Executive Summary

An AGV that drifts off its path, or that stops in the wrong place, has three families of possible causes: navigation and calibration (sensor mounting angle, calibration parameters, floor marker accuracy), mechanical structure (frame deformation, suspension condition, assembly error), and the executing components - above all the wheels. The wheel-side problem is easy to underestimate because it is silent. No alarm is raised and no error code is logged; the only visible sign is a vehicle that slowly wanders off the line. When everything electronic has already been checked, wheel geometry is often where the remaining error lives.

Five geometric indicators define how a wheel behaves under a moving vehicle: diameter consistency between the wheels of the same machine, tread roundness, radial runout, face runout of the hub, and concentricity between the tread and the wheel core. Together they determine two things that matter to every AGV - heading-hold capability and odometry accuracy. Diameter mismatch is the most deceptive of them. A 0.5 mm difference between the left and right drive wheels, a relative difference of only 0.25%, accumulates an estimated 2.9 degrees of heading deviation and roughly 25 cm of lateral drift over a 10 m straight run when both sides are commanded to the same speed.

Closed-loop navigation does not remove wheel error; it converts it into observable symptoms. Once geometric error enters the control loop, it typically appears as an S-shaped trajectory with repeated corrections, faster wear on one wheel because that wheel performs more of the corrective work, and a falling first-attempt success rate at docking stations and charging posts. That is why a higher grade of navigation accuracy raises, rather than lowers, the precision demanded from the wheel.

This guide summarizes the mechanics behind wheel-induced drift, breaks the causes into six root categories, and gives you an executable five-step field method: visual inspection, diameter measurement, runout dial indication, hardness comparison, and a road test with odometry recalibration. It also provides a precision control checklist that can be written directly into purchasing and acceptance specifications. The final section takes the supplier-side view. Drift is a precision problem that is hard to remedy after the fact but controllable beforehand, and mapping each root cause to the process where it originates and the inspection gate where it is intercepted is the practical answer. HANKE's process control and batch inspection records are used here as a worked example of the verifiable data a customer can request.

Key AGV Drift Troubleshooting Data at a Glance

1. Start with the Symptom: Four Typical Ways an AGV Goes Off-Course

The most common mistake when troubleshooting an AGV that cannot track straight or stop accurately is to replace the sensor the moment someone reports drift. In reality, the same complaint covers several entirely different fault chains. Sorting the symptom first, before deciding which direction to investigate, saves a large amount of trial and error.

Field feedback from wheel-side precision problems converges on four patterns. The first is persistent one-sided drift: along a straight section the vehicle keeps pulling to one side, and the further it travels the worse the deviation becomes. The second is an S-shaped path, where the trajectory oscillates between the two sides of the planned route and the controller corrects continuously. The third is inaccurate stopping: repeated docking attempts at the same station land in different positions, so the first-attempt success rate falls. The fourth is in-place rotation error: a differential-drive vehicle commanded to spin 360 degrees does not return to its starting heading, or angular repeatability degrades.

The mapping between these four symptoms and the wheel indicators is not complicated, and the table below can be used directly for a first on-site localization.

One caveat applies to this table: it lists the main suspects, not conclusions. Navigation calibration error, frame deformation and suspension failure can all produce very similar symptoms. The argument for inspecting wheels anyway is cost. A caliper, a dial gauge and a Shore durometer cover most of the items, so wheel checks should run in parallel with navigation troubleshooting rather than queue behind it.

2. How Wheels Steer an AGV Off-Course: Three Fundamental Principles

To treat drift as an engineering problem rather than something mysterious, you need three basic relationships between wheel behaviour and vehicle motion. Once these three layers are clear, every troubleshooting action that follows has a reason behind it.

