AMR and Mobile Robotics Terminology

A mobile robot that follows a fixed, predefined path within a facility, guided by infrastructure embedded in or applied to the floor—magnetic tape, painted lines, embedded wires, or reflective markers. AGVs cannot dynamically reroute around obstacles; if the path is blocked, the AGV stops and waits until the obstruction is cleared. AGVs are the predecessor technology to AMRs and remain in use for simple, repetitive transport tasks where the path does not change and throughput requirements are predictable. The key distinction from AMRs: AGVs require physical infrastructure to define their path; AMRs navigate using onboard sensors and software without physical path infrastructure.

A mobile robot that navigates a facility independently using onboard sensors (LiDAR, cameras, depth sensors), a digital map of the facility, and navigation software that plans and executes paths in real time without requiring fixed infrastructure (magnetic tape, wires, or rails) on the floor. AMRs dynamically detect and avoid obstacles, reroute around congestion, and adapt to changes in the facility layout. In warehouse applications, AMRs transport totes, shelves, pallets, or carts between zones (picking, packing, shipping, storage) either autonomously or in collaboration with human operators. AMR fleet sizes in mid-market operations typically range from 10 to 75 robots, managed by a fleet management system that assigns tasks, coordinates traffic, and optimizes throughput.

The electronic control system embedded in the AMR’s battery pack that monitors and manages the battery’s state of charge, state of health, cell voltage balancing, temperature, charge/discharge current, and fault detection. The BMS protects the battery from conditions that reduce lifespan (overcharge, over-discharge, excessive temperature, cell imbalance) and communicates the battery’s status to the fleet management system, which uses the data to schedule charging and predict battery replacement. A BMS failure can result in undetected cell degradation that reduces the robot’s runtime from the expected 8–10 hours to 4–5 hours, effectively removing the robot from a full shift of productive operation.

A charging station that delivers electrical power to the AMR through physical contact between metal pins on the charging station and corresponding contacts on the robot. The robot docks with the station by aligning the contacts (typically using a V-shaped guide or magnetic alignment) and charging begins automatically. Contact charging provides high charging power (typically 400W–1,500W, enabling charge times of 45–90 minutes from 10% to 90% state of charge), is mechanically simple, and is the most common charging method for warehouse AMRs. The contact surfaces require periodic cleaning to maintain charging efficiency and prevent resistance heating.

A charging station that delivers electrical power to the AMR through electromagnetic induction across an air gap, without physical contact between the robot and the station. The robot positions itself over an inductive pad embedded in the floor or mounted on a pad surface, and power transfers through coupled coils. Wireless charging eliminates the mechanical alignment requirement and contact wear of pin-based charging, but operates at lower efficiency (typically 85–92% vs. 95–98% for contact charging), generates more heat, and typically delivers lower charging power, resulting in longer charge times. Wireless charging is advantageous for environments where contact contamination (dust, moisture, debris) would degrade pin-based charging reliability.

A picking methodology in which an AMR carries a cart or rack with multiple order containers (typically 4–12 positions), and the operator picks items for multiple orders simultaneously during a single pass through the pick zone. The AMR’s fleet management system assigns a cluster of orders with overlapping SKU requirements to minimize the travel distance per pick. The operator picks each item and places it in the designated order container, guided by the AMR’s display or pick-to-light indicators on the cart. Cluster picking with AMRs typically achieves 150–250 picks per hour, compared to 80–120 for manual discrete picking, because the AMR optimizes the pick path and the multi-order consolidation reduces the number of visits to each pick location.

An AMR designed to work alongside human operators in a shared picking workflow, where the robot handles the transport (carrying totes, carts, or shelves between zones) and the human handles the manipulation (physically picking items from shelves and placing them in order containers). The robot and the operator share the same workspace without physical barriers, with the robot’s safety systems (per ISO 3691-4 and ANSI B56.5) ensuring safe operation in proximity to humans. Collaborative picking is the most common AMR deployment model for mid-market warehouses because it combines the robot’s transport efficiency with the human’s dexterity and judgment for the pick task.

