SCARA arms deliver 4-axis pick-and-place at payloads of 5-10 kg and 650 mm reach in fixed cells, while AGVs and AMRs handle floor-level material transport between those cells.
Specifying the wrong class wastes capital: a SCARA robot bolted to a conveyor will not move totes across a 200 m warehouse, and an AGV robot rolling under a feeder cannot solder a PCB. The two are complementary, not interchangeable.
Kinematic Class, Reach, and Payload
A SCARA arm uses selective compliance in the horizontal plane and stacked vertical axes for Z and rotation, giving it 4 controlled axes optimised for vertical pick-and-place strokes. The EPSON GX10-B, for example, ships as a 4-axis packaging unit with a 650 mm reach, ceiling- or tabletop-mount options, and a 5-10 kg payload window [S2]. The arm's horizontal rigidity makes it the default choice for high-cycle assembly of small parts where repeatability in the XY plane matters more than long reach.
An AGV robot is a wheeled vehicle, not a manipulator: it carries totes, pallets, or racks on a flat deck, navigates by magnetic tape, QR, or SLAM, and unloads at fixed stations. The kinematic job is to translate payloads across a facility footprint — typically hundreds of metres — not to position an end-effector within sub-millimetre tolerance. Reach and payload figures therefore describe deck capacity and towing force, not arm articulation.
Selection Criteria: Stationary Precision vs Mobile Transport
The decision pivot is whether the work is station-bound or floor-bound. If the value-add step is assembling, dispensing, inspecting, or packaging at a fixed cell, a SCARA robot integrated with a feeder, conveyor, or vision system is the correct spec. The Kollmorgen SCARA line has been deployed for pick-and-place, parts assembly, and life-sciences applications since the early 1980s, with servo motor sizing matched to torque and speed per axis [S4]. The ABB SCARA family, similarly, targets fast-cycle material handling, machine tending, and small-parts assembly using a configurable robot-selector and AR-viewer toolchain [S1].
If the work is moving bin-totes from a warehouse rack to a production line, an AGV robot — or its more autonomous cousin, an AMR — is the correct spec. AGVs follow predetermined paths; AMRs localise and re-plan dynamically. For heavier mobile work like pallet transport, the comparison shifts to reach trucks and forklifts, where aisle width, lift height, and load capacity become the binding constraints as outlined in pallet jack vs forklift selection for 2026 warehouses.
Decision Matrix: SCARA vs AGV Across Four Criteria
Four criteria cleanly separate the two classes. Workstation footprint: a SCARA cell occupies roughly a 1-2 m² tabletop footprint with a 650 mm reach envelope [S2]; an AGV path occupies floor lanes measured in tens of metres. Positioning accuracy: SCARAs hit ±0.01-±0.02 mm repeatability in the XY plane typical of the class; AGVs dock to ±10-±20 mm with fiducial aids. Cycle metric: SCARAs are rated in picks-per-minute (often 60-120 ppm in light assembly); AGVs are rated in missions-per-hour and travel speed (typically 1-1.5 m/s). Payload form factor: a SCARA moves a part in a gripper weighing grams to low kilograms; an AGV moves a tote or pallet weighing tens to hundreds of kilograms on its deck.
The matrix rules out cross-deployment. A SCARA cannot traverse a plant aisle; an AGV cannot perform a soldering stroke. Where they meet is in a hybrid cell: a SCARA loads a tote at station A, an AGV shuttles that tote to station B, a second SCARA unloads it — a pattern common in 3C, pharmaceutical, and daily-chemical lines served by integrators like the Phoenix Bat (parallel), Python (SCARA), and Camel (AGV) product series [S3].
Standards, EMC, and Integration Discipline
SCARAs sold into Chinese markets must meet the T/CES 254-2023 group standard for electromagnetic compatibility, which sets emission and immunity test methods specific to SCARA robot architectures and was published 2023-12-27 with implementation from 2023-12-28 [S6]. For the AGV side, fleet managers typically reference ISO 3691-4 for driverless industrial truck safety and IEC 61508 for safety-rated navigation functions — the binding document depends on the deployment country and the presence of onboard personnel. Open architectures are converging: a ROS 2 + Gazebo simulation stack such as the SCARA_ROBOT_ROS2 reference repository lets integrators validate kinematics and motion control offline before commissioning a cell [S5].
On the controls side, multi-model motion algorithms and one-controller-many-robots architectures are now common in Chinese SCARA/AGV hybrid lines; the Phoenix platform advertises a single main controller dragging multiple robot configurations, a VisionPower vision platform, and PLCopen-compliant motion instruction sets for hybrid cells [S3]. For plant engineers weighing material handling equipment more broadly, the spec-first comparison in reach truck vs forklift aisle and load selection covers the lift-truck neighbour of this class.
Use Cases and Failure Modes
SCARAs dominate 3C electronics assembly, pharmaceutical packaging, medical-device sub-assembly, and small automotive sub-components where 5-10 kg payloads and sub-millimetre repeatability drive yield. EPSON's GX10 series is explicitly positioned for confined-space work in automotive components, mobile phones, computer parts, medical equipment, medication packaging, and lab automation [S2]. Failure modes are predictable: overreach past the rated 650 mm envelope, Z-axis stack deflection under excess payload, and harmonics from mismatched servo sizing — all addressed by axis-specific motor selection per the Kollmorgen co-engineering model [S4].
AGVs dominate warehousing, e-commerce fulfilment, and inter-line tote shuttling where payload outweighs positioning precision. Failure modes are different: floor flatness, fiducial degradation, traffic-management deadlocks, and battery duty cycle. Phoenix's 3C, pharmaceutical, daily-chemical, food, semiconductor, warehousing, and auto-parts solution verticals all assume an AGV/AMR shuttling totes between SCARA cells rather than a single robot doing both jobs [S3]. For facilities weighing the broader detection side of safety hardware, the multi-gas detector vs single gas detector spec map and the heat detector vs gas detector hazard comparison follow the same selection-first logic.
Who Each Robot Is For — and Who It Is Not For
A SCARA is for the process engineer specifying a fixed workstation: a 3C assembler, a pharma packager, a semiconductor handler who needs a 5-10 kg payload, 650 mm reach, and four-axis vertical pick-and-place under tight cycle times [S2]. It is NOT for the logistics engineer moving 500 kg pallets across a 10 000 m² warehouse — that is forklift or reach-truck territory as mapped in the reach truck vs forklift comparison. An AGV is for the logistics engineer running totes between cells at 1-1.5 m/s on a fixed lane network; it is NOT for a precision assembly task that demands ±0.02 mm XY repeatability.
The cleanest spec rule for 2026: if the value-add step is motion of the part relative to a tool, specify a SCARA; if the value-add step is motion of the part relative to the factory, specify an AGV or AMR; if a line needs both, spec them as a coupled cell with a shared controller and a single ROS 2 or PLCopen motion backbone rather than two isolated islands.
Two trackable signals for the next planning cycle: (1) ROS 2-native SCARA controllers maturing past prototype status into production cells, evidenced by active repositories such as the SCARA_ROBOT_ROS2 simulation stack [S5]; (2) Chinese integrators like Phoenix continuing to fold SCARA, parallel, six-axis, and AGV lines under one controller architecture [S3], which compresses integration cost for hybrid cells. Engineers specifying in late 2026 should ask vendors whether the SCARA and the AGV in their bid share a motion-control kernel, because that single question decides whether the cell behaves as one system or two.