7 Types of Cobots Explained With Real-World Applications
The seven practical cobot categories are power-and-force-limited arms, monitored-stop applications, speed-and-separation applications, hand-guided systems, dual-arm cobots, mobile manipulators and high-payload cobots. They are not seven mutually exclusive robot species. Some describe a safety method; others describe a machine architecture. The right choice starts with the complete application risk—not the arm brochure.
Choose the collaboration concept first; choose the robot model second.
A cobot arm is only one component of a collaborative application. Grippers, workpieces, welding torches, lasers, fixtures, conveyors, scanners and operator behavior can all create hazards that the robot’s built-in safety functions do not remove. Start by defining how and when people enter the workspace, then select the architecture that can meet payload, reach and cycle-time requirements inside the validated safety concept.
Useful for lower-force shared-space tasks, frequent changeovers, inspection and machine tending—but still subject to an application risk assessment.
Useful when human access is intermittent and the cell can use safety scanners or other protective devices to control speed and separation.
Choose these when two-handed manipulation or movement between workstations creates value that a fixed single arm cannot deliver.
Robot, tool, part, process energy, fixture, access and foreseeable misuse must be evaluated together before claiming collaborative operation.
From safety concept to production cell
A cobot is not automatically a safe, fence-free application.
ISO describes collaborative robotics as an application in which a robot system and people share a workspace. That definition matters because the safety outcome belongs to the complete application—not to a label on the arm. The same robot can run one task as a power-and-force-limited application, another behind a scanner-controlled separation zone, and a third inside conventional guarding.
The 2025 editions of ISO 10218-1 and ISO 10218-2 distinguish robot requirements from integration and application requirements. ISO/TS 15066 remains a published technical specification while a revision project is underway. Your local laws, adopted standards and customer requirements determine the compliance route.
Four collaborative safety methods sit beneath the seven buyer-facing types.
These methods can be combined within one application. Their feasibility depends on the risk assessment, validated safety functions, stopping performance, separation distances, contact measurements and hazards created by the tool and workpiece.
Safety-rated monitored stop
The robot is held in a monitored standstill while a person occupies the collaborative workspace. It is useful for sequential work: the robot performs a task, stops safely, and the operator enters to load, inspect or adjust. It is not simultaneous motion beside a person.
Hand guiding
After the required safe state is established, an operator uses a hand-guiding device to command robot motion. This can simplify teaching, positioning and skilled process setup, but the tool, part and potential trapping points still require assessment.
Speed and separation monitoring
Safety-rated sensing tracks people relative to the robot. The system changes speed or stops before protective separation is violated. Performance depends on validated sensing, reaction time, stopping distance and realistic approach paths.
Power and force limiting
The application limits forces and pressures during foreseeable contact. This requires more than torque sensing: geometry, speed, payload, tooling, contact area, clamping and body region all influence whether contact is acceptable.
OSHA’s robotics technical guidance emphasizes application-specific hazard analysis and risk assessment, including end-effector and workpiece hazards. It also notes that PFL and SSM are commonly combined.
Compare the seven practical cobot types by the problem they solve.
| Practical type | What defines it | Best application pattern | Main strength | Main constraint |
|---|---|---|---|---|
| 1. PFL cobot arm | Contact-limiting robot design and safety functions | Shared-space tending, assembly, inspection | Flexible access and redeployment | Safe speed and force can limit cycle time |
| 2. Monitored-stop application | Robot stops safely before a person enters | Sequential loading, checking or changeover | Robot can run faster when zone is clear | Frequent entry can destroy throughput |
| 3. SSM application | Safety-rated distance sensing controls motion | Cells with predictable intermittent foot traffic | Adaptive speed instead of every visit becoming a full stop | Sensor coverage and stopping distance need space |
| 4. Hand-guided process cobot | Operator physically commands or teaches motion | Welding path teaching, finishing, ergonomic positioning | Captures skilled path knowledge quickly | Recorded paths still need process editing and validation |
| 5. Dual-arm cobot | Two coordinated arms share a controller or workspace | Light assembly, kitting, laboratory manipulation | One arm can hold while the other works | Coordination, collision avoidance and tooling complexity |
| 6. Mobile manipulator | Cobot arm mounted on an autonomous mobile platform | Multi-station material handling and inspection | One system can serve several locations | Docking, localization, battery and floor conditions |
| 7. High-payload cobot | Collaborative-capable arm with higher payload/reach class | Palletizing, heavy tending, large-part handling | Extends collaboration to heavier loads | Higher moving mass increases kinetic and trapping risk |
This is a purchasing framework, not a standards taxonomy. A dual-arm, mobile or high-payload system still needs one or more validated safety methods for its actual task.