The fleet management system’s algorithms and strategies for preventing, detecting, and resolving traffic congestion when multiple robots converge on the same area simultaneously. Congestion management techniques include zone capacity limits (restricting the number of robots allowed in a zone at any time), dynamic rerouting (redirecting robots to alternate paths when the primary path is congested), task sequencing (staggering task assignments to avoid simultaneous arrival at popular locations), and speed reduction (slowing robots approaching congested areas to reduce the severity of congestion events). Congestion is the primary non-linear scaling factor in AMR fleet deployments: throughput per robot decreases as fleet size increases because more robots compete for the same aisle space, intersections, and charging stations.

A condition in which two or more robots are mutually blocked—each robot’s path is obstructed by another robot that is itself waiting for the first robot to move—creating a gridlock that none of the involved robots can resolve independently. The simplest deadlock involves two robots facing each other in a single-width aisle where neither can pass. Complex deadlocks involve 3+ robots in a circular wait pattern. Deadlocks halt the affected robots’ tasks and, if they occur at critical intersections or aisle entries, can cascade to block additional robots and reduce fleet-wide throughput. See deadlock avoidance algorithm.

The fleet management system’s logic for preventing deadlock conditions before they occur, rather than resolving them after they form. Techniques include reservation-based path planning (each robot reserves the path segments it intends to traverse, and the system prevents another robot from reserving a conflicting segment), priority assignment (robots with higher-priority tasks or less flexibility receive path priority), and look-ahead conflict detection (the system evaluates the projected paths of all active robots and identifies potential conflicts before the robots reach the conflict point). Effective deadlock avoidance is a key differentiator between mature and immature fleet management systems: a mature system prevents 99%+ of potential deadlocks through proactive planning; an immature system reacts to deadlocks after they form, requiring manual intervention or recovery protocols that cost 30–60 seconds per event.

The centralized software platform that manages, coordinates, and optimizes an AMR fleet’s operations. The FMS performs five core functions: task allocation (assigning transport or picking tasks to specific robots based on location, battery status, and task priority), path planning (calculating the optimal route for each robot from its current position to the task’s destination), traffic management (coordinating the movements of all robots to prevent congestion, deadlocks, and collisions), battery management (monitoring each robot’s state of charge and scheduling charging to maintain fleet availability), and performance analytics (tracking fleet throughput, utilization, exception rates, and identifying optimization opportunities). The FMS is the software that converts a collection of individual robots into a coordinated fleet—two vendors with identical robot hardware produce 20–30% different fleet throughput depending on the FMS’s intelligence.

The analytical process of determining the number of robots required to meet the buyer’s throughput target, accounting for the non-linear relationship between fleet size, throughput, and congestion. Fleet throughput does not scale linearly with robot count: the first 10 robots in a zone may each contribute 10 tasks per hour; robots 11–20 may each contribute 8 tasks per hour (due to increased congestion); robots 21–30 may each contribute only 5 tasks per hour (due to severe congestion at intersections and in narrow aisles). Fleet size optimization models this diminishing-return curve to identify the fleet size that achieves the throughput target at the lowest cost per task. Formula (simplified): optimal fleet size = f(throughput target, tasks per robot per hour, congestion factor, charging downtime).

An AMR operating mode in which the robot follows the operator through the facility at a set distance, carrying the operator’s picked items on its platform or cart. The operator walks the pick path and places items on the robot; when the robot’s capacity is reached or the pick route is complete, the robot autonomously transports the picked items to the packing station or staging area while the operator begins a new pick route with a different robot. Follow-me mode eliminates the operator’s transport time without requiring the operator to adapt to the robot’s path—the robot adapts to the operator’s path. This mode is commonly used in the early adoption phase because it requires minimal operator training and workflow change.

A software-defined geographic boundary on the facility map that restricts or permits robot access to specific areas. Exclusion geofences prevent robots from entering areas that are unsafe (pedestrian-only zones, loading docks during active truck operations), operationally inappropriate (manual picking zones where robots would interfere with operators), or physically incompatible (areas with floor conditions the robot cannot navigate). Inclusion geofences define the boundaries of the robot’s permitted operating area. Geofences are configured in the FMS and can be modified in real time to respond to changing operational conditions (opening or closing zones during shift changes, maintenance activities, or emergency situations).

