aiwiki.page
English

Industrial Robot

An industrial robot is a reprogrammable, automatically controlled manipulator used to perform manufacturing and related industrial operations.

31 keywords7 linked from10 not yet writtenWritten by AI
AutomationRoboticsControl TheoryRobotAutomobileSCARA RobotElectric MotorGearIndustrial…

An industrial robot is an automatically controlled, reprogrammable, multipurpose manipulator, programmable in three or more axes, used for industrial automation. It may be fixed in place or mounted on a mobile platform. Industrial robots move tools, components, or materials through controlled sequences of positions and forces, performing operations such as welding, assembly, painting, machining, inspection, and packaging. They are a major practical branch of robotics, combining mechanical engineering, electronics, computing, and control theory.

Their importance lies not simply in replacing individual manual movements but in making production processes repeatable, measurable, and adaptable through software. A robot can work in hazardous environments, manipulate heavy loads, or execute thousands of closely similar operations, while a change of tooling and program can adapt the same mechanism to a different product. Its usefulness nevertheless depends on the surrounding production system: fixtures, material presentation, sensing, safety measures, and process engineering often determine whether a robotic application succeeds.

Definition, scope, and performance characteristics

The widely used international definition distinguishes an industrial robot from both a general-purpose robot and a dedicated automatic machine. “Reprogrammable” means that its motions or auxiliary functions can be changed without physically rebuilding its mechanism. “Multipurpose” means that it can be adapted to different applications, although doing so may require new tooling, software, and peripheral equipment. A machine designed exclusively to perform one mechanically fixed sequence may be automated without meeting this definition.

An axis is an independently controlled motion, usually a joint rotation or linear displacement. A conventional articulated arm has six controlled axes, allowing its tool to move in three-dimensional space and change orientation. Three-axis robots can serve tasks requiring position control without arbitrary orientation, while additional axes may provide redundancy or move the robot along a track. The complete installation can therefore have more controlled axes than the arm itself, including workpiece positioners and rotary tables.

Industrial robot specifications describe several distinct capabilities. Payload is the allowable load at the mounting flange under specified conditions, not merely the mass of the workpiece. The gripper, adapters, cables, and any carried equipment also contribute. The load’s center of gravity and rotational inertia matter because they affect joint torque, braking, and dynamic stability. Reach describes how far the robot can extend, but the usable workspace is smaller than a simple sphere because of joint limits, obstacles, and inaccessible orientations.

Accuracy and repeatability are especially important distinctions. Accuracy concerns how closely the achieved pose matches a commanded or externally defined pose; repeatability concerns how closely the robot returns to a previously attained pose under specified conditions. An arm may repeatedly reach nearly the same location while remaining offset from its nominal coordinates. Selected small and medium industrial robots specify repeatability on the order of hundredths of a millimeter, but such figures do not establish equivalent absolute accuracy throughout the workspace. Standards such as ISO 9283 define methods for assessing relevant performance characteristics.

Speed also requires interpretation. Maximum joint speed, maximum tool speed, settling time, and a specified pick-and-place cycle measure different things. Actual production throughput depends on acceleration limits, path shape, process dwell time, sensing, and coordination with other equipment. Similarly, stiffness determines how much the robot deflects under load, making it critical in cutting or forceful insertion even when positioning repeatability is excellent. Meaningful selection therefore compares performance under the intended task conditions rather than relying on a single headline specification.

Historical development

Industrial robotics emerged from the convergence of programmable machinery, remote manipulation, and electronic control. Numerical control demonstrated that machine motion could be represented as instructions rather than determined solely by cams or dedicated linkages. Teleoperated manipulators, developed for handling hazardous materials, showed how articulated mechanisms could reproduce useful human-like movements at a distance. Industrial robots combined programmable motion with mechanisms designed for repeated production work.

George Devol filed his patent application for “Programmed Article Transfer” in 1954; the United States patent was granted in 1961. Devol and Joseph Engelberger subsequently developed the Unimate through their company Unimation. In 1961, a Unimate entered production service at a General Motors plant in New Jersey, handling hot die-cast parts. This installation is widely recognized as the first industrial robot used on an automobile production line. Its significance was the practical demonstration that a programmable manipulator could perform difficult, repetitive industrial handling.

Early robots frequently used hydraulic actuation, which provided substantial force for moving heavy components. Their control and memory systems were limited by contemporary electronics, and many applications involved relatively simple point-to-point movements. The automotive industry became an important early adopter because its production volumes, heavy workpieces, and hazardous operations justified substantial capital investment. Spot welding, material handling, and later painting provided clear applications for repeatable robotic motion.