2.1 Odometry: How Far the Wheel Turns Is Not How Far the Vehicle Travels

An AGV controller estimates distance travelled by reading encoder pulses from the wheel or the motor shaft and multiplying by the wheel circumference: revolutions multiplied by circumference equals distance. The formula assumes that the wheel rolls effectively, does not slip, and has a stable circumference. Once the rolling radius carries a systematic error, the distance estimate shifts as a whole, and that shift accumulates with time.

One concept is frequently overlooked here: rolling radius is not the same as geometric radius. The tread compresses under load, so the effective rolling radius is slightly smaller than the drawing radius, and the amount of compression depends on hardness, load and temperature. This explains a common site experience - odometry is calibrated when new wheels are fitted, and then has to be recalibrated again after some time in service. The encoder is rarely at fault. The rolling radius has changed, whether through tread wear, a change in load, or a change in temperature. A rolling radius error of 0.25% corresponds to roughly 2.5 m of accumulated odometry error for every kilometre travelled.

2.2 Differential Drive: How a Diameter Difference Becomes a Heading Error

A differential-drive AGV steers by the speed difference between its left and right wheels. When the controller issues an equal-speed command but the two wheels have different actual diameters, their linear speeds differ, and the vehicle travels along an arc instead of a straight line.

The deviation can be estimated with two simple relationships. The difference in distance travelled equals the travel distance multiplied by the relative diameter difference (dL = L x dD/D). The heading deviation equals that distance difference divided by the wheel spacing (theta = dL / W). For a differential-drive AGV with 500 mm wheel spacing and 200 mm drive wheels, the theoretical deviation over a 10 m straight run is shown in the table below.

The table is a theoretical calculation that assumes 500 mm wheel spacing, 200 mm wheel diameter, equal-speed commands on both sides, and no closed-loop correction. Its purpose is to show the trend - a diameter difference amplified by travel distance. A real vehicle corrects continuously, so the deviation does not accumulate all the way to the end of the run, but the order of magnitude and the growth pattern still hold. This is why a difference of 0.2 mm, which looks negligible on a drawing, becomes a visible deviation on the factory floor.

2.3 Closed-Loop Correction Is Not the Same as Eliminating Error

Many sites treat the presence of laser navigation as a reason to leave the wheels alone. That is a misunderstanding. What closed-loop navigation does is measure the deviation and correct for it. It changes how the error appears, not the error itself.

When wheel geometry error enters a closed-loop system, it usually shows up in three observable ways. The trajectory takes on an S shape as the vehicle corrects back and forth across the path. One wheel wears noticeably faster than the other, because it absorbs more of the corrective motion. Positioning confidence drops, which appears as a lower first-attempt success rate at docking stations and a need for second micro-adjustments. Frequent correction also brings a small rise in energy consumption and a change in travel noise. In other words, the higher the grade of navigation accuracy, the higher - not lower - the precision demanded from the wheel.

3. Six Root Causes, One by One

Ranked by how often they appear in the field and by their weight in the outcome, wheel-side causes of drift and inaccurate stopping fall into six categories. Each entry below covers the cause, the typical symptom and the direction of the remedy, and is meant to be worked through against your own site.

3.1 Left and Right Drive Wheels with Different Diameters

This is one of the most frequent categories. Diameter differences come from three main sources. The left and right wheels may come from different production batches or different suppliers, so a difference already exists inside the drawing tolerance. The two sides may wear at different rates, so the diameter gap widens with service time. And the two sides may compress differently, for example when uneven load distribution leaves one side with a smaller effective rolling radius.

The key remedy is pairing. Before a wheel set is assembled onto a machine, measure and record the diameter of every drive wheel, then pair the two closest wheels onto the same axle position. The effort is small, but the effect on suppressing drift is direct. Taking a flanged drive wheel as an example, HANKE's GSFH and GSFM series cover a diameter range of 100 to 300 mm and carry 300 to 1100 kg per wheel at 4 km/h. Key dimensions are sampled by batch and recorded before shipment, and a second round of diameter pairing verification at the customer's assembly station brings the probability of this type of drift down to a low level.