A navigation method in which the AMR travels along a predefined grid of paths marked on the warehouse floor (typically using QR codes, barcodes, or magnetic markers at grid intersections). The robot moves from grid point to grid point in straight lines and 90-degree turns, following the grid’s geometry rather than calculating free-form paths. Grid-based navigation is simpler and more deterministic than SLAM-based navigation (the robot’s position is precisely known at each grid marker) but less flexible (the robot can only travel along grid paths, and path changes require physical modification of the floor markers). Grid-based systems are common in goods-to-person shelf-carrying robot deployments where the storage pods are arranged on a grid.

A battery design that allows the operator or an automated system to remove the AMR’s depleted battery and insert a fully charged battery without tools and without powering down the robot’s control system, returning the robot to service within 1–3 minutes. Hot-swap batteries eliminate the charging downtime that reduces robot availability: instead of a robot sitting at a charging station for 45–90 minutes, the battery is swapped in minutes and the robot resumes operation immediately. The depleted battery is placed in a charging rack to charge offline. Hot-swap capability requires additional battery inventory (typically 1.2–1.5 batteries per robot) and a battery management infrastructure (charging racks, battery health monitoring, rotation scheduling).

The FMS’s logic for managing the sequence in which robots pass through an intersection where two or more robot paths cross. Without intersection management, robots arriving at an intersection simultaneously would stop and wait indefinitely (deadlock) or collide. Intersection management assigns priority based on configurable rules: task priority (a robot carrying an urgent order proceeds first), first-arrival (the robot that reached the intersection first proceeds), direction priority (robots traveling on the primary aisle have priority over robots entering from secondary aisles), or minimum-delay (the system calculates which sequence minimizes total delay across all waiting robots). Poorly designed intersection management is a primary cause of fleet congestion and throughput loss.

A sensor that measures distances by emitting laser pulses and measuring the time for each pulse to reflect back from surrounding surfaces, generating a 2D or 3D point cloud that represents the robot’s environment. In warehouse AMRs, 2D LiDAR (a single scanning plane, typically mounted at ankle or waist height) is the primary sensor for navigation (SLAM localization) and safety (obstacle detection in the robot’s path). Key specifications: scan frequency (how many times per second the sensor completes a full 360° scan, typically 10–50 Hz), range (maximum detection distance, typically 10–30 meters for warehouse LiDAR), and angular resolution (the angular spacing between consecutive laser pulses, typically 0.25°–1.0°—smaller values provide finer detail). Units: Hz, m, degrees. See SLAM, obstacle detection.

A specific implementation of SLAM that uses LiDAR sensor data to simultaneously build a map of the environment and determine the robot’s position within that map. The LiDAR’s point cloud is matched against the stored facility map (or against the accumulating map during initial mapping) using algorithms such as iterative closest point (ICP) or scan matching. LiDAR SLAM is the dominant navigation technology for warehouse AMRs because LiDAR provides reliable, high-resolution distance measurements in the warehouse’s typical environment (structured geometry with walls, racking, columns, and shelving that provide consistent reference features). LiDAR SLAM’s primary vulnerability is featureless environments (long, uniform aisles without distinguishing features) where the scan-matching algorithm cannot distinguish the robot’s current position from adjacent positions.

A lithium-ion battery chemistry characterized by high cycle life (2,000–5,000+ charge cycles to 80% capacity), superior thermal stability (does not undergo thermal runaway under normal abuse conditions), and a flat voltage discharge curve (providing consistent power output throughout the discharge cycle). LiFePO₄ is the preferred battery chemistry for warehouse AMRs because the high cycle life minimizes battery replacement frequency (3–7+ years of daily cycling vs. 2–3 years for NMC), and the thermal stability eliminates the fire risk associated with higher-energy lithium chemistries in a warehouse environment with combustible materials. The trade-off: LiFePO₄ has lower energy density (90–160 Wh/kg) than NMC (150–220 Wh/kg), resulting in a heavier battery for equivalent runtime. See lithium nickel manganese cobalt.

A lithium-ion battery chemistry with higher energy density (150–220 Wh/kg) than LiFePO₄, providing longer runtime per charge or lighter weight for equivalent runtime. NMC batteries are common in consumer electronics and electric vehicles. In warehouse AMR applications, NMC’s higher energy density is offset by shorter cycle life (800–2,000 cycles to 80% capacity, compared to 2,000–5,000+ for LiFePO₄), lower thermal stability (greater susceptibility to thermal runaway under abuse conditions), and higher long-term cost due to more frequent replacement. The buyer should verify the battery chemistry in the AMR vendor’s proposal and calculate the battery replacement cost in the 10-year TCO: NMC batteries at $3,000–$5,000 per robot replaced every 2–3 years represent a significant operating cost that LiFePO₄ batteries avoid.