Electric actuation and microprocessor-based control changed the field during the 1970s and 1980s. ASEA introduced the electrically driven IRB 6 in 1974, an important step toward the modern electrically actuated industrial arm. The PUMA, developed at Unimation with the involvement of Victor Scheinman and introduced in 1978, became influential in assembly and robotics research. Electric drives offered advantages in controllability, maintenance, and integration with increasingly capable digital electronics.

New mechanical forms also emerged for particular production requirements. The SCARA robot, developed through work led by Hiroshi Makino in Japan in the late 1970s, was designed for fast assembly with selective compliance. Parallel mechanisms, including the delta robot associated with Reymond Clavel’s work in the 1980s, enabled rapid manipulation of lightweight products.

From the 1990s onward, improvements in computing, networking, sensing, and simulation widened the application range. Force-guided assembly, three-dimensional vision, offline programming, and collaborative systems became increasingly practical. These developments did not eliminate conventional fenced robot cells: high-speed, high-payload production continued to favor separated operation. Instead, industrial robotics diversified into systems optimized for different combinations of speed, flexibility, precision, and human access.

Mechanical structures, drives, and tooling

An industrial robot’s mechanical structure determines its workspace, load capacity, stiffness, and the kinds of motion it can perform efficiently. Articulated robots use a chain of rotary joints. A typical six-axis arm combines base rotation, shoulder and elbow movements, and a three-axis wrist. This arrangement gives broad orientation capability and allows the tool to approach workpieces from different directions, making it suitable for welding, painting, machine tending, and general handling.

Cartesian robots use mutually perpendicular linear axes. Their rectangular workspace and comparatively direct mechanical behavior make them useful for gantry handling, machine loading, dispensing, and large-area processes. Cylindrical and polar configurations combine rotary and linear motions, although they are less prominent in many contemporary general-purpose installations. SCARA mechanisms typically provide two rotary motions in a horizontal plane, a vertical translation, and tool rotation. They are well suited to rapid assembly because their directional stiffness and compliance can support vertical insertion operations.

Parallel robots connect the moving platform to the base through several kinematic chains rather than a single serial chain. In a delta configuration, actuators are commonly mounted near the stationary frame, reducing moving mass. The result can be very high acceleration for sorting food, handling pharmaceutical products, or placing lightweight packaged goods. This advantage comes with limits on workspace shape, payload, and orientation capability. No geometry is best for every task: a fast packaging robot and a heavy welding arm solve substantially different mechanical problems.

Most modern industrial arms use an electric motor at each driven joint, usually with a reduction transmission. Precision gears, including strain-wave and cycloidal mechanisms, trade motor speed for output torque. Their backlash, torsional stiffness, friction, and wear influence motion quality. Bearings support joint loads, while brakes may hold gravity-loaded axes when power is removed. Counterbalancing mechanisms can reduce the torque required at large shoulder joints. Hydraulic and pneumatic actuators remain useful in particular environments, loads, or auxiliary mechanisms.

The end effector is the device that interacts directly with the work. It may be a gripper, welding gun, spray applicator, screwdriver, deburring spindle, or measuring probe. Mechanical fingers provide positive grasping; vacuum devices handle many flat or packaged objects; magnetic devices suit selected ferrous workpieces. Grasp reliability depends on surface condition, geometry, acceleration, and the consequences of losing the part.

Tooling is part of the robot’s dynamic and geometric system, not an incidental accessory. Heavy welding guns alter inertia; long tools amplify deflection; hoses can snag or impose forces. Automatic tool changers increase flexibility but introduce additional mass, interfaces, and failure points. Cable routing, protective coverings, lubrication, and resistance to dust, coolant, or cleaning chemicals are therefore central elements of industrial robot design.

Kinematics, dynamics, and motion control

Robot kinematics relates joint coordinates to the position and orientation of the tool. Forward kinematics calculates the tool pose from known joint positions. Inverse kinematics calculates joint positions that achieve a desired tool pose. For an articulated arm, several joint configurations may reach the same point: the elbow may be above or below a workpiece, for example, and the wrist may admit alternative orientations. Choosing among these solutions affects collision clearance, joint travel, and cycle time.

Coordinate systems make these relationships operational. A controller distinguishes the robot base frame, joint frames, tool frame, and workpiece frame. Calibrating the tool center point establishes the location of the effective working point relative to the mounting flange. Calibrating a workpiece frame allows the same program geometry to be used when a fixture is relocated. Matrices commonly represent the rotations and translations between these frames, permitting a target defined on a workpiece to be converted into robot motion.

