From Fenced Cells to Shared Workspaces: Choosing the Right Human-Robot Collaboration Architecture for Your Line

Choosing the right human-robot collaboration (HRC) architecture is crucial for manufacturing lines, depending on how often humans need to enter the robot's workspace. The article outlines progressive collaboration levels, from fenced cells to simultaneous work, and provides a framework for selecting the appropriate pattern based on task demands and safety requirements.

Kaleo Kekoa

October 29, 2026 · 5 min read

Human-robot collaboration in manufacturing is not one architecture but a spectrum, and the right choice depends on how often a person actually needs to enter the robot's space. At one end sits the classic fenced cell, where hard guarding eliminates contact entirely; at the other, a person and a robot work simultaneously on the same part. Robot manufacturer KUKA, whose engineering guidance lays out six progressive stages of collaboration, frames the decision bluntly: "it is the application" that should drive the configuration, not the robot's capabilities. This guide walks through the recognized collaboration levels, the safety mechanisms that make each one permissible, and a practical matrix for matching a pattern to your task.

Key Architectural Patterns for HRC: A Comparative Framework

Start with a definition. Researchers at Queen's University Belfast, who published a taxonomy of human-robot interaction, note that "cobot" is used universally for robots with built-in safety features that enable collaborative applications — but the robot itself is only half the story. The architecture of the collaboration — where the human works, when, and under which safety regime — is what actually defines the system. Their review also found that numerous classification frameworks exist, using three or four terms to describe interaction levels, so you will see slightly different vocabularies across vendors and standards bodies. The underlying progression, however, is consistent: from physical separation, through temporal separation, to genuine simultaneous cooperation.

Stage 1: Separated operation with fixed guarding

The baseline pattern is the one most factories already run. A fixed safety fence — hard guarding — separates the operator from the robot completely, eliminating the possibility of contact. KUKA describes this as the classic configuration and notes that it still covers the majority of robot applications. If your task never requires a person to enter the robot's envelope, this remains a perfectly legitimate architecture, and it buys you the simplest safety case: no contact can occur because no shared space exists.

Stage 2: Occasionally shared workspace with virtual guarding

The first step toward collaboration keeps the fence but makes it virtual. A laser-based system replaces the physical barrier, defining a separation zone around a workspace that is only occasionally shared — for refilling components or changing out pallets, for example. The governing mechanism here is a safety-rated monitored stop: power stays on, but contact is only possible with a stationary robot. The robot does not need a full shutdown-and-restart cycle when a person steps in; it holds position safely and resumes when the person clears the zone. This pattern suits tasks where human intervention is predictable but infrequent.

Stage 3: Frequently shared workspace with speed and separation monitoring

When intervention becomes routine, the architecture shifts again. In this pattern, the workspace is intended to be shared, with frequent operator presence — KUKA's example is an operator who manually loads and unloads parts in the work cell. Safety is maintained through speed and separation monitoring: the system continuously tracks the distance between human and robot and slows the robot as a person approaches, with contact again only possible when the robot is stationary. This is the first level where the safety mechanism is dynamic rather than a static barrier, responding to where the person actually is rather than assuming they stay outside.

Toward simultaneous collaboration and autonomy

Beyond monitored separation lies true simultaneous work — human and robot occupying the same space at the same time, whether on different parts or the same workpiece. KUKA's staged model continues toward collaborative robots that respond to the human's actions and, ultimately, autonomous collaborative systems. Each step up this ladder trades simplicity for flexibility: the safety case grows more demanding as physical and temporal separation disappear, and the engineering burden shifts from barriers to sensing, monitoring, and control.

HRC Collaboration Model Selection Matrix

Use this matrix to map your task's interaction pattern to the corresponding roles and safety mechanism. Read it left to right: identify how often the human needs access, then check the safety regime that pattern requires. The rows follow the progression established in KUKA's staged guidance and the interaction-level frameworks catalogued in the Belfast taxonomy review.