Seven types of cobots, their applications and their limits
Power-and-force-limited cobot arms
PFL arms are the category most buyers picture when they hear “cobot.” They commonly use lightweight structures, rounded surfaces, joint torque sensing and safety-rated motion functions. They are attractive for high-mix, lower-volume automation because operators can often access the station more easily than a conventional fenced cell.
Real applications: loading a CNC machine, presenting parts to a camera, screwdriving, packing, simple assembly and tending test equipment. The application is strongest when parts are manageable, hazards are limited and human access is frequent enough to justify a shared layout.
The common mistake is assuming the arm’s PFL capability makes every tool safe. A sharp gripper, rotating spindle, hot weldment or heavy workpiece can make direct contact unacceptable even when the robot itself can limit torque.
Safety-rated monitored-stop applications
This is an application pattern rather than a special robot shape. The robot performs the automated portion while the shared space is clear, then enters a safety-rated monitored stop before a person accesses the zone. Because collaboration is sequential, the system may support higher process speeds than a contact-limited approach when people are absent.
Real applications: a robot loads parts while an operator remains outside, then stops for gauge inspection; a palletizing station pauses while a pallet is exchanged; or a welding fixture becomes accessible only after motion and process energy are brought to the defined safe state.
Its economic performance depends on access frequency. If operators enter every few seconds, repeated stopping and restarting can overwhelm the cycle. If access occurs every few minutes, the concept can preserve productivity without designing the whole motion phase for human contact.
Speed-and-separation monitoring applications
SSM uses safety-rated sensing and control to maintain protective separation between people and hazardous robot motion. A typical design uses outer warning zones, reduced-speed zones and an inner stop boundary. The system can run faster when no person is nearby and reduce speed as someone approaches.
Real applications: palletizing beside an operator aisle, large-fixture assembly with occasional inspection access, or machine tending where staff cross the edge of a cell but do not need intentional contact with the moving robot.
The hard work is not buying a scanner; it is proving coverage and stopping performance. Racks, fixtures, pallets and the robot itself can occlude sensors. Approaches can be diagonal or unexpected. The protective distance must account for approach speed, sensing and control response, robot stopping time and position uncertainty.
Hand-guided cobots for teaching and skilled processes
Hand guiding lets an operator command robot motion through an enabling device or dedicated guiding interface after the required safe conditions are established. Its value is not that every recorded point becomes production-ready; its value is faster capture of geometry and human process knowledge.
Real applications: tracing an irregular welding seam, teaching a polishing path, positioning a heavy tool, defining inspection viewpoints or recording a spray trajectory. A skilled operator can establish a useful starting path, after which the integrator refines speed, orientation, blending, force and process timing.
Hand-guided teaching does not eliminate programming. Recorded paths may contain noise, inconsistent speed, singularities or poor tool orientation. Production replay also needs fixtures, calibration, collision checking and process validation. For welding, the seam, torch angle, stick-out, travel speed and filler delivery still require engineering.
Dual-arm cobots for coordinated manipulation
Dual-arm cobots provide two manipulators that can coordinate around a shared task. The second arm can replace fixtures in some applications: one hand holds or orients a part while the other inserts, fastens, scans or presents another component. ABB positions dual-arm YuMi for small-parts assembly, illustrating the architecture’s strength in light, precise work.
Real applications: electronics assembly, cable routing, kit preparation, vial handling, connector insertion and packaging tasks that require part stabilization. The benefit is greatest when flexible two-handed handling avoids a large family of dedicated fixtures.
The trade-off is complexity. Two arms create more self-collision possibilities, shared reach constraints and synchronization work. Payload per arm can be modest, and the business case depends on whether bimanual coordination truly removes fixtures, staging or manual hand-offs.
Mobile manipulators that combine an AMR and cobot arm
A mobile manipulator mounts a robotic arm on an autonomous mobile robot. Omron describes its MoMa concept as a cobot arm combined with an AMR for flexible production. The architecture extends one manipulator across several workstations—but adds a second automation system with its own navigation, safety and availability requirements.
Real applications: collecting samples from multiple machines, moving bins between assembly stations, loading low-volume equipment at scheduled stops, inspecting distributed assets and delivering components before performing a pick/place task.
Accurate manipulation usually requires a repeatable docking or localization strategy. Floors, ramps, doorways, Wi-Fi coverage, traffic rules and charging behavior become production variables. The arm’s payload does not tell the whole story: stability, center of gravity, base locking and dynamic limits affect the usable work envelope.
High-payload cobots for palletizing and heavy tending
High-payload cobots extend collaborative-capable platforms into case handling, larger workpieces and longer reach. As one current example, Universal Robots lists the UR20 with a 20–25 kg payload range and 1,750 mm reach. Specifications change by model and software configuration, so the application should always be checked against the current manufacturer data sheet.
Real applications: case palletizing, handling heavy castings, loading larger machine tools, positioning fixtures and moving assemblies that would otherwise require a lift assist or conventional industrial robot.
