; retrieved 2026-08-07; market AU.
- RET-S10 — GOAT navigation-beacon abnormal communication FAQ — ECOVACS; retrieved 2026-08-07; market GLOBAL.
- RET-S11 — Optimising an Automower installation for steep slopes — Husqvarna; retrieved 2026-08-07; market US.
- General information: Nerd Mango provides general informational content. It is not legal, financial, medical, investment or other professional advice.
- Pricing & availability: Pricing, features, availability and compatibility may change. Verify current details with the provider before purchasing or subscribing.
- AI assistance: AI tools assisted research and drafting. Every article is edited and approved by a real human editor; AI is never the accountable author and never publishes autonomously.