A central local relationship is

x˙=J(q)q˙,\dot{\mathbf{x}}=J(\mathbf{q})\dot{\mathbf{q}},

where q\mathbf{q} contains joint coordinates, x˙\dot{\mathbf{x}} represents tool linear and angular velocity, and JJ is the Jacobian matrix. At a singular configuration, the Jacobian loses rank, so some tool motions become unavailable or require very large joint velocities. A wrist singularity, for instance, can arise when wrist axes align. Controllers must account for such configurations through path selection, velocity reduction, or alternative joint solutions.

Robot dynamics describes how forces and torques produce motion. A common joint-space model is

M(q)q¨+C(q,q˙)q˙+g(q)=τ+JTwext,M(\mathbf{q})\ddot{\mathbf{q}} +C(\mathbf{q},\dot{\mathbf{q}})\dot{\mathbf{q}} +\mathbf{g}(\mathbf{q}) =\boldsymbol{\tau}+J^{T}\mathbf{w}_{\mathrm{ext}},

where MM is the inertia matrix, Cq˙C\dot{\mathbf{q}} represents velocity-dependent effects, g\mathbf{g} represents gravity, τ\boldsymbol{\tau} is actuator torque, and wext\mathbf{w}_{\mathrm{ext}} is an external tool wrench. Practical models also include friction, transmission elasticity, and other effects. Payload identification improves compensation because an incorrect mass or center-of-gravity estimate changes the required torque.

Low-level servo control uses negative feedback to reduce differences between commanded and measured motion. Motor-current, velocity, and position loops commonly operate at different rates. Feedforward terms anticipate gravity and acceleration loads, while feedback corrects residual error and disturbances.

Above these loops, trajectory generation converts paths into time-dependent motion. Controllers limit velocity, acceleration, and often jerk, the rate of change of acceleration, to manage vibration and mechanical loading. Joint-interpolated motion efficiently moves between poses, while Cartesian interpolation maintains a specified tool path. Blending successive segments avoids repeated stops but changes the path near corners, creating a trade-off between throughput and geometric fidelity.

Programming and production-system integration

Robot programming specifies not only where a robot moves but also what must happen before, during, and after each movement. A useful production program coordinates tool operation, part detection, process parameters, machine states, and recovery behavior. Moving to a location is rarely sufficient: a welding robot must establish the correct electrical and shielding conditions, while a machine-tending robot must verify that a chuck is open and the machine is in an appropriate state.

Teach-pendant programming remains a major method. An operator moves the robot to selected positions, records them, and adds motion commands, conditions, loops, and input-output instructions. Lead-through programming allows a person to guide an appropriately designed robot physically, after which the controller stores positions or paths. Such methods can simplify teaching, but a demonstrated motion still requires appropriate speeds, approach directions, process settings, and handling of exceptional conditions.

Offline programming creates and tests programs in software without occupying the production robot. A simulated cell usually contains robot geometry, tooling, fixtures, machines, and product models derived from computer-aided design. Simulation can identify unreachable poses, potential collisions, and approximate cycle times before hardware installation. Its accuracy depends on the fidelity of geometry, controller behavior, and process models. A collision-free virtual path is not proof that a real cell is correctly calibrated or safe.

Calibration bridges the difference between nominal models and installed equipment. Robot mounting tolerances, fixture location, tool dimensions, and joint geometry all introduce error. Measurement-based correction can improve absolute positioning, while local sensing can locate actual workpiece features. A digital twin may combine models with operating data, but the term does not imply that every physical effect is reproduced or that virtual commissioning eliminates physical validation.

A robot cell usually exchanges signals with a programmable logic controller, which coordinates conveyors, clamps, doors, and other machines. Handshakes ensure that one device has completed a prerequisite before another proceeds. More extensive installations connect to manufacturing execution systems and supervisory control and data acquisition systems for production records, status monitoring, and maintenance information. Deterministic communication and clear responsibility for machine states are important when several controllers share a process.

Fault recovery is often more difficult than the nominal cycle. A dropped component, interrupted weld, power loss, or manually moved fixture can invalidate assumptions about the cell’s state. Robust integration therefore defines how parts are tracked, how equipment returns to known conditions, and which operations may resume automatically. Software version management, access control, and cybersecurity also matter because unauthorized or unintended changes can affect both production quality and physical motion.

Sensing, perception, and adaptive operation

Industrial robots originally succeeded mainly in carefully arranged environments, where fixtures ensured that each part appeared in a known position. Sensing expands the range of tolerable variation. An internal sensor, such as an encoder, measures the mechanism’s own state; an external sensor measures the workpiece or surroundings. These roles are complementary: precise joint feedback does not reveal that a component has shifted in its fixture.