3.2 Tread Roundness and Radial Runout Out of Tolerance

A tread is never a perfect circle. Its radius varies periodically within one revolution, and the magnitude of that variation is the radial runout. Excessive runout has two consequences. Angular velocity stays constant while linear velocity fluctuates periodically, which produces low-frequency vibration and a crawling feel. And the distance estimate is modulated once per wheel revolution, which accumulates into a position deviation over time.

The usual causes are insufficient machining allowance or a tool-setting deviation in the finish turning step after casting, uneven polyurethane thickness caused by mold wear, and insufficient rigidity in large-diameter wheels. The standard industry control is dial indication after finish turning: a dial gauge is placed against the tread and the wheel is rotated one full turn to read the runout. At HANKE, finished-product inspection includes a CHOTEST coordinate measuring machine with 1 µm precision that inspects key dimensions batch by batch, and dimensional anomalies of this kind are detected mainly at this stage.

3.3 Hub Concentricity and Face Runout

A wheel is an assembly of a polyurethane tread and a metal core, and their relative position determines whether the wheel rotates true. If the tread is not concentric with the core, the wheel rotates with the equivalent of a fixed eccentricity that pushes the vehicle once per revolution. If the core face runout is large, the wheel wobbles axially, which causes uneven pressure and one-sided wear.

Both indicators originate in the turning accuracy of the wheel core, the press-fitting or assembly process, and the clearance of the bore-and-shaft fit. In the field, the fast check is to lift the wheel, turn it slowly by hand, and watch whether the relative position of tread and core remains stable, then measure face runout with a dial gauge. Note that a concentricity problem cannot be solved by changing the tread material; it has to be addressed in core machining and assembly.

3.4 Inconsistent Hardness and Friction Coefficient

Hardness affects two things: the friction coefficient between tread and floor, and how much the tread compresses under load. When the two wheels of a machine differ in hardness, both the available traction and the rolling radius differ. Under the same drive torque, the softer or lower-friction side tends to slip slightly, and the distance covered during that slip is not counted by the encoder, so the odometry error becomes irregular rather than systematic.

Two common misconceptions are worth clearing up. First, higher hardness is not automatically better. Raising hardness sacrifices floor protection and may change noise behaviour, so it has to match the duty. Second, hardness is not the same as wear resistance. Wear resistance is mainly determined by the formulation structure and is normally measured with an indicator such as the DIN 53516 abrasion value. HANKE uses two tread systems: Eamflex 93A, oriented toward high wear resistance and traction, and Saxflex 75A, oriented toward floor protection and low noise. Hardness is controlled within tolerance bands such as 75 ± 3 Shore A, and on a single machine, wheels of the same system, same batch and same hardness are recommended.

3.5 One-Sided Wear and Uneven Abrasion

One-sided wear is an amplifier of drift. A small initial diameter difference makes the vehicle pull to one side. When the controller corrects, that wheel takes on more slipping and friction, so it wears faster and its diameter shrinks faster, which widens the diameter gap further. The result is a positive feedback loop: the more it drifts, the more it drifts.

The check is direct. Remove both wheels, measure tread wear and diameter, and compare. A clear difference means the loop is already running. At that point, replacing only one wheel usually has limited effect. Replace them as a pair, and correct the initial diameter difference or assembly problem at the same time, or the new wheel will reproduce the same wear pattern before long.

3.6 Bond Layer Slippage and Delamination

The polyurethane tread and the metal core are bonded into a single piece. If the bond layer fails locally, the tread rotates slightly relative to the core. The signature of this behaviour is that the encoder registers revolutions while the vehicle does not travel the corresponding distance, so the odometry error appears suddenly and unpredictably. Unlike the systematic offset produced by a diameter difference, bond slippage is abrupt and random, and it is often accompanied by abnormal noise or tread displacement under load. It belongs to the category that should be taken out of service for inspection immediately.

