Autonomous Mobile Robots (AMRs): Functions and Deployment
Autonomous mobile robots represent a distinct and rapidly expanding category within the broader robotic systems landscape, defined by their ability to navigate dynamic environments without fixed infrastructure or continuous human control. This page covers how AMRs are classified, the sensor and software mechanisms that enable self-directed movement, the operational environments where AMRs are deployed, and the decision criteria that determine when an AMR is appropriate versus alternative automation approaches. Understanding these boundaries is essential for engineers, facility managers, and procurement teams evaluating mobile automation.
Definition and scope
An autonomous mobile robot is a self-propelled machine capable of perceiving its environment, planning a path, and executing movement to complete tasks without relying on physical guidance systems such as magnetic floor tape, reflective markers, or embedded wires. This distinguishes AMRs from their predecessor category, automated guided vehicles (AGVs), which require pre-installed infrastructure to define travel routes.
The International Organization for Standardization addresses mobile robot safety under ISO 3691-4, which specifies safety requirements for industrial trucks, including automated functions. The Association for Advancing Automation (A3) and the Robotic Industries Association (RIA) have published supplemental guidance classifying AMRs as a subset of mobile robots that achieve spatial awareness through onboard sensing rather than environmental modification.
AMR payloads range from under 100 kilograms for light-duty inventory transport units to over 1,500 kilograms for heavy-load variants used in automotive and aerospace manufacturing. Speed limits for industrial AMRs in occupied workspaces are typically governed by site-specific risk assessments under OSHA 29 CFR 1910.217 and the broader ANSI/RIA R15.08 standard, the first US-specific safety standard developed specifically for industrial mobile robots. Facilities integrating AMRs into shared human workspaces should also consult the regulatory context for robotic systems that governs mixed-occupancy deployments.
How it works
AMR operation depends on three integrated subsystems: perception, localization, and motion planning.
Perception is achieved through a combination of sensor modalities. Lidar (light detection and ranging) is the most common primary sensor for obstacle detection, with 2D lidar units scanning horizontal planes and 3D lidar units building volumetric maps. Depth cameras, ultrasonic sensors, and infrared arrays supplement lidar by detecting low-lying obstacles, transparent surfaces, or objects below the lidar scan plane.
Localization refers to the robot's ability to determine its position within a facility map. Two dominant methods are used:
- Simultaneous Localization and Mapping (SLAM) — the robot constructs a map of its environment in real time while tracking its own position within that map. SLAM is infrastructure-free and allows deployment in environments where layout changes frequently.
- Natural feature navigation — the robot matches observed environmental features (walls, columns, rack structures) against a pre-built reference map stored onboard or on a fleet management server.
Motion planning translates the destination instruction and current position into a collision-free path. Modern AMR motion planners use dynamic replanning algorithms — distinct from the static route tables used by AGVs — allowing the robot to reroute around unexpected obstacles in under 200 milliseconds on commercial-grade hardware.
Fleet management software coordinates multiple AMRs simultaneously, managing traffic, charging cycles, task queuing, and exception handling. NIST's Engineering Laboratory Robotics Program has published research on performance metrics for autonomous navigation systems, including path deviation and task completion rate as standardized evaluation criteria.
Common scenarios
AMRs are deployed across five primary operational contexts:
- Warehouse and fulfillment — AMRs transport goods-to-person picking stations, move pallets between storage and staging areas, and support inventory cycle counts. Deployments in this category typically involve fleets of 20 to 300 units operating across shifts. More detail on this application appears in warehouse and logistics robotics.
- Manufacturing intralogistics — AMRs deliver components to assembly lines, remove finished subassemblies, and synchronize with production scheduling systems. Automotive plants with high-mix production schedules benefit from AMR flexibility versus fixed conveyor systems.
- Healthcare and hospital logistics — AMRs transport medications, lab specimens, linens, and meal trays between hospital departments. The FDA regulates software functions in medical-use robots under 21 CFR Part 820, and facilities using AMRs in patient-care areas must complete site-specific hazard analyses.
- Retail and hospitality — AMRs perform automated inventory scanning in retail stores and deliver room service or supplies in hotels. These environments present navigation challenges due to unpredictable pedestrian density and irregular floor surfaces.
- Airport and transit hub logistics — AMRs handle baggage sorting, cleaning, and supply replenishment in controlled zones, operating under facility-specific access protocols and applicable FAA or TSA site requirements.
Decision boundaries
Selecting an AMR over an AGV, a conveyor system, or a collaborative robot (cobot) depends on four structured criteria:
Infrastructure flexibility — AMRs require no floor modifications; AGVs require tape, wire, or reflective marker installation. Facilities with frequently changing layouts, multiple tenants, or floor surfaces that preclude embedded infrastructure favor AMRs.
Path variability — Fixed-route tasks with high throughput (e.g., a single loop between two fixed stations running 24 hours) may be served more efficiently by AGVs or conveyors at lower per-unit cost. Tasks involving conditional routing, dynamic destination selection, or multi-stop sequences are better matched to AMR capabilities.
Obstacle density and human co-presence — AMRs are designed for dynamic, human-occupied spaces. ANSI/RIA R15.08 defines three risk-based operating modes — free navigation, reduced speed, and stopped — triggered by proximity detection. AGVs in human-shared spaces require additional physical guarding or exclusion zone enforcement.
Payload and cycle requirements — AMRs with payloads above 1,000 kilograms are available but represent a smaller product segment. Applications requiring precise positioning repeatability under ±1 millimeter — such as machine tending at CNC equipment — may require a fixed-arm robot or a mobile manipulator rather than a standalone AMR.
The artificial intelligence in robotic systems domain intersects directly with AMR decision-making, as machine-learning-based perception systems are increasingly embedded in commercial AMR platforms to handle previously problematic scenarios such as reflective flooring, low-light conditions, and partial occlusion of obstacles.