The process by which an AMR determines its precise position and orientation (x, y, θ) within the facility’s digital map. Localization is the foundation of autonomous navigation: the robot must know where it is before it can plan a path to where it needs to go. LiDAR SLAM is the primary localization method for warehouse AMRs, supplemented by odometry (wheel encoder data), IMU (inertial measurement unit) data, and in some systems, visual features from cameras. Localization accuracy in warehouse environments is typically ±20–50mm positional and ±1–2° angular, sufficient for aisle navigation and docking at pick stations and charging stations.

A condition in which the AMR’s localization system cannot determine the robot’s position with sufficient confidence to continue navigation safely. Localization failure triggers a protective stop: the robot halts, reports the failure to the FMS, and waits for operator intervention or an automated recovery attempt (rotating in place to acquire new sensor data that resolves the positional ambiguity). Common causes include featureless environments (long aisles without distinguishing features), reflective surfaces (the LiDAR pulse bounces unpredictably off reflective racking or shrink-wrapped pallets), dynamic environment changes (a section of racking was removed or reconfigured since the map was created), and sensor occlusion (the LiDAR’s field of view is blocked by a pallet or cart). Localization failure rate is a key performance metric: a fleet with a 0.1% localization failure rate per task generates 5 failures per 5,000 daily tasks, each requiring 30–60 seconds of downtime; a fleet with a 1.0% rate generates 50 failures per day—a significant productivity drain.

The digital representation of the facility that the AMR’s navigation system uses for localization and path planning. The standard map format is a 2D occupancy grid: a grid of cells (typically 5–10cm resolution) where each cell is classified as occupied (wall, racking, column, permanent obstacle), free (navigable floor space), or unknown. The map is typically generated by driving a robot through the facility with the LiDAR scanner active (the mapping process), and the resulting map is loaded to all robots in the fleet. The map must be updated when the facility layout changes (racking reconfiguration, new obstacles, zone changes); failure to update the map causes localization failures and navigation exceptions as the robot’s sensor data conflicts with the stored map.

An area of the facility where AMRs and human workers (pedestrians, forklift operators, pallet jack users) share the same floor space simultaneously. Mixed-traffic zones are the standard operating environment for collaborative warehouse AMRs and require the robot’s safety systems to detect, track, and avoid humans in real time per ISO 3691-4 and ANSI B56.5. The robot’s behavior in mixed-traffic zones includes reduced maximum speed (typically 1.0–1.5 m/s vs. 2.0+ m/s in robot-only zones), expanded safety detection fields, and mandatory stop or slow responses when humans are detected within defined proximity thresholds. See safety-rated speed, warning zone.

A navigation methodology in which the AMR uses SLAM with the facility’s existing structural features (walls, columns, racking, shelving, fixed equipment) as landmarks for localization, without requiring any infrastructure modifications to the facility (no floor markers, no magnetic tape, no reflectors, no embedded wires). Natural feature navigation is the defining capability that distinguishes AMRs from AGVs: the robot adapts to the environment as it exists rather than requiring the environment to be modified for the robot. The advantages are zero infrastructure cost and rapid deployment; the limitation is vulnerability to featureless or highly dynamic environments where the natural features are insufficient or change too frequently for reliable localization.

An event in which the AMR deviates from its planned path or stops unexpectedly due to an obstacle, a localization issue, a safety event, or a system fault. Navigation exceptions are classified by severity: a reroute (the robot detects an obstacle and calculates an alternate path without stopping—minimal time impact), a temporary stop (the robot stops, waits for the obstruction to clear, then resumes—10–30 seconds typical), and a persistent stop (the robot cannot resolve the situation autonomously and requires operator intervention—30–60+ seconds). Navigation exception frequency is a primary fleet performance metric, measured as exceptions per 1,000 tasks or exceptions per robot per shift.