Higher payload does not mean the application can safely contact people at full speed. Moving mass, reach, center of gravity, sharp workpieces and trapping against fixed structures can dominate the risk. Many high-payload installations use collaborative-capable hardware with SSM, monitored stops or partial guarding rather than relying on contact-limited motion alone.
The end effector and workpiece can change the safety concept completely.
A smooth cobot arm holding a padded inspection camera is a different application from the same arm carrying a sharp metal part, a spinning deburring tool, a hot welding torch or an energized laser head. Before deciding on a fence-free layout, identify every process hazard and every place a person can be trapped between moving and fixed objects.
- Include gripper, adapter, cables and part in payload calculations.
- Check retention if loss of air or power could drop a load.
- Review edges, points, heat, radiation, fumes and stored energy.
- Assess predictable misuse, maintenance and recovery—not just normal cycles.
Match the cobot architecture to the production pattern.
Palletizing
Often favors high-payload arms and clear human access rules. SSM, monitored stops or guarding may be more productive than limiting every motion for contact.
Light assembly
Dual arms create value when one arm can stabilize, present or reorient a part while the other performs the process.
Mobility moves the bottleneck from reach to logistics.
A mobile manipulator can serve multiple stations, but utilization depends on traffic, docking, charging, queue logic and recovery from blocked paths. A fixed arm can be the better investment if the process needs continuous attendance at one machine. Mobility pays when travel is genuinely variable and the system can spend enough time doing productive manipulation rather than waiting or navigating.
- Measure travel time and workstation waiting separately.
- Define docking accuracy for the actual gripper and part tolerance.
- Plan charging around production peaks and shift patterns.
- Create safe recovery procedures for lost localization or blocked aisles.
Use five gates to eliminate the wrong cobot types early.
Define the process
Part, task, quality target, cycle time, operating hours, changeover pattern and failure modes.
Map human access
Who enters, when, how often, from which direction and whether contact is intended or only foreseeable.
Calculate the load
Part, gripper, adapter, hoses and cable package; then check center of gravity and inertia at required poses.
Select safety concept
Choose PFL, hand guiding, SSM, monitored stop or a combination based on the application risk assessment.
Prove the cell
Measure real cycle time, stop performance, contact/separation conditions, quality, uptime and operator recovery.
Start with must-have constraints
Payload and reach are necessary filters, but they are not sufficient. The robot must reach all process poses without living at singularities or joint limits. It must carry the complete moving load at the required orientation and acceleration. It must also meet cycle time after collaborative speed reductions, scanner zones, process dwell and operator interactions are included.
For machine tending, add door time, chuck or fixture time, part presentation and inspection. For welding, add tack loading, gas pre-flow, arc or laser-on time, filler delivery and positioner movement. For palletizing, model each layer and the worst reach—not only the easiest box position.
Price the working application—not the robot arm.
Robot prices change by payload, reach, controller, software and region. More importantly, the arm is only one line item. A meaningful business case includes engineering, end-of-arm tooling, fixtures, safety devices, process equipment, training, commissioning, maintenance and the production losses caused by changeovers or stops.
Investment boundary
- Robot, controller and mounting
- Gripper, tool changer and cable management
- Vision, scanners, safety PLC and guarding
- Fixtures, conveyors and process equipment
- Programming, risk assessment and validation
- Training, spares and facility changes
Benefit boundary
- Direct labor actually redeployed or avoided
- Additional production that demand can absorb
- Reduced scrap, rework and inspection variation
- Longer machine utilization across shifts
- Ergonomic and exposure-risk reduction
- Changeover flexibility and redeployment value
Do not turn theoretical freed labor into automatic savings. If the operator remains assigned to the cell, count only the time that can be used for measurable productive work. Likewise, value extra output only when there is demand, downstream capacity and material availability to sell or use it.
A collaborative arm does not make laser cleaning or welding safe for open human contact.
Laser cleaning, laser welding and laser marking add optical radiation, reflected beam, hot material, fumes, fire and process-specific hazards. ISO 10218-1 explicitly notes that applications such as welding, laser cutting and machining create additional hazards that must be addressed during application design. The robot’s PFL features do not control laser exposure.
In practice, a cobot can still provide valuable flexible motion while the laser process is enclosed, interlocked or otherwise engineered for the applicable laser class and local regulations. Collaborative features can simplify setup, teaching, loading or low-energy positioning—but laser emission should occur only in the validated process state.
Validate the application with representative parts and real operator behavior.
A polished demo proves that a robot can move. A useful pilot proves that the proposed system can make acceptable parts, recover from common interruptions and meet the required cycle under the intended safety concept.
