Industrial robot types and applications

Industrial robots are the most widely deployed robots in the world. The International Federation of Robotics (IFR) has estimated the operational stock of factory robots at more than four million units, with more than half a million new installations a year in recent reports, and the automotive and electronics industries still account for a large share of them. ROB-440 is about how these machines are chosen, laid out, programmed, connected, safeguarded, maintained and paid for. Earlier courses gave you the theory underneath: kinematics in ROB-210, dynamics and control in ROB-300, real-time software in ROB-320, vision in ROB-330, planning in ROB-350, grasping in ROB-400 and human-robot collaboration in ROB-430. Here we take the integrator's and the plant engineer's point of view, where the question is not whether a robot can do a task in principle but whether a particular robot cell will run three shifts a day for ten years at an acceptable cost and with nobody hurt.

This first lesson surveys the main mechanical configurations of industrial robots, the specifications that appear on every datasheet, and the applications that make up most of the market. By the end you should be able to read a robot datasheet critically, match a configuration to a task, and estimate whether a candidate robot can reach, lift and keep up with a job before anyone spends money on a simulation.

What makes a robot industrial

The international vocabulary standard ISO 8373 defines an industrial robot as an automatically controlled, reprogrammable, multipurpose manipulator, programmable in three or more axes, which can be either fixed in place or mobile, for use in industrial automation applications. Each part of that definition matters. Automatically controlled means the robot runs a stored program without an operator guiding each motion. Reprogrammable means the motion or auxiliary functions can be changed without physical alteration. Multipurpose means it can be adapted to a different application by reprogramming or by changing the end effector. A pick-and-place mechanism built from fixed cams, or a dedicated press-loading shuttle, can be fast and reliable, but it is fixed automation rather than a robot. The boundary matters in practice because fixed automation is often cheaper and faster for one product made in enormous volume, while robots win when products change, volumes are moderate, or the task needs dexterity.

Industrial robots are part of a larger hierarchy of automation. Fixed or hard automation, such as transfer lines and dedicated machines, has the highest throughput and the least flexibility. Programmable automation, such as CNC machine tools and robots, can be changed between batches. Flexible automation can switch between products with little or no downtime, often using robots with quick tool changers and vision. Choosing the right level is the first engineering decision in any project, and it is an economic decision as much as a technical one. Lesson 11 returns to the money.

Mechanical configurations

Robots are usually classified by the arrangement of their first three joints, which position the wrist, and by how many axes they have in total. The remaining joints orient the tool. ROB-210 derived the kinematics of these arrangements; here we focus on what each is good for on a factory floor.

Articulated arms dominate the market because a six-axis arm can place a tool at any position and orientation within its envelope. A seventh axis, as on some collaborative and dual-arm robots, adds redundancy that helps avoid obstacles and singularities. Articulated arms are also commonly mounted on a linear track, called a seventh axis or range extender, to serve several machines.

SCARA stands for Selective Compliance Assembly Robot Arm. The arm is compliant in the horizontal plane but stiff vertically, which suits pushing a pin into a hole. Typical SCARA cycle times on the industry-standard test path are well under half a second.

Delta robots, invented by Reymond Clavel in the 1980s, hang from a frame above a conveyor. Because the motors are fixed to the frame and only light carbon-fibre arms move, they can exceed 10 g of acceleration and perform more than 150 picks a minute on light products.

Gantry robots are built from linear modules. Their envelope is a rectangular box, which matches pallets, machine beds and storage racks. Their accuracy is limited mainly by the rails and drives rather than by a chain of rotary joints, so large gantries can hold tight tolerances over several metres.

Reading a robot datasheet

A datasheet compresses a robot into a page of numbers. Integrators learn to read them sceptically, because each number is measured under specific conditions.

Repeatability and accuracy are often confused. ISO 9283 defines pose repeatability as the radius of a sphere that contains the measured positions when the robot returns many times to the same commanded pose, calculated as the mean distance from the centroid plus three standard deviations. Pose accuracy is the distance between the commanded pose and the centroid of the reached poses. A taught point relies on repeatability only, because the operator moved the robot to the real location and recorded the joint angles. A point calculated from a CAD model or from a camera relies on accuracy, which depends on how well the robot's kinematic model matches the real arm. Calibration, covered in Lesson 4, closes that gap.