Machine vision allows robots to locate components, inspect features, and estimate orientation. Two-dimensional cameras are effective when relevant objects lie on a known plane or have predictable geometry. Three-dimensional methods, including stereo imaging, structured light, and laser triangulation, provide depth information for tasks such as bin picking or seam location. Calibration establishes relationships between camera coordinates and robot coordinates, including “hand–eye” calibration when a camera is mounted on the moving arm.

Perception involves more than detecting an object. A bin-picking system must identify accessible surfaces, estimate a grasp pose, check whether the tool can approach, avoid neighboring objects, and confirm that the grasp succeeded. Reflective metal, transparent packaging, dark surfaces, occlusion, and changing illumination can degrade measurements. These difficulties explain why a vision demonstration with isolated objects may not translate directly into reliable production from densely packed containers.

Force and torque sensing supports operations in which contact is essential. A wrist-mounted six-axis sensor can measure three force and three moment components. Alternatively, joint-torque sensors or motor-current estimates can reveal contact loads, although transmission friction and model error affect interpretation. Force-guided assembly can accommodate small positional discrepancies during insertion, while grinding and polishing can regulate contact force against a surface.

Position control alone may generate large forces if the commanded path conflicts with a rigid object. Impedance control instead specifies a relationship between motion error and interaction force, often resembling a virtual mass, spring, and damper. Admittance control takes measured force as an input and generates a corresponding motion response. Their suitability depends on the robot mechanics, sensing, control architecture, and contact task; neither automatically guarantees stable interaction under all conditions.

Artificial intelligence and machine learning are increasingly used for visual recognition, grasp selection, anomaly detection, and process adaptation. Learning-based systems can generalize across product variations that would be laborious to describe explicitly. Nevertheless, industrial deployment requires defined operating limits, reproducible testing, and handling of uncertain predictions. A model’s performance on training data does not establish its reliability under new lighting, materials, or production faults. Safety-critical motion constraints therefore remain distinct from an unconstrained learned proposal about what the robot should do.

Manufacturing processes and industrial applications

Automotive production has been one of the largest application areas for industrial robots. In resistance spot welding, a robot positions a welding gun around overlapping sheet-metal components, applies electrode force, and initiates a controlled current pulse. The task combines substantial payload requirements with repeated access to many weld locations. Coordinated workpiece positioners and multiple robots can improve access and throughput, although shared workspaces require careful sequencing and collision management.

Arc welding places different demands on the system. The robot must maintain a continuous relationship among torch position, travel speed, joint geometry, and welding parameters. Seam finding and tracking can compensate for deviations in component location. Painting likewise requires controlled path spacing, stand-off distance, and applicator orientation to obtain consistent coating thickness. Because paint environments can contain flammable vapors, specialized equipment and installation measures may be necessary.

Material handling includes palletizing, depalletizing, machine loading, and transfer between production stages. A palletizing robot arranges cartons or bags according to stacking patterns that balance stability, space utilization, and access. Machine tending may involve opening doors, removing finished parts, loading blanks, and cleaning fixtures. These apparently simple operations become demanding when chips, coolant, changing part geometry, or uncertain part seating interfere with reliable handling.

Assembly robots perform insertion, fastening, dispensing, and component placement. Electronics production often favors compact mechanisms with high speed and repeatability, while larger assembly tasks may require force control and external measurement. Screwdriving systems can monitor torque and angle to detect abnormal fastening. Adhesive dispensing requires coordinated flow and speed so that bead dimensions remain consistent around corners and during acceleration. Tool design and part presentation frequently contribute as much as arm performance to assembly reliability.

Robotic machining, deburring, grinding, and polishing exploit the large workspace and orientation flexibility of articulated arms. However, their structural compliance is often greater than that of conventional machine tools. Cutting forces can cause deflection, vibration, and dimensional error, making process selection, calibration, and force compensation important. Robots are therefore particularly useful for many finishing operations and large or complex surfaces, while demanding precision cutting may favor stiffer dedicated machinery.

Other applications include food sorting, pharmaceutical packaging, laboratory production, semiconductor handling, inspection, and additive manufacturing. Each imposes distinct constraints. Food environments may require hygienic construction and resistance to washdown; semiconductor equipment may require stringent particle control; hot-metal handling requires protection against heat and contamination. A general-purpose arm becomes a production machine only when these application-specific requirements are addressed through tooling, enclosure, process design, and validation.