Bond quality stability depends on the control of three critical processes: whether sandblasting has completely removed the oxide layer from the core surface, whether the adhesive has been applied evenly and completely, and whether casting temperature and pressure parameters are stable. HANKE's manufacturing flow covers sandblasting, adhesive application and casting with batch records for each, which is one of the conditions behind 48 months of operation on OEM production lines such as Mercedes-Benz with no after-sales record.

4. A Five-Step Field Troubleshooting Method (Easiest to Hardest)

The sequence below follows one principle: start with the cheap checks that cover the most frequent problems, and move gradually into checks that require tools and time. Most drift problems are localized within the first three steps.

Step 1: Visual and Tactile Inspection (Very Low Difficulty)

Lift or jack the vehicle so that the drive wheel turns freely, then rotate it slowly by hand. Check whether the tread profile shows obvious ovality, whether there is a sloped one-sided wear surface, whether a gap or displacement mark has appeared between the tread edge and the core, and whether metal chips, sand or other debris are embedded in the tread. This step needs no tools, yet it rules out a substantial share of obvious faults, so it is worth making it a fixed action in every scheduled inspection.

Step 2: Multi-Point Diameter Measurement (Low Difficulty)

Do not measure the diameter in a single direction. Take three measurements about 120 degrees apart on each wheel, repeat each direction twice and average the result. This reveals a diameter that differs between directions on the same wheel, which is a roundness problem, and it also gives you a diameter value you can use for pairing. Then compare the diameters of the left and right drive wheels on the same machine, focusing on whether the relative difference exceeds 0.1% to 0.2%. Take the measurements with the wheels at the same thermal state; do not mix hot and cold readings.

Step 3: Radial and Face Runout Dial Indication (Medium Difficulty)

Fix a dial gauge to the frame with a magnetic base and place the contact point square against the middle of the tread, near the centre of the tread width. Rotate the wheel slowly through one full revolution and record the total indicator reading: that is the radial runout. Then reposition the contact point against the core face and rotate through one revolution again to read the face runout. When measuring, exclude the interference of bearing clearance: rotate repeatedly two or three times at the same position and watch whether the reading stays stable. If the readings differ substantially each time, check bearing preload and assembly condition.

Step 4: Multi-Point Hardness Comparison (Low Difficulty)

Use a Shore durometer to take three to five readings on the same circumference of the tread, making sure the contact point is pressed squarely and the surface is clean. Focus on three comparisons: the spread across points on one wheel, the difference between the left and right drive wheels, and the change between the current hardness and the factory value of a new wheel. The left-to-right hardness difference on one machine is best kept within 2 Shore A. Ambient temperature also affects readings, so when measuring in a cold store or other low-temperature site, record the temperature condition as the baseline for later comparison.

Step 5: Road Test Straightness and Odometry Calibration (High Difficulty)

Mark a 10 m straight path on a flat floor and let the vehicle travel along it at low speed. Measure the lateral deviation at the end relative to the start point, and the heading deviation of the vehicle, then compare them against the calculation table in Section 2.2 to judge whether the deviation exceeds expectation. A second, more convenient test is a 360-degree in-place rotation: mark the initial position and heading on the floor, command the vehicle to rotate one full turn, and observe the deviation of the return position and heading. This single action exposes both diameter difference and wheel-spacing calibration error. After every wheel change or load adjustment, run the odometry calibration again.

5. Solution Matrix

Once the root cause is localized, the response splits into two categories: temporary mitigation, used to get production running again quickly, and fundamental correction, used to prevent recurrence. The table below gives the corresponding measures and verification method for each root cause.

6. Prevention: Put Precision Indicators into Purchasing and Acceptance Specifications

The most cost-effective way to handle drift is prevention, and the vehicle for prevention is an explicit precision control checklist. Many drift disputes are hard to attribute because the purchasing specification states only a 200 mm diameter and a 75A hardness, and says nothing about runout, concentricity or pairing tolerance. The items in the table below can be used directly as the basis for quotation requests and acceptance, with the specific values adjusted to the precision grade of the vehicle and the duty.