Bring to the demonstration
- Representative good, bad and borderline parts
- Actual CAD, joint drawings and tolerance requirements
- Gripper contact limits and surface-finish constraints
- Target cycle time and shift schedule
- Photos or layout of operator access and adjacent equipment
- Utilities, network and environmental requirements
Record before approval
- Cycle-time distribution—not one best cycle
- First-pass quality and measurement method
- Stop, restart and fault-recovery behavior
- Operator interventions per shift
- Tool wear, calibration and maintenance tasks
- Safety assumptions requiring formal validation
A supplier demonstration supports concept selection; it does not replace the integrator’s documented risk assessment, safety validation or site acceptance process.
Start with the part, process and safety concept—not a generic cobot model.
Share the workpiece, task, payload, reach, cycle target, operator access pattern and required laser process. Oceanplayer can help outline a robotic cleaning, welding or marking test around the real production constraint.
Use the next tool that matches your automation stage.
Questions buyers ask about cobot types
The answers below use application-level safety language. Final decisions depend on the current standards adopted in your jurisdiction and a documented risk assessment.
What are the seven types of cobots?
A practical buyer-facing list is: power-and-force-limited cobot arms, safety-rated monitored-stop applications, speed-and-separation monitoring applications, hand-guided systems, dual-arm cobots, mobile manipulators and high-payload cobots. The first four relate mainly to collaborative safety methods; the last three describe hardware or system architecture. A single installation can belong to more than one category.
What is the difference between a cobot and an industrial robot?
“Cobot” usually describes a robot designed with capabilities that support collaborative applications, but collaboration is a property of the complete application. Conventional industrial robots and collaborative-capable robots are both covered by industrial robot safety requirements. The integrator decides how the robot, tool, workpiece and safeguards form a safe application.
Can every cobot operate without safety fencing?
No. A risk assessment may require scanners, interlocked guarding, partial enclosure or a conventional cell. Sharp tools, hot parts, heavy loads, crushing points, hazardous process energy and required production speeds can make unrestricted access unacceptable even when the arm has PFL functions.
Which cobot type is best for machine tending?
PFL arms are attractive when operators need frequent access and parts are manageable. A monitored-stop or SSM concept can be better when the robot needs higher clear-zone speed and human access is intermittent. The correct choice depends on door/chuck timing, payload, reach, part presentation and how often people enter the workspace.
Which cobot type is best for welding?
Hand-guided teaching can simplify path creation for high-mix welding, while a fixed arm with monitored access may provide better throughput for repeat production. Welding fumes, heat, arc or laser radiation, wire feed, sharp parts and fixtures remain process hazards. A collaborative arm does not remove the need for process-specific protection.
What is the difference between PFL and speed and separation monitoring?
PFL controls risk from foreseeable contact by limiting power, force and pressure within the validated application. SSM aims to prevent hazardous contact by maintaining protective separation through safety-rated sensing and motion control. Many applications combine them: run faster while the zone is clear, then slow or stop as a person approaches.
When does a dual-arm cobot make sense?
It makes sense when one arm must hold, orient or tension a part while the second performs another action, and when that coordination can replace fixtures or staging. If a simple fixture and one arm can achieve the same result, the single-arm solution is usually easier to program, validate and maintain.
When should I choose a mobile cobot?
Choose a mobile manipulator when one system can create value across several workstations and travel is a meaningful part of current labor. Validate docking accuracy, floor conditions, traffic, charging, wireless coverage, base stability and what happens when a destination is occupied or the route is blocked.
Are high-payload cobots safe around people?
They can support collaborative applications, but higher moving mass increases kinetic energy and can increase trapping or crushing severity. Safe use depends on payload, inertia, speed, stopping performance, gripper and workpiece geometry, separation measures and the complete risk assessment—not the payload label alone.
How should I compare cobot costs and ROI?
Compare complete installed systems and benefits that can actually be realized. Include tooling, fixtures, safety devices, engineering, process equipment, training, validation, maintenance and facility changes. Model labor that can truly be redeployed, quality savings, additional saleable output, uptime and realistic downtime rather than relying on a generic payback claim.
Primary sources used for this guide
- ISO 10218-1:2025 — safety requirements for industrial robots as partly completed machinery.
- ISO 10218-2:2025 — requirements for industrial robot applications and robot cells.
- ISO/TS 15066:2016 — collaborative industrial robot systems and work environments; ISO lists it as published and under revision.
- OSHA Technical Manual, Industrial Robot Systems and Industrial Robot System Safety — risk assessment and collaborative application guidance.
- Universal Robots UR20 product information — current example of a high-payload, long-reach cobot platform.
- ABB collaborative robot portfolio — YuMi and other collaborative architectures.
- Omron Mobile Manipulator — example of a cobot arm combined with an autonomous mobile robot.
This guide supports early application planning. It is not a machine-safety assessment, legal determination, standards certification or substitute for qualified integration and validation.