The AMR’s capability to detect an obstacle in its planned path and calculate an alternative route around the obstacle in real time without stopping or requiring human intervention. Obstacle avoidance uses the robot’s LiDAR and/or depth camera data to identify the obstacle, the navigation software to evaluate alternative paths, and the path planner to select the optimal detour. Effective obstacle avoidance maintains the robot’s forward progress with minimal speed reduction; less effective systems require a full stop, reroute calculation, and restart, adding 5–15 seconds per avoidance event. See obstacle detection, path planning.

The AMR’s sensor subsystem that identifies objects in the robot’s path or proximity. Detection sensors are layered for comprehensive coverage: LiDAR (primary—long range, high resolution, 2D scanning plane at a specific height), depth cameras (supplementary—3D detection for objects above or below the LiDAR scanning plane, including low-profile obstacles on the floor and overhead obstructions), and ultrasonic sensors (close-range detection, typically used for docking precision and detecting transparent surfaces that LiDAR cannot reliably sense). The obstacle detection system feeds the safety controller (triggering speed reduction or stops per the safety field configuration) and the navigation system (providing obstacle data for path replanning).

A localization method that estimates the robot’s position by measuring the rotation of its drive wheels using shaft encoders and calculating the distance traveled and the heading change from the accumulated wheel rotations. Odometry provides continuous position updates at high frequency (typically 100–1,000 Hz) but accumulates error over time because of wheel slip, floor surface variation, and encoder resolution limits. In warehouse AMRs, odometry serves as the short-term position estimator between LiDAR SLAM updates: the robot uses odometry to track its position between LiDAR scans (every 20–100ms) and corrects the odometry-estimated position when the next SLAM update provides a more accurate fix. Odometry alone is insufficient for warehouse navigation—after 10–20 meters of travel, accumulated error can exceed 0.5–1.0 meter.

A charging strategy in which the AMR navigates to a charging station during brief idle periods between tasks (typically 30 seconds to 5 minutes) rather than waiting for the battery to reach a low state of charge before initiating a dedicated charging session. Opportunity charging maintains the battery at a higher average state of charge throughout the shift, extending the robot’s operational availability and reducing the depth of discharge per cycle (which extends battery cycle life). The FMS manages opportunity charging by evaluating the robot’s state of charge, the current task queue, and the proximity of available charging stations to determine whether a charging opportunity is worthwhile.

The navigation algorithm that calculates the optimal route from the robot’s current position to the task’s destination, considering the facility map, known static obstacles (racking, columns, walls), dynamic obstacles (other robots, reported temporary obstructions), traffic density, speed zones, and geofences. Path planning algorithms in warehouse AMRs include A* (graph-based shortest path), D* Lite (dynamic replanning as conditions change), and proprietary algorithms that incorporate fleet-level optimization (routing multiple robots simultaneously to minimize total fleet travel time and congestion). Path planning quality directly affects fleet throughput: a fleet with efficient path planning achieves 10–20% higher throughput than a fleet with the same hardware and naive shortest-path routing.

The maximum weight the AMR can carry on its transport platform during operation. Payload capacity is specified as static payload (maximum weight when the robot is stationary, including loading and unloading) and dynamic payload (maximum weight during travel, which may be lower than static payload due to the additional forces from acceleration, deceleration, and turning). Mid-market warehouse AMRs typically range from 30–50 kg for tote-carrying robots, 200–600 kg for shelf-carrying robots, and 500–1,500 kg for pallet-carrying AMRs. Units: kg (international) or lb (US). The buyer verifies that the AMR’s dynamic payload rating accommodates the heaviest load unit in the operation, including the weight of the container plus its contents.

A structured deployment methodology for AMR fleets in which the buyer deploys a small initial fleet (typically 5–15 robots) in a defined zone for a pilot period (8–12 weeks), evaluates performance against predefined success criteria, and then scales to the full fleet size in planned tranches based on the pilot’s validated results. The pilot-to-scale framework is the buyer’s primary risk management tool for AMR investments because the pilot validates throughput, operator adoption, integration quality, and navigation reliability in the buyer’s actual environment before committing the full capital.

A safety-rated response in which the AMR decelerates to a complete stop when a person or obstacle is detected within the robot’s protective safety field. The protective stop is a mandatory safety function per ISO 3691-4 and ANSI B56.5: the robot must stop before contacting any detected object within the protective field, regardless of the object’s size or type. The protective stop distance depends on the robot’s speed, mass (including payload), floor friction, and deceleration capability. The safety field’s geometry is configured to ensure that the robot stops completely before reaching the field’s boundary at every speed within the robot’s operating range. See safety-rated speed, warning zone.