One item on the checklist is easily overlooked: incoming sampling inspection. However good the factory inspection is, impact during transport, storage and handling can still change the effective precision of a wheel, so a quick recheck at the assembly line is worthwhile. HANKE inspects key dimensions batch by batch with a CHOTEST coordinate measuring machine at 1 µm precision at the finished-product stage, and the customer performs a sampling recheck on arrival. With gates at both ends, uncertainty is held to a low level.

7. How HANKE Approaches It: Moving Drift Risk Upstream into the Process

The previous sections described the industry-standard criteria and troubleshooting methods. This section shifts the viewpoint to the supply side: given the same six root causes of drift, which control points does HANKE set in its own manufacturing and inspection flow, and what verifiable information can a customer obtain from those control points.

7.1 One Premise: Drift Is Hard to Remedy After the Fact and Controllable Beforehand

Drift differs from failures such as slippage or abnormal noise in one obvious respect: its cost usually surfaces only after the wheels are on the vehicle. Once the machine drifts and docking fails, troubleshooting means dismantling, re-measuring, replacing parts and recalibrating, and the time and labour involved far exceed the cost of the wheel itself. That is why the cost-effective answer to this class of problem is to close the risk points before the wheel leaves the factory.

HANKE's approach can be summarized in one sentence: translate each of the six root causes listed in Section 3 into control points and inspection gates inside the process, so that every deviation capable of causing drift is identified and intercepted at the process step where it is created. The table below shows the complete mapping.

7.2 Four Control Gates: From Wheel Core to Finished Product

Gate one is core material selection and turning accuracy. The core is the precision datum of the whole wheel; the geometric relationship between its outer diameter and its face sets the starting point for concentricity and face runout downstream. HANKE drive wheel cores can be specified in 45 carbon steel, cast iron, die-cast aluminium alloy or stainless steel according to the duty, with surface treatments including painting, phosphating, zinc plating and Dacromet. Deviations created at this stage are difficult to compensate for later by adjusting the tread, so machining accuracy has to be held here.

Gate two is the three processes at the bond interface. The quality of the bond surface treatment determines whether tread and core stay concentric and synchronous. HANKE's bonding flow consists of sandblasting, adhesive application and casting. Sandblasting removes the oxide layer from the core surface and creates a roughness that favours bonding. Adhesive application forms an even, complete bond layer. Casting takes place under controlled temperature and pressure so that the polyurethane elastomer and the core are joined into one body. All three processes have work instructions that define upper and lower limits for key parameters, with batch records kept. This is both the guarantee of bond strength and the precondition for tracing back to a process step and parameter when a delamination-type anomaly occurs.

Gate three is formed dimension and roundness control. After casting, the wheel passes through finish turning to obtain its final outer diameter, width and roundness. Whether radial runout and roundness meet specification depends mainly on the machining allowance and tool-setting condition at this step. The finished wheel then goes directly to key dimension inspection, and the results are archived by batch so that the customer can cross-check them during incoming sampling.

Gate four is finished-product inspection and batch archiving. Finished-product inspection is the common exit for all six risk categories. HANKE uses a CHOTEST coordinate measuring machine with 1 µm precision to inspect key dimensions batch by batch, with data recorded and archived by batch. Hardness is sampled by batch and tied to the corresponding formulation batch. For the customer, this means the precision state of the wheel is not a verbal promise; it is data that exists, can be cross-checked and can be traced.

7.3 What Customers Can Obtain: Turning Trust into Verifiable Actions

The most effective way to judge whether a wheel supplier really manages precision is not to listen to the introduction, but to see whether it can produce data. All of the items below can be requested directly from HANKE's technical team during selection and purchasing.