# ISO 9283-style repeatability and accuracy from simulated measurements
import numpy as np
rng = np.random.default_rng(440)
commanded = np.array([800.0, 200.0, 500.0])            # mm, commanded position
bias = np.array([0.35, -0.20, 0.15])                   # systematic model error, mm
pts = commanded + bias + rng.normal(0, 0.006, size=(30, 3))
centroid = pts.mean(axis=0)
l = np.linalg.norm(pts - centroid, axis=1)
RP = l.mean() + 3 * l.std(ddof=1)
AP = np.linalg.norm(centroid - commanded)
print(f"repeatability RP = {RP:.3f} mm")
print(f"accuracy AP      = {AP:.3f} mm")

repeatability RP = 0.022 mm
accuracy AP      = 0.431 mm

Payload and wrist load diagrams

Rated payload alone is not enough. Every manufacturer publishes a load diagram that shows the permitted mass as a function of how far the load's centre of gravity sits from the tool flange, both along the flange axis and sideways. A robot rated for 20 kg may allow the full 20 kg only when the centre of gravity is within about 100 mm of the flange, and much less at 300 mm. The wrist motors also have limits on static torque and on moment of inertia about each wrist axis. Lesson 3 shows how to check a gripper and part against these limits. For now, the rule is that the payload includes everything mounted on the flange: tool changer, gripper, hoses, sensors and the heaviest part, with a margin for future changes.

Applications that dominate the market

Industrial robots are concentrated in a handful of applications. Handling, which includes machine tending, palletizing, packing and pick-and-place, is the largest category in IFR statistics, followed by welding and soldering, assembly, and dispensing and painting.

Arc welding robots carry a welding torch and wire feeder, often working with a positioner that rotates the part so that every weld is made in the flat position. Spot welding robots in automotive body shops carry heavy servo guns and are among the largest robots made. Painting robots are built with explosion-protected motors, or pressurized housings, so that they can operate in solvent-laden air, and use rotary bell atomizers with electrostatic charge to improve transfer efficiency. Machine tending robots load and unload CNC lathes, mills, presses and moulding machines, an application that grew rapidly with collaborative arms because many small shops have one or two machines and no room for a fenced cell. Palletizing robots stack cases at the end of packaging lines. Assembly robots insert, press, screw and snap components together, often with force sensing. Dispensing robots apply adhesives and sealants along a path whose speed must be tightly controlled for a consistent bead.

Each application drives the robot choice. Palletizers, for example, are often four-axis robots because the tool only needs to stay horizontal and rotate about a vertical axis; removing the wrist axes increases payload and speed for the same cost. A spot welding robot needs a large payload for the gun and a stiff structure to resist reaction forces. A painting robot needs a hollow wrist so that paint and air lines can run inside the arm without catching on the part.

Throughput: the first feasibility check

Before a detailed design, engineers estimate whether a robot can keep up. If a production line must produce a certain number of parts per shift, the available time divided by the required number of parts gives the takt time, the maximum time per part. The robot's cycle time must be shorter than takt, with margin for variation, faults and future growth.

# Quick feasibility: robots needed for a picking rate with variability
import numpy as np
rng = np.random.default_rng(440)
demand_per_min = 120            # products arriving per minute
pick_cycle_s = rng.normal(0.55, 0.06, 100_000)   # delta robot pick cycles, s
mean_rate = 60 / pick_cycle_s.mean()
p95_rate = 60 / np.percentile(pick_cycle_s, 95)
print(f"mean picks per robot per min: {mean_rate:.0f}")
print(f"rate at slow (95th pct) cycle: {p95_rate:.0f}")
for n in (1, 2, 3):
    print(f"{n} robot(s): utilization {demand_per_min / (n * mean_rate):.0%}")

mean picks per robot per min: 109
rate at slow (95th pct) cycle: 92
1 robot(s): utilization 110%
2 robot(s): utilization 55%
3 robot(s): utilization 37%

High utilization looks efficient on paper but leaves no room for slow cycles, missed picks or restarts after a fault. Planners commonly aim for 75 to 85 per cent utilization of the bottleneck robot, a rule of thumb that Lesson 10 justifies with availability and OEE numbers.

Market trends

Several trends shape industrial robotics in the mid-2020s. Collaborative applications continue to grow, particularly in small and medium enterprises. Vision and AI-based picking have made it practical to handle mixed parts that once required careful fixturing. Robot prices per unit have fallen while integration costs, including tooling, guarding, programming and commissioning, often make up two-thirds or more of total project cost. Offline programming and simulation, discussed in Lessons 2 and 4, are now standard for large projects. Labour shortages in manufacturing and logistics in both Canada and the United States have accelerated adoption, and government incentives for capital investment, covered in Lesson 11, affect the payback calculation. At the same time, many robots still perform the same welding and handling tasks that defined the field in the 1980s, and the fundamentals of reach, payload and cycle time have not changed.

Industrial robot density, the number of robots per 10,000 manufacturing employees, is a common international comparison. IFR reports have placed South Korea, Singapore, China, Germany and Japan near the top, with the United States and Canada lower but rising. Density reflects industry mix as much as policy: economies dominated by automotive and electronics manufacturing naturally have more robots.