A navigation method in which QR codes or 2D barcodes are affixed to the warehouse floor at regular intervals (typically in a grid pattern), and the AMR uses a downward-facing camera to scan the codes as it travels. Each code encodes the position’s coordinates on the facility map, providing the robot with absolute position fixes that correct any accumulated odometry error. QR code navigation provides high positional accuracy (±10–20mm at each code) and deterministic behavior (the robot knows its exact position at every code) but requires initial code placement (a significant installation effort for large facilities) and ongoing maintenance (codes damaged by forklift traffic, cleaning, or wear must be replaced).

The FMS’s algorithm for sequencing and prioritizing the tasks in each robot’s work queue and the global task pool. Queue management determines which task a robot receives next based on multiple factors: task priority (urgent orders before standard orders), robot proximity to the task’s origin (minimizing empty travel), robot battery status (assigning nearby tasks to robots with low battery rather than distant tasks that might not be completed before charging), task type compatibility (some robots may be configured for specific task types), and workload balancing (distributing tasks evenly across the fleet to prevent some robots from being overloaded while others are idle). Queue management quality directly affects fleet utilization and operator wait time.

A commercial model in which the buyer pays a monthly or annual per-robot subscription fee (typically $1,500–$4,000/robot/month) rather than purchasing the robots outright ($25,000–$50,000 per robot). The RaaS fee typically includes the robot hardware, fleet management software, maintenance, and replacements, converting the AMR investment from a capital expenditure (CapEx) to an operating expense (OpEx). RaaS reduces the buyer’s upfront investment and financial risk but typically results in a higher total cost over a 5–10 year horizon compared to capital purchase. RaaS also creates a contractual dependency: the buyer does not own the robots and cannot continue operating them if the RaaS provider ceases operations.

The number of robots operating per unit of automated floor area, used as a planning metric for congestion risk and infrastructure requirements (charging stations, Wi-Fi access points). Formula: robot density = number of robots / (automated zone area in square feet / 1,000). Typical mid-market densities: 1–3 robots per 1,000 sq ft in low-density deployments (transport-focused), 3–6 robots per 1,000 sq ft in medium-density deployments (collaborative picking), 6–12+ robots per 1,000 sq ft in high-density goods-to-person grid systems. Robot density above the vendor’s recommended threshold for the zone’s aisle width and intersection geometry triggers increasing congestion and diminishing throughput returns.

The percentage of a robot’s available operating time spent performing productive tasks (transporting loads, traveling to pick locations, docking for load transfer) as opposed to non-productive activities (waiting for tasks, charging, recovering from exceptions, sitting idle). Formula: utilization rate = (active task time / total available time) × 100. Target utilization for warehouse AMRs is typically 70–85%: below 70% suggests the fleet is oversized or the task allocation is inefficient; above 85% suggests the fleet has insufficient capacity for demand spikes and robots have no opportunity for charging during peak periods. Units: percentage.

The process by which the AMR’s Wi-Fi radio transitions its network connection from one wireless access point to another as the robot moves through the facility. Roaming is critical for AMR fleet operations because the robot must maintain continuous communication with the FMS for task updates, position reporting, and safety status. A roaming event that interrupts the connection for more than 100–200ms can cause the FMS to lose the robot’s position and the robot to miss a task update. The Wi-Fi infrastructure must be designed for fast, seamless roaming using protocols such as 802.11r (fast BSS transition) and 802.11k (neighbor reports) with a target roaming time under 50ms. See Wi-Fi latency.

The maximum speed at which the AMR is permitted to travel in each operating zone, defined by the risk assessment per ISO 3691-4 and constrained by the robot’s ability to detect obstacles and stop within the safety field’s boundaries at that speed. Safety-rated speed varies by zone: robot-only zones (no personnel access) may permit 2.0–3.0 m/s; mixed-traffic zones (shared with pedestrians) typically limit the robot to 1.0–1.5 m/s; high-traffic zones (near workstations, intersections, dock doors) may further restrict speed to 0.5–1.0 m/s. Higher speed requires larger safety fields (more floor space committed to detection) and longer stopping distances. Units: m/s.