7.4 One Point That Needs to Be Said Clearly

No manufacturing system should promise that deviation will never occur. That is neither realistic nor consistent with engineering practice. What deserves more attention is a different question: whether deviation is held within a range that is predictable, verifiable and traceable. When a wheel shows an out-of-specification dimension, can the supplier retrieve the inspection records for that batch, identify which process step and which parameter has drifted, and give a corrective action? That is the substance of precision management capability.

In terms of actual results, a batch of HANKE drive wheel products has run on a Mercedes-Benz automotive production line for 48 months with no after-sales record, and HANKE also supplies the bonded wheel solution for Kinco's AGV servo wheels. What records like these demonstrate is not the absence of problems, but that when precision is continuously managed as a process indicator, long-term stability is achievable.

7.5 Four Operating Condition Factors to Confirm at the Selection Stage

Controlling drift risk starts at selection, not after-sales. When HANKE's technical team works with a customer, it normally confirms four operating condition factors first: load per wheel and load distribution, travel speed and acceleration/deceleration frequency, floor material and flatness, and ambient temperature range. These four factors determine the choice of tread system and hardness direction. Where high wear resistance and traction are the priority, the Eamflex 93A system is the natural direction. In clean workshops or on fine floors where floor protection and low noise come first, the Saxflex 75A system is the natural direction. Being clear about the duty at the selection stage is far less trouble than adjusting after the wheels are on the vehicle.

8. Frequently Asked Questions

Q1: Is AGV drift always a wheel problem?

No. The causes of drift fall into three categories: navigation and calibration (sensor mounting angle, calibration parameters, floor marker position accuracy), mechanical structure (frame deformation, suspension condition, assembly error), and the executing components (wheel geometric precision and friction consistency). The wheel side is distinctive because it raises no alarm and logs no error, only a slow deviation, which is why it often gets investigated last. There are two useful decision rules. First, the deviation persists after the sensor is replaced or the vehicle is recalibrated. Second, the direction of the deviation is consistently linked to one particular wheel. When either rule applies, the wheel moves up the suspect list.

Q2: How large a left/right diameter difference starts to affect tracking?

In practice, when the relative diameter difference is held within 0.1%, the correction capability of conventional navigation covers it easily. Above 0.2%, which is about 0.4 mm on a 200 mm wheel, a visible deviation appears on open-loop straight runs. The recommendation is to write a limit of 0.1% to 0.2% for the left/right diameter difference into the assembly work instruction, and to complete pairing measurements before assembly.

Q3: What can cause drift immediately after fitting new wheels?

There are three common reasons. The left and right wheels are not from the same batch and differ in diameter or hardness. The rolling radius of the new wheel differs from the old one and the odometry calibration parameters have not been updated, which is especially pronounced when moving from a worn wheel to a new one, or from a hard tread to a soft tread. Or the wheel was not fitted true, the tightening torque is uneven, or the bearing was not pressed in correctly. A diameter pairing recheck and an odometry recalibration resolve most post-change drift at this step.

Q4: Does replacing a pair of wheels solve drift?

It depends on the root cause. If the cause really is a diameter difference or one-sided wear, replacing the pair works. If the cause lies in hub concentricity, assembly quality, floor conditions or navigation parameters, replacing wheels only relieves the symptom temporarily. The safer approach is to localize the root cause with the five-step method first, then decide whether to change wheels, change the process or change parameters, rather than replacing parts repeatedly.

Q5: With laser navigation or QR-code navigation fitted, why does the AGV still drift?

What closed-loop navigation does is measure the deviation and correct for it. It changes how the error appears, not the error itself. Wheel geometry error inside a closed-loop system turns into three symptoms: an S-shaped trajectory repeatedly corrected, faster wear on one wheel, and a lower first-attempt success rate at docking stations. The higher the navigation precision grade, therefore, the higher the precision demanded from the wheel.

Q6: Does higher hardness mean more stability, and is hardness related to drift?