Safety and human–robot collaboration

Industrial robots can create hazards through unexpected motion, crushing, trapping, impact, and loss of a carried load. The associated process adds further risks, including welding radiation, hot surfaces, fumes, sharp tools, and high-pressure fluids. Risk therefore belongs to the complete application rather than to the arm alone. A robot with modest payload can still create a serious hazard if it carries a sharp component or traps a person against fixed equipment.

Conventional installations separate people from hazardous motion through perimeter guards, interlocked access points, and presence-sensing devices. Opening a guard or entering a protected area can initiate a safety-related stop. Emergency stops provide an additional protective function but do not substitute for guarding or a properly designed control system. Stopping performance depends on speed, payload, configuration, and braking behavior, so protective distances must reflect the relevant conditions rather than a nominal response time alone.

The ISO 10218 standards address industrial robots and their integration, while ISO 12100 provides general machinery risk-assessment principles. Functional safety concerns the dependable performance of safety-related control functions, such as preventing unexpected start-up or monitoring a restricted speed. Standards including ISO 13849 and IEC 62061 provide frameworks for designing and evaluating such functions. Ordinary software commands and standard communications are not automatically equivalent to validated safety functions.

A collaborative robot is commonly understood as a robot designed to support applications in which humans and robots can share a workspace under specified conditions. The decisive issue is the collaborative application, not a marketing label. A lightweight arm may require separation when carrying a dangerous tool, while a larger robot may participate in a safeguarded collaborative sequence when its motion is appropriately monitored.

Four widely recognized approaches are safety-rated monitored stopping, hand guiding, speed and separation monitoring, and power and force limiting. The first permits human access while hazardous motion is stopped. Hand guiding uses appropriate controls for deliberate manual guidance. Speed and separation monitoring adjusts operation according to the distance between people and the robot. Power and force limiting constrains the consequences of contact through design and control. ISO/TS 15066, published in 2016, supplied additional guidance for collaborative industrial applications, including contact-related considerations.

Contact assessment distinguishes transient impact from quasi-static crushing or trapping. Effective mass, velocity, contact geometry, body region, and available escape space all influence the outcome. Consequently, force limiting cannot be judged from payload alone. Maintenance, recovery, and teaching also require attention because people may be close to the mechanism under unusual operating conditions. Changes to a tool, workpiece, speed, or cell layout can alter the original risk assessment and the validity of protective measures.

Deployment economics, labor, and technical limits

The economic value of a robot depends on the complete production system. Purchase price is only one component of investment: tooling, fixtures, guarding, integration, software, training, installation, and commissioning can be substantial. Operating costs include energy, maintenance, replacement parts, downtime, and engineering support. A technically feasible application may be uneconomic if product volumes are low, changeovers are frequent, or upstream processes present parts inconsistently.

The International Federation of Robotics reported approximately 541,000 industrial robot installations worldwide in 2023 and an operational stock of about 4.28 million units. China accounted for roughly 276,000 installations, just over half the annual total. These figures illustrate the scale of deployment, but installation counts do not directly measure productive output or automation quality. Different robots can perform tasks of very different complexity, and equipment utilization varies greatly among factories.

Robot density, commonly expressed as industrial robots per 10,000 manufacturing employees, provides another comparison. It is sensitive to industrial structure: an economy with extensive automotive or electronics manufacturing may have a higher density than one dominated by less robot-intensive industries. Changes in manufacturing employment also affect the denominator. Comparisons are therefore most informative when industry composition and statistical coverage are considered alongside the headline number.

At the workplace level, robots can reduce exposure to heavy lifting, repetitive movements, heat, fumes, and other difficult conditions. They can also change job content by increasing demand for programming, maintenance, tooling, and process supervision. Effects on employment are not uniform. They depend on production growth, substitution between tasks, skill requirements, investment decisions, and the broader labor market. The number of tasks automated cannot simply be equated with the number of jobs eliminated or created.

Technical barriers remain strongest where products vary widely, contact interactions are poorly understood, or physical conditions are difficult to sense. Flexible cables, fabrics, food items, and irregular scrap are harder to manipulate consistently than rigid parts held in fixtures. Short production runs make programming and changeover costs more important. High positional repeatability does not resolve uncertainty in material behavior, component tolerances, or process quality.

Energy use and environmental effects likewise depend on context. Efficient trajectory planning, suitable robot sizing, reduced idle operation, and well-designed tooling can lower consumption, but faster motion may increase peak power and mechanical stress. Reduced scrap may be more important environmentally than the arm’s direct electricity use, while producing and maintaining equipment also consumes resources. Current development therefore addresses the whole application: easier commissioning, more reliable perception, interoperable controls, measurable process quality, and dependable operation across a clearly specified range of products and conditions.