An AMR that lifts and transports an entire mobile storage shelf (pod) from a storage area to an operator pick station, where the operator picks the required items from the shelf, and the robot returns the shelf to a storage location. The shelf-carrying model (pioneered by Kiva Systems, now Amazon Robotics) is a goods-to-person technology: the operator remains stationary and the product is delivered to the operator. Shelf-carrying robots operate on a grid-based navigation system and achieve high storage density because the shelves are stored without aisles (only the robot’s clearance is needed between shelves), with aisles created dynamically as robots navigate between shelf positions.

The computational process by which a robot simultaneously constructs a map of an unknown environment and determines its own position within that map, using sensor data (LiDAR, camera, or both). SLAM solves the circular dependency between mapping and localization: to build a map, the robot must know its position (to place sensor observations correctly on the map); to determine its position, the robot must have a map (to match sensor observations against known features). SLAM algorithms resolve this dependency using probabilistic methods that maintain and update estimates of both the map and the position simultaneously. In warehouse applications, SLAM is performed once during initial facility mapping and then the localization component runs continuously during operation to track the robot’s position within the stored map.

A defined area on the facility map where the FMS enforces a specific maximum robot speed, overriding the robot’s default speed limit for that zone. Speed zones are configured based on the risk assessment for each area: open transport corridors may allow full speed, aisle intersections require reduced speed for safe stopping distances, areas near operator workstations require reduced speed for pedestrian safety, and zones with floor surface transitions (expansion joints, ramps, dock plates) require reduced speed to prevent load shifting and robot instability.

The FMS’s core algorithm that assigns each incoming task to a specific robot based on multiple weighted factors: robot proximity to the task’s origin (minimizing empty travel time), robot battery level (ensuring the robot can complete the task without needing to charge mid-task), robot payload compatibility (matching the task’s load to the robot’s capacity), current fleet congestion (avoiding tasks that route the robot through congested areas), task priority (assigning high-priority tasks to the nearest available robot rather than waiting for the optimal robot), and workload balance (preventing overloading specific robots while others are idle). The task allocation algorithm’s quality is the primary differentiator between AMR vendors with comparable hardware: a superior algorithm achieves 15–25% higher fleet throughput from the same number of robots.

An AMR designed to transport one or more totes on its top surface from one location to another within the facility. The robot receives a tote at a source station (pick zone, receiving dock, AS/RS output), autonomously transports the tote to a destination station (packing, shipping, another pick zone), and transfers the tote at the destination via a conveyor interface, lift mechanism, or manual operator transfer. Tote-carrying robots are the most common AMR type for mid-market collaborative picking operations because they integrate with the facility’s existing tote-based workflows without requiring the storage infrastructure changes that shelf-carrying robots require.

The FMS’s real-time coordination of all active robots’ movements to prevent collisions, minimize congestion, and maximize fleet throughput. Traffic management encompasses intersection management (sequencing robots through shared intersections), congestion management (rerouting robots to avoid crowded areas), speed management (adjusting robot speeds based on local traffic density), and deadlock prevention (ensuring robot paths do not create mutual blocking conditions). Traffic management operates at the fleet level rather than the individual robot level: each robot reports its position and intended path to the FMS, and the FMS evaluates the collective impact of all robots’ movements before approving each path segment.

An open communication interface standard, developed by the German Association of the Automotive Industry (VDA) and the VDMA (German Mechanical Engineering Industry Association), that defines a standardized protocol for communication between a fleet management system (master control) and autonomous vehicles from different manufacturers. VDA 5050 specifies four communication layers: connection (MQTT-based transport), protocol (message types and formats for order, state, visualization, and factsheet topics), vehicle behavior (how the vehicle interprets and executes commands), and integration (how the FMS discovers and manages vehicles). VDA 5050 enables multi-vendor fleet interoperability: a buyer can operate robots from Vendor A and Vendor B under a single FMS, provided both vendors implement the VDA 5050 interface.

