AMR Positioning Accuracy vs. Docking Accuracy: Why a 10 mm Navigation Spec Can Still Miss the Interface

In AMR procurement, accuracy-related numbers are quoted often but rarely understood correctly. Four metrics are often mixed up or lumped together as “accuracy” in spec sheets, but they measure completely different things.

A buyer reviewing a spec sheet that claims “positioning accuracy ±10 mm” may ask: “At that accuracy, half my docks will fail.” This pushback is not unreasonable. It reflects a widespread problem: the accuracy number on the spec sheet and the docking accuracy the buyer actually needs on-site are often not the same metric.

Understanding the difference between these four numbers is the first step to avoiding selection-stage misjudgment.

Accuracy Terminology Comparison

Spec sheets typically give the first two numbers (positioning accuracy and repeatability), but what buyers actually need are the latter two (docking accuracy and interface error). This information gap is the main source of selection mistakes.

Map Localization Accuracy vs. Repeatability

Map localization accuracy answers “does the robot know where it is?” It measures the deviation between the robot’s sensor-based position estimate (LiDAR SLAM, visual SLAM, QR codes) and its true position.

Localization accuracy is affected by:

• Sensor resolution and precision
• Map quality (feature richness during mapping)
• Environmental changes (layout changes after mapping, floor wear causing feature loss)
• Dynamic obstacle interference

Repeatability answers “how much does the robot deviate each time it comes to the same point?” It measures position scatter across multiple arrivals at the same target.

Repeatability and localization accuracy measure different things — repeatability measures consistency of arrival, while localization accuracy measures how close the estimated position is to the true position. One is not inherently better than the other; they answer different questions.

But good repeatability does not mean good docking accuracy. If the robot consistently deviates to the left, repeatability is excellent (perfectly consistent), but if the docking interface requires tight tolerance, this robot still cannot dock successfully.

Docking Accuracy: Actual Deviation at the Station or Carrier

Docking accuracy is the combined result of localization accuracy, repeatability, load geometry, floor error, and docking strategy. It is not a number you can read directly from a spec sheet.

Factors affecting docking accuracy:

Load effects: Under full load, the load weight changes chassis attitude (suspension compression, tire deformation), and the navigation sensor’s height and angle may shift slightly. A robot’s positioning accuracy when empty may degrade under load. Buyers who report that spec-sheet accuracy feels “half-useless” on-site are likely experiencing accuracy degradation under load.

Floor effects: If the floor near the docking position is uneven, sloped, or has gaps, the robot will produce position deviations during the final few centimeters of fine adjustment. Floor wear can cause visual navigation to lose positioning — buyers have reported “navigation unreliable, no positioning sustained” messages when floor texture degrades.

Docking strategy: Many systems may use local fine-positioning methods (QR codes, reflectors, visual features, mechanical locating pins) for the final docking approach, depending on the interface tolerance required. If the docking position lacks these aids, the robot relies on global positioning accuracy — which may be insufficient for tight-tolerance interfaces.

Load geometry transfer: Even if the robot’s body positioning accuracy is good, if the load CG offset causes the load interface to shift relative to the robot, the final interface error increases.

Why Load Geometry and Floor Error Show Up at the Interface

The accuracy error transfer chain:

Robot localization error → chassis attitude error → top module mounting error → load interface offset → final interface error

Each step can amplify or introduce new error sources:

These errors do not simply add (some may cancel out), but in the worst case they do stack. This is why a robot with good positioning accuracy on paper may produce significant interface error during on-site docking.

Vision, Reference Markers, and Local Repositioning

Most AMRs switch to a local fine-positioning strategy in the final stage of docking, not relying on global map positioning.

Common local positioning methods:

Confirm which local positioning method the supplier’s docking accuracy is based on. If it relies on QR codes or reflectors, the installation accuracy, maintenance requirements, and environmental durability (wear, dirt, obstruction) of these markers need to be included in the operations plan.

Floor marker maintenance is a long-term factor affecting docking accuracy, not a one-time installation fix. Buyers have reported that floor codes break often, resin coating gets dirty, and PET film gets crushed by forklift traffic.

How to Test Accuracy Under Full Load

Accuracy numbers seen in empty-load demos cannot be directly used as expectations for loaded conditions.

Full-load accuracy test method:

1. Prepare actual load: Use your actual load (weight, geometry, CG), not a counterweight — load geometry affects accuracy as much as weight.
2. Test multiple docking positions: At least 3 different positions (illustrative protocol; project acceptance count to be agreed), including straight approach and post-turn approach.
3. Repeat testing: At least 10 repetitions per position (illustrative protocol; project acceptance count to be agreed), recording each deviation.
4. Measure final interface error: Do not just measure robot position — measure the actual deviation between the load interface and the target interface.
5. Record failure count: How many needed manual correction? How many completely failed?
6. Different floor conditions: If it passes on clean flooring, also verify on oily or worn floors.

Docking accuracy is not about “the best attempt achieved X mm” but “the worst attempt deviated how much” and “how often manual intervention is needed.” 9 out of 10 good docks with 1 at a large deviation is completely different from 10 consistent moderate deviations for a production line.

Measuring Failed Docks and Manual Corrections, Not Just Millimeters

Accuracy evaluation should not look only at millimeters but also at operational-level metrics:

If spec-sheet accuracy translates to a 70% docking success rate, with 30% needing a manual push or repositioning, then on a production line it is indeed problematic. But if docking success rate is 99% and deviations are within tolerance, the same accuracy number may be completely adequate. The difference is in the application scenario and tolerance requirements.

Docking Accuracy Acceptance Template

Use this template during FAT and SAT to systematically verify docking accuracy.

Accuracy Verification Inputs

1. Distinguish four accuracy metrics — positioning accuracy, repeatability, docking accuracy, and interface error are not the same number.
2. Spec sheet accuracy is usually based on empty-load and ideal conditions — ask the supplier for accuracy data under full load.
3. Confirm the local positioning method — global positioning accuracy is usually insufficient for docking; QR codes, reflectors, or mechanical pins are needed.
4. Floor marker maintenance is a long-term cost — wear, dirt, and crushing all affect docking accuracy.
5. Measure final interface error, not robot position — load geometry transfer error may be larger than robot positioning error.
6. Look beyond millimeters — look at success rate and manual intervention frequency — low success rate and high success rate mean completely different things for a production line.
7. Do full-load docking tests at both FAT and SAT — empty-load demos cannot replace loaded verification.