Hardness relates to drift indirectly. Hardness affects the tread friction coefficient and the amount of compression under load. A left/right hardness mismatch produces different traction and different rolling radius on the two sides, which generates yaw. But higher hardness is not better: raising it sacrifices floor protection and may change noise behaviour. Hardness is also not the same as wear resistance; wear resistance is determined mainly by the formulation structure and is normally measured with an indicator such as the DIN 53516 abrasion value.

Q7: Can wheels of different hardness be mixed on the same AGV?

Mixing different hardness values between left and right wheels is not recommended. On a single machine, the left and right drive wheels should be of the same system, same batch and same hardness, with the hardness difference kept within 2 Shore A. If different hardness values are genuinely required, for example when the two sides run on different floor surfaces, differential compensation should be applied through the controller with recalibration, rather than simply mixing the wheels.

Q8: Is drift more pronounced in cold stores or outdoors?

Yes. Polyurethane hardness and dimensions change with temperature. In cold conditions the material becomes harder and the rolling radius and friction characteristics change, so in a wide temperature range such as -30°C to +70°C a single set of odometry calibration parameters may not suit the whole range. HANKE drive wheel products have a normal operating temperature range of -30°C to +70°C, with short-term exposure to +90°C, and load capacity decreases above +40°C. When calibrating precision in a cold store or outdoors, record the temperature condition alongside the result.

Q9: How should newly purchased wheels be accepted so that drift is not discovered only after fitting?

Use a three-check-one-test approach: check the diameter (multi-point measurement with left/right pairing), check the runout (radial and face dial indication), check the hardness (multi-point comparison), and then test on the vehicle (10 m straight-run deviation and 360-degree return-to-origin). Writing these four items into the acceptance form is more economical than reworking later.

Q10: How can you tell whether a wheel supplier's precision control capability is reliable?

Look at four things. Can the supplier provide key dimension inspection records for the batch, not just the nominal values printed in a catalogue? Do the critical processes - core turning, sandblasting, adhesive application, casting and finish turning - have controlled work instructions with parameter limits? Can the supplier perform an operating-condition assessment covering load, speed, floor and temperature at the selection stage, instead of pushing the same formulation at every application? And when an anomaly occurs, can it be traced back to a batch and a process step? A supplier with these four capabilities usually gives clearer answers on precision issues such as drift.

9. Summary and Recommendations

Treating AGV drift and inaccurate stopping as a precision problem that can be decomposed produces a much clearer set of conclusions. Six recommendations follow, and all of them can be put into practice directly.

First, classify before troubleshooting. Separate one-sided drift, S-shaped paths, inaccurate stopping and rotation error, and use the symptom-to-cause mapping table to set the direction, instead of immediately replacing sensors and navigation modules.

Second, take diameter consistency seriously. Keep the left/right diameter difference on one machine within 0.1% to 0.2%, and measure and pair every wheel before assembly. This has one of the strongest returns on effort of any measure available.

Third, put runout and concentricity into the purchasing specification. Stating diameter and hardness alone is not enough to constrain wheel precision. Radial runout, face runout and concentricity should be written in as well, with the inspection stage named.

Fourth, understand the limits of closed-loop correction. The navigation system corrects the trajectory, not the wheel error. The combined result is an S-shaped path, faster wear on one side and a lower docking success rate.

Fifth, establish a scheduled inspection routine. Put visual inspection, diameter measurement, runout dial indication, hardness comparison and the road test into the inspection form, carry it out on a fixed cycle, and keep the records, so that deviation is found before it becomes a failure.

Sixth, include the supplier's process control capability in the evaluation. Batch inspection records, work instructions with parameter limits for critical processes, operating-condition assessment capability and batch traceability are the four hard indicators that keep drift risk out of the vehicle in the first place, and they reflect real capability far better than the nominal figures in a catalogue.

For wheel selection, precision is not an abstract concept. It is an engineering indicator built from diameter tolerance, runout, concentricity, hardness consistency and bond stability. Once these indicators are under control, the walking precision of an AGV stops being a topic that invites repeated argument and becomes a calculation with a definite answer.