SLAM Navigation: How It Works and Its Applications in AGVs and AMRs

Warehouse rack positions, transport routes, and workstation layouts may change as operational needs evolve. For AGVs that rely on fixed navigation markers, a change that makes the original route unsuitable may require magnetic tape, QR codes, or other markers to be repositioned.

SLAM Navigation uses environmental features to help robots determine their position, reducing their reliance on fixed markers. It is widely used in industrial AMRs as well as AGVs, although localization performance still depends on the warehouse environment and sensor configuration.

What Is SLAM Navigation?

SLAM stands for Simultaneous Localization and Mapping. It allows a robot to build a map of its surroundings while estimating its own position as it explores an environment.

Imagine an AMR or AGV entering a warehouse for the first time, before a map of the facility has been created. As the vehicle moves, its sensors gather information about nearby racks, walls, and columns. The robot estimates how far it has traveled and how its direction has changed while gradually assembling this information into a map.

Localization and mapping are closely connected. The robot needs an approximate position estimate to place newly observed features in the right locations on the map. Once the map takes shape, those features can in turn help refine its position estimate.

After mapping, AMRs and AGVs can generally use the existing map for everyday localization, working with path-planning and motion-control systems to carry out transport tasks. If people or obstacles appear along the way, whether the vehicle can reroute also depends on functions such as obstacle detection and local path planning.

SLAM vs. LiDAR: What Is the Difference?

In technical materials about AMRs and AGVs, SLAM and LiDAR are often discussed together, which can make them sound like two competing navigation technologies.

In fact, SLAM is a method for localization and mapping, while LiDAR (Light Detection and Ranging) is a sensor technology used to collect distance information about the surroundings.

The two can work together. LiDAR scans the environment, and a SLAM algorithm uses the scan data to localize the robot and build a map. This is known as LiDAR SLAM.

SLAM can also use images captured by cameras for localization and mapping, an approach commonly called Visual SLAM.

Likewise, an AMR or AGV equipped with LiDAR does not necessarily use SLAM. The sensor can also support localization against an existing map, obstacle detection, and other functions.

How Does LiDAR SLAM Work?

Consider a warehouse AMR or AGV being prepared for operation. Before it starts regular transport tasks, the site typically needs to be mapped.

As the vehicle travels along warehouse aisles, LiDAR continuously scans the nearby racks, walls, and columns, recording their distance and direction relative to the vehicle.

After the vehicle moves, the same row of racks appears in a different position in the next scan. The SLAM algorithm compares successive scans, identifies corresponding environmental features, and estimates how far the vehicle has moved and how its orientation has changed. This process commonly involves scan matching.

As the vehicle continues moving, the system combines data collected from different positions to build a warehouse map. For a more stable motion estimate, a real-world system may also combine data from wheel odometry and an inertial measurement unit (IMU).

Even with these additional inputs, position estimates can accumulate errors as the vehicle travels.

For example, if an AMR or AGV completes a loop through several aisles and returns to a previously visited location, the system may use recognized features to correct the estimated trajectory and map. This is known as loop closure.

Once the map has been created and checked, the vehicle can begin normal operation.

When it receives a transport task, it continuously collects laser scan data and matches those observations against the existing map to estimate its current position. The navigation system then uses that position information to guide the vehicle along its planned route.

This means an AMR or AGV generally does not need to rebuild the entire warehouse map for every trip. If the environment changes substantially, or if the map no longer represents the site accurately, the map may need to be updated or rebuilt.

Mapping and localization can use either two-dimensional or three-dimensional laser scan data. Common industrial mobile robot approaches include 2D LiDAR SLAM and 3D LiDAR SLAM.

2D LiDAR SLAM primarily uses scans taken in a plane and is common in indoor ground-based mobile robots. 3D LiDAR SLAM uses three-dimensional spatial information, allowing it to recognize environmental features at different heights.

The right choice depends on the facility and its localization requirements; a more complex 3D mapping approach is not automatically the better option.

SLAM Navigation in AMRs and AGVs

AMRs and AGVs using SLAM navigation can handle pallet transport, rack movement, and material delivery in warehouses and factories. A SLAM AGV carrying pallets, for example, has different docking requirements from an AMR moving racks, even though both rely on SLAM for localization.

Warehouse Pallet Transport

Moving pallets between receiving, storage, and shipping areas is a common task for forklift AGVs.

A forklift AGV can use the warehouse map to locate itself and follow a planned route to the pickup point. Reaching the correct storage location, however, is only the first step. Its forks must also align accurately with the pallet openings before the load can be picked up safely.

For this reason, some forklift AGVs use additional load-detection or positioning functions to improve accuracy when picking up and placing pallets.

Goods-to-Person Picking and Rack Transport

In a goods-to-person system, a lifting AGV can travel underneath a storage rack and carry the entire rack to a manual picking workstation. After an employee finishes picking, the robot returns the rack to its designated storage area.

One distinctive feature of this application is that the racks themselves may move frequently.

If a robot relies heavily on movable racks as localization features, changes in their positions can make parts of the map inconsistent with the actual environment. Mapping and localization therefore need to make appropriate use of more stable features, such as walls and columns, together with the system’s map-maintenance approach.

In warehouses with tightly packed racks, repetitive aisle layouts and similar-looking features also need to be considered so that the robot can reliably distinguish one location from another.

In-Plant Material Delivery

In factories, AMRs and AGVs can transport bins, pallets, and other materials between raw-material warehouses, production lines, and temporary storage areas.

When a production line is rearranged, pickup and delivery locations may change as well. Vehicles using SLAM navigation can have their transport tasks and travel routes adjusted to the new workstation locations without relaying fixed navigation markers. If environmental features change substantially, the existing map should also be checked to confirm that it is still suitable.

If new equipment blocks an original aisle, vehicle access conditions and localization performance also need to be reassessed.

Advantages and Limitations of SLAM Navigation

Less Installation and Maintenance of Fixed Navigation Markers

In warehouses with long transport routes and many aisles, SLAM navigation can reduce the need to install magnetic tape, QR codes, and other fixed localization markers, as well as the work involved in maintaining them.

Easier Transport Route Changes

When a warehouse adds workstations or changes transport tasks, destination points and travel routes can be adjusted to match the new requirements. If the physical environment also changes, the map may need to be checked or updated.

Although fixed navigation markers generally do not have to be laid again, a new route must still meet vehicle access, safety clearance, and workstation docking requirements.

Localization Depends on Environmental Features

In long warehouse aisles lined with highly similar racks, SLAM may have difficulty distinguishing locations based on environmental features. Large open areas with few distinctive features can also make localization more challenging.

For vehicles using LiDAR SLAM, glass, highly reflective surfaces, dust, and sensor occlusion by goods can affect the quality of laser scan data.

Environmental Changes Can Affect Map Matching

Warehouses see new pallets placed in aisles, goods moved around, and temporary equipment occupying passageways every day. A suitable localization system can often handle limited dynamic changes, but if many features change for an extended period, the original map may no longer be suitable.

For warehouses that frequently move racks or equipment, it is important to consider which environmental features will remain stable over time and whether the map needs periodic inspection and maintenance.

If you are planning to deploy SLAM-guided AMRs or AGVs in your warehouse or factory, Coolyne can provide a free automation feasibility assessment and ROI analysis based on your facility layout, material-handling needs, and operating processes, helping you choose an appropriate automation solution.

Contact Coolyne for a Free Assessment