Smart warehouse / Article
SLAM Navigation Explained: How an AMR Maps and Drives Your Floor
SLAM stands for simultaneous localisation and mapping, and it is the reason a mobile robot needs nothing installed in your floor. The robot builds a map of your building from its own sensors and works out where it is inside that map at the same time, continuously. In laser SLAM a LiDAR scanner measures distances to permanent structure, walls, columns, racking, and matches what it sees against the stored map to fix its position. The route then lives in software rather than in the slab, so re-slotting a zone is a configuration change instead of a construction project. QR markers are added where docking precision matters. This article explains how each method works, where it fits, and the floor conditions that decide whether it works at all.
The idea in four parts
01 / The problem SLAM solves
A robot has to answer two questions at once
Any vehicle that drives itself has to know where it is. The older answer to that question was to put the answer in the building: lay magnetic tape, bury a wire in the slab, or survey reflective markers onto columns, then let the vehicle follow what it finds. It works, and it is still the right answer for some applications, but it has one structural consequence. The route is a physical asset. Changing it means changing the floor.
SLAM inverts that. Instead of reading a path someone installed, the robot builds its own map of the space and locates itself within it, both at the same time, which is where the name comes from. As one vendor describes the mechanism, AMRs use sensor fusion, combining data from LiDAR sensors, laser scanners, cameras and onboard navigation in real time to determine position, create digital maps, detect obstacles and optimise routes, with no magnetic strips or floor guides required.
The chicken-and-egg quality of doing both jobs simultaneously is what makes SLAM a genuine piece of engineering rather than a marketing word. To know where you are you need a map. To build a map you need to know where you were when you took each measurement. SLAM resolves the circularity by doing both incrementally and correcting as evidence accumulates, which is why a robot driven once through a building comes out the other side with a usable map of it.
What this buys you operationally
Three things, and all of them are about change rather than speed. Layout freedom: when racks move or a staging area is added, the map can simply be updated without modifying the warehouse floor. Traffic tolerance: an AMR continuously analyses its surroundings and calculates an alternative route, where an AGV meeting an obstacle typically stops and waits until the path clears. And a shorter install, because there is no slab cutting, epoxy work or reflector survey standing between the purchase order and a working route.
02 / How laser SLAM works
The building is the reference
Laser SLAM, sometimes called natural feature navigation, is the workhorse method in industrial settings and the one on our machines. The sequence is simple enough to describe in four steps, which is worth doing because it makes the failure modes obvious.
Two properties follow directly from step one, and both matter on a Malaysian floor. Laser SLAM does not depend on light, so it works day and night without ambient light and requires nothing glued to the floor, which suits a warehouse that runs a dark night shift or has uneven lighting between zones. And it needs geometry to look at, which is why the same source notes the method is sensitive to distance and environment, best suited to workshops and warehouses where machinery and equipment are permanently fixed, with added reflectors needed in unfavourable environments.
That second property defines the honest limits, and the same source is specific about them: the method becomes unsuitable when the floor area changes by more than roughly half, or where a corridor runs more than about 100 metres with nothing distinctive along it. A very long, very uniform aisle is a hard problem for scan matching because every position along it looks alike. This is not a defect, it is a design constraint, and it is one of the things a floor study is looking for before anything is quoted.
03 / The methods compared
Four ways to tell a robot where it is
The choice is not really about technology preference. It is about how much your routes change, how precise the docking has to be, and how much you are willing to install in the building.
| Method | Reference used | Floor infrastructure | Strength | Limit |
|---|---|---|---|---|
| Laser SLAM | Permanent structure: walls, columns, racking | None | Route lives in software; works without ambient light; reroutes around obstacles | Needs stable geometry; long featureless corridors and heavily changing layouts are hard |
| QR code navigation | Markers applied to the floor, read by downward cameras | Markers, applied and maintained | Exceptional repeatable positioning; common in very high-throughput environments | Markers must be re-laid when layout changes; wear and contamination matter |
| Visual SLAM | Natural features or floor texture seen by cameras | None | No added infrastructure, fine positioning accuracy | Depends on visual conditions and floor texture quality |
| Fixed-path AGV guidance | Magnetic tape, buried wire or laser reflectors | Tape, wire or reflector survey | Cheap, predictable, very mature | Any route change is a physical change; an obstruction stops the vehicle |
Scroll the table sideways on a phone / Characteristics from published vendor and integrator material
Industry commentary summarises the first two rows neatly: in practice, SLAM offers greater flexibility while QR code navigation delivers maximum positioning precision, with QR systems using downward-facing cameras that continuously scan floor markers and requiring maintenance whenever layouts change. A guidance comparison makes the same distinction structurally, grouping magnetic tape, buried rods, RFID tags, QR codes and barcodes together as low-cost solutions that require predefined infrastructure, against LiDAR navigation which can be reflector-based or natural, where natural LiDAR eliminates reflectors by interpreting the existing environment.
It is also worth noting that the categories are blurring. The same comparison observes that most navigation systems other than fixed-path options can be adapted for both AGVs and AMRs, with barcode grids and SLAM especially versatile. So the useful question in a vendor conversation is not whether a machine is called an AMR. It is what it references, what you have to install, and what happens when something is in the way.