The software component on the AMR that implements the VDA 5050 communication protocol, translating VDA 5050 commands from the fleet management system into the robot’s native control language and translating the robot’s state information into VDA 5050 messages for the FMS. The vehicle agent is the abstraction layer that enables a VDA 5050-compliant FMS to control the robot without understanding the robot’s proprietary control system. The vehicle agent’s implementation quality determines the depth of VDA 5050 compliance: a fully compliant vehicle agent supports all four communication layers; a partially compliant agent may support basic order execution but not advanced features such as instant actions, visualization data, or factsheet exchange.

A boundary defined in the FMS’s facility map that the robot treats as an impassable obstacle, preventing the robot from crossing the boundary even though no physical barrier exists at that location. Virtual walls are used to restrict robot access to specific areas without installing physical barriers: keeping robots out of office areas, preventing robots from crossing loading dock thresholds during active truck operations, and creating temporary no-go zones during maintenance activities. Virtual walls can be activated and deactivated dynamically through the FMS’s map management interface.

The outer detection zone of the AMR’s safety system, defined by the safety scanner’s field configuration, in which the detection of a person or obstacle causes the robot to reduce speed (but not stop) to ensure the robot can stop safely if the person or obstacle enters the inner protective zone. The warning zone extends beyond the protective zone and provides an early-warning margin: the robot begins decelerating in the warning zone so that it approaches the protective zone boundary at a reduced speed, requiring a shorter stopping distance. Warning zone dimensions are configured based on the robot’s maximum speed and the floor’s friction characteristics.

A specific coordinate on the facility map that the robot uses as an intermediate destination or decision point along a path. Waypoints are used for routing robots through specific locations (requiring the robot to pass through an intersection in a specific direction), defining standby positions (where robots wait when no tasks are assigned), and creating ordered sequences (the robot must visit waypoint A before waypoint B before reaching the final destination). Waypoints are configured in the FMS’s map editor and are referenced by the path planning algorithm when generating routes.

The round-trip time for a data packet to travel from the AMR to the FMS server and back, measured in milliseconds. Wi-Fi latency affects fleet responsiveness: the FMS cannot update the robot’s task, adjust its path, or receive its position report until the message completes the round trip. Target latency for real-time AMR fleet management is below 50ms under normal conditions and below 100ms during peak Wi-Fi utilization. Latency above 200ms causes perceptible delays in task assignment and position reporting, leading to suboptimal path planning and increased congestion risk. Wi-Fi latency is affected by the access point density, the number of connected devices, interference from other wireless systems, and the quality of the roaming configuration. Units: milliseconds (ms).

A defined area of the facility where AMRs are authorized to operate, delineated by geofences or virtual walls in the FMS’s facility map. The work zone encompasses the aisles, intersections, staging areas, and workstation approaches that the robots use for task execution. Areas outside the work zone are off-limits to robots. The work zone’s boundaries are defined based on the facility’s layout, the safety assessment, and the operational design: some zones are robot-only (no personnel access during robot operation), some are mixed-traffic (shared with pedestrians), and some are exclusion zones (no robot access at any time).

The geographic line on the facility map where one operational zone ends and another begins. At a zone boundary, the robot’s operating parameters may change: speed limit (transitioning from a high-speed transport corridor to a low-speed pick zone), safety field configuration (expanding the protective field when entering a mixed-traffic zone), and fleet management assignment (the robot may transfer from one zone controller to another in distributed FMS architectures). The zone boundary configuration ensures that the robot applies the correct operating parameters for each zone before entering the zone, not after.

The maximum number of robots that the FMS allows to operate simultaneously within a defined zone, enforced to prevent congestion that would reduce per-robot throughput below acceptable levels. When the zone reaches its capacity limit, additional robots are queued at the zone boundary or rerouted to alternative zones. Zone capacity limits are determined empirically during the pilot phase or through simulation: the operations team increases the robot count in each zone until the per-robot throughput begins to decline, and sets the capacity limit at the point of maximum aggregate throughput.

The process by which an AMR transitions from one zone’s management to another zone’s management as the robot crosses a zone boundary. In distributed FMS architectures (where each zone has a local controller that manages the robots within that zone), the zone handoff involves transferring the robot’s task assignment, position tracking, and traffic management responsibility from the departing zone’s controller to the arriving zone’s controller. The handoff must be seamless: the robot should not stop or slow during the transition. In centralized FMS architectures, the zone handoff is a logical transition (the central FMS updates the robot’s zone assignment in its database) rather than a physical controller transfer.