04 / The hybrid
QR where precision pays, SLAM everywhere else
A route is not homogeneous. Crossing an open floor is one problem; sliding forks under a pallet, docking into a conveyor handover or seating onto a charger is a different one. The first rewards flexibility. The second rewards repeatable accuracy at one specific spot, every time, for years.
Hybrid navigation resolves that by using each method where it is strongest. As one vendor describes their implementation, additional QR codes placed on the floor are scanned by onboard cameras to support navigation and precise positioning, which is particularly effective in changing environments or applications requiring high positioning accuracy. This is exactly the configuration on our Latent Lift AMRs: QR and Laser SLAM navigation together, with 360-degree obstacle avoidance, carrying payloads from 60kg to 3,000kg across pallets, racks and trolleys. Our Forklift AMRs run Laser SLAM, carrying 1,500kg to 3,000kg with a lifting height of 1,000 to 4,000mm for pallets and cages.
The practical benefit is that markers are used sparingly. You are not tiling a warehouse with QR codes, you are placing them at the handful of positions where a few millimetres decide whether a fork enters a pallet cleanly. Everywhere else the building is the reference, so when a zone gets re-slotted next year the map changes and the markers stay where the work is. That is a maintenance argument as much as an engineering one, and it is the reason we would not design a whole-floor marker grid for a typical Malaysian manufacturing site.
05 / What the floor has to give
SLAM removes infrastructure. It does not remove requirements.
The most useful sentence in the published material on this subject is a corrective one. SLAM natural feature navigation often requires less fixed infrastructure than tape, wire or reflector-based systems, but it still requires mapping, environmental consistency, traffic planning and a suitable floor condition. Nothing installed in the slab is not the same as nothing to prepare.
Guidance on selecting a navigation method makes the same point as a checklist, advising that you evaluate not only the physical space but temperature, dust levels, floor quality and lighting conditions, and match the technology to whether the application needs precision, flexibility or cost-efficiency. Every one of those is something we look at during a floor study, and every one of them is cheaper to find with a tape measure than during commissioning.
One prerequisite sits outside navigation entirely and is worth stating plainly. A robot that navigates perfectly still collects from the location your record names. If your stock count cannot be trusted, flawless navigation delivers the wrong tote promptly. Navigation is a movement problem; knowing what is where is a data problem, and it comes first.
06 / Commissioning
Mapping is one afternoon. Annotating is the work.
Building the map is the part that looks impressive and takes the least time: the robot is driven through the building and comes out with a geometric model of it. What takes the time is everything the map cannot infer, because a map knows shapes and not intentions.
Annotation is where a floor becomes navigable. Routes and preferred directions. One-way segments in aisles too narrow to pass. Speed zones near doorways, blind corners and pedestrian crossings. Docking positions, with markers where precision matters. Charging positions and the charging strategy, since opportunity charging in gaps behaves very differently from taking a unit out of service. Door and lift interfaces. Yield rules where robots meet people. Priority rules for what may legitimately wait during the dispatch wave.
Which is why our sequence puts design and simulation between the floor study and the build, and why commissioning is done by the people who designed it. We design the layout and simulate throughput so the system is proven before install, then install, commission, integrate with your WMS and train your team. Simulation is where you learn that one doorway is the fleet's real ceiling, or that two routes crossing at a single point costs more throughput than a slower robot would have. Those are annotation decisions, and they are far cheaper to make in a model. The fleet-count arithmetic that sits on top of this is in the movement layer comparison.
07 / When not to use it
SLAM is a premium paid for change
An honest article about a technology should say where it loses. SLAM navigation and the machines that carry it cost more per unit than fixed-path guidance, and published comparison puts the trade plainly: AMRs generally require a higher initial investment per unit, but significantly less warehouse infrastructure. You are buying the ability to change the route without changing the building. If the route will never change, you are paying for an option you will not exercise.
So three situations point away from it. A single high-volume flow that runs constantly along a fixed line, where a conveyor is usually cheaper and mechanically simpler. A completely static environment where guidance precision matters more than flexibility, where a reflector or tape-guided vehicle can be the more economical answer. And a floor whose real constraint is storage rather than movement: if the building is full and nobody is walking far, no navigation technology addresses that, and the answer is density instead. We set out that test in the storage capacity article.
Where SLAM clearly earns its premium is the situation most Malaysian manufacturing and 3PL floors are actually in. Routes that change with the product mix. Shared aisles with forklifts and people. A building you cannot take offline to cut a slab. And a three-year plan nobody can describe precisely. In that shape, a route that lives in software is not a luxury. It is the only version that survives the plan changing.
08 / On your floor
We map the real building, not a drawing of it
Our sequence is fixed because the expensive mistakes all happen before installation. A floor study first: flows, volumes, SKU profile, real distances, aisle widths, doors, lifts, floor condition and where the queues form today. Then design and simulate, so routes, fleet size and throughput are proven in a model against your own volumes. Then deploy: install, commission, integrate with your WMS and train your team.
Navigation on our Latent Lift AMRs is QR and Laser SLAM with 360-degree obstacle avoidance, and Laser SLAM on the Forklift AMRs, so nothing is installed in your slab to carry a route. Payloads run from 60kg to 3,000kg on the Latent Lift and 1,500kg to 3,000kg on the Forklift AMR, which lifts from 1,000 to 4,000mm. For cleanroom work, our Mobile Manipulator AMR handles FOUP, FOSB, reticle and magazine transfer and meets Cleanroom Class 1.
CODETRACE integrates on site from Shah Alam in Selangor and Batu Kawan in Penang, so the team that maps your floor is the team that commissions the fleet and the team you call when a route needs changing. Where the movement layer connects into production rather than only dispatch, the same discipline extends upstream into factory automation.
Nothing in the slab. The route lives in software, so it changes when your floor does.
FAQ / SLAM navigation
Questions, answered.
01What is SLAM navigation on an AMR?
SLAM stands for simultaneous localisation and mapping. The robot builds a map of your building from its own sensors and works out where it is inside that map at the same time, continuously. In laser SLAM a LiDAR scanner sweeps the surroundings, measures distance to permanent structures such as walls, columns and racking, and matches the live scan against the stored map to fix position. Because the reference is the building itself, no tape, wire, reflectors or rails are installed in the floor. The route lives in software, so changing it is a configuration change rather than a construction job.
02Why does an AMR need no floor rails or tape?
Because it navigates by what is already there. A fixed-path AGV needs magnetic tape, buried wire or reflective markers because the guidance is external to the vehicle: the path is in the floor and the vehicle follows it. A SLAM robot carries its own map and its own sensors, so the path is in software. Practically this means no slab cutting, no epoxy work, no reflector survey, and layout changes handled by updating the map rather than modifying the floor. It also means a blocked route is a reroute rather than a stopped vehicle waiting for the obstruction to clear.
03What is the difference between laser SLAM and QR code navigation?
They trade flexibility against precision. Laser SLAM references permanent structure, needs nothing added to the floor, and works day or night without ambient light, but it is sensitive to environments that change substantially or to very long featureless corridors. QR navigation uses downward-facing cameras reading floor markers and delivers exceptional repeatable positioning, which is why it is common in very high-throughput environments, but it requires markers to be applied and maintained whenever the layout changes. In practice SLAM offers greater flexibility while QR delivers maximum positioning accuracy.
04Why use both QR and laser SLAM on the same robot?
Because the two jobs on a route have different requirements. Travelling across an open floor rewards flexibility, since the route may change and obstacles appear. Docking into a charger, a conveyor handover, a rack pick-up or a station needs repeatable accuracy at a specific point. A hybrid setup uses laser SLAM for travel and QR markers at the points where precision pays, which is why CODETRACE Latent Lift AMRs run QR and Laser SLAM navigation with 360-degree obstacle avoidance, while Forklift AMRs use Laser SLAM for pallet and cage handling.
05What floor and building conditions does SLAM need?
SLAM needs less infrastructure than tape or reflectors, but it still needs mapping, environmental consistency, traffic planning and a suitable floor condition. Three things matter most in practice. Feature stability: the robot localises against permanent structure, so a floor where more than about half the visible layout changes regularly is difficult, and very long featureless corridors give the scanner little to match against. Floor quality: slope, joints, ramps, gratings and standing water all affect traction and wheel odometry. Traffic: shared aisles with forklifts and pedestrians need width, sightlines and rules rather than only sensors.
06Does SLAM cope with a warehouse that keeps changing?
Up to a point, and the distinction is between transient and structural change. Pallets left in an aisle, a person crossing, a trolley parked in the wrong place: these are obstacles, and the robot reroutes around them. Racking relocated, a new mezzanine, a whole zone re-slotted: that is a map change, and the map is updated without touching the floor. What genuinely defeats laser SLAM is an environment where the permanent reference itself is not permanent, and that is where QR markers or added reflectors earn their place.
07How is a SLAM robot commissioned on a real floor?
The robot is driven through the building to build the map, then the map is annotated with the things the map cannot infer: routes, priorities, one-way segments, speed zones, docking positions, charging points, doors and lifts, and where robots must yield to people. Then it is tuned against the real floor. Our sequence puts a floor study before any of this, then layout design and throughput simulation, then install, commission, integrate with your WMS and train the team, supported from Shah Alam in Selangor and Batu Kawan in Penang.
08Is SLAM the right choice for every application?
No. If the same load moves along the same path continuously and the route will never change, a conveyor is usually cheaper and simpler. If guidance precision matters more than flexibility and the environment is completely static, a fixed-path AGV can be the more economical answer. SLAM earns its premium where the route changes, where robots share space with people and forklifts, and where you do not want to modify the building to install a route. The honest test is whether change has value on your floor.
Sources / Every claim in this article
Where the detail comes from
Sources are listed by what they are rather than by brand name. All of them are robot vendors, integrators or trade publications writing about a category they sell into, so read each characterisation as one drawn by a company with a product on one side of the comparison. The mechanism of SLAM itself is standard robotics and belongs to nobody.