HRC collaboration patterns mapped to roles, safety mechanisms, and typical manufacturing tasks
Collaboration Level	Human Role	Robot Role	Key Safety Mechanism	Typical Manufacturing Tasks
Coexistence / separated operation	Works entirely outside the robot's guarded zone	Performs its tasks autonomously behind fixed guarding, isolated from human presence	Fixed safety fence; contact is eliminated by physical separation	Continuous automated processes with no operator access needed
Sequential collaboration (occasionally shared workspace)	Enters the workspace occasionally to intervene	Runs autonomously, then holds at a safety-rated monitored stop while the person is inside	Virtual guarding (laser-based separation); safety-rated monitored stop — power stays on, contact only possible with a stationary robot	Refilling components, changing out pallets
Cooperation in a frequently shared workspace	Manually loads and unloads parts in the work cell on a routine basis	Operates in the same cell, slowing in response to the operator's proximity	Speed and separation monitoring; robot slows as the human approaches, contact only possible with a stationary robot	Cells where operator intervention is frequent and intended
Simultaneous collaboration and autonomous systems	Works alongside the robot at the same time, on the same or different parts of the workpiece	Responds to the human's actions and, at the most advanced stage, acts autonomously	Dynamic, sensing-driven safety regimes beyond static separation — the most demanding safety case in the progression	Shared work on workpieces where human dexterity and robot capacity must combine in real time

HRC collaboration patterns mapped to roles, safety mechanisms, and typical manufacturing tasks

Collaboration Level

Human Role

Robot Role

Key Safety Mechanism

Typical Manufacturing Tasks

Coexistence / separated operation

Works entirely outside the robot's guarded zone

Performs its tasks autonomously behind fixed guarding, isolated from human presence

Fixed safety fence; contact is eliminated by physical separation

Continuous automated processes with no operator access needed

Sequential collaboration (occasionally shared workspace)

Enters the workspace occasionally to intervene

Runs autonomously, then holds at a safety-rated monitored stop while the person is inside

Virtual guarding (laser-based separation); safety-rated monitored stop — power stays on, contact only possible with a stationary robot

Refilling components, changing out pallets

Cooperation in a frequently shared workspace

Manually loads and unloads parts in the work cell on a routine basis

Operates in the same cell, slowing in response to the operator's proximity

Speed and separation monitoring; robot slows as the human approaches, contact only possible with a stationary robot

Cells where operator intervention is frequent and intended

Simultaneous collaboration and autonomous systems

Works alongside the robot at the same time, on the same or different parts of the workpiece

Responds to the human's actions and, at the most advanced stage, acts autonomously

Dynamic, sensing-driven safety regimes beyond static separation — the most demanding safety case in the progression

Shared work on workpieces where human dexterity and robot capacity must combine in real time

Notice the pattern in the safety column: each level replaces a cruder guarantee with a more dynamic one. The fence makes contact impossible; the monitored stop makes contact possible only with a stationary robot; speed and separation monitoring makes the robot's motion itself respond to the human. The more often a person needs to be in the cell, the more of that dynamic machinery — sensors, monitoring logic, rated safety functions — your architecture has to carry.

Selecting the Right HRC Architecture for Your Task

Begin with two questions. First, how often does a person actually need to enter the robot's workspace? If the answer is never, stop at stage one and keep the fence — it is the cheapest safety case you will ever write. If intervention is occasional and predictable — replenishing stock, swapping pallets — the virtual-guarding pattern with a safety-rated monitored stop removes the physical barrier without requiring the robot to shut down. If the operator is in the cell constantly, loading and unloading by hand, you need speed and separation monitoring, and your engineering effort moves into sensing and safe-motion control.

Second, what does the task itself demand of each party? KUKA's guidance warns against putting the emphasis on the robot rather than the application — the reverse is true, in its words. Match the work, not the machine. Tasks that need human dexterity or judgment stay with the person; repetitive, heavy, or high-precision steps go to the robot; and the collaboration level is chosen to make that handoff safe, not to showcase the robot's collaborative features. The ISO technical specification ISO/TS 15066:2016 governs the technical requirements for collaborative robot applications, including the speed-and-separation and monitored-stop techniques these patterns depend on, so any level beyond the fixed fence should be designed against that specification's requirements for your jurisdiction. (This article explains the technical requirements only; it is not a legal interpretation.)

Be honest about the limits of current systems when your task is complex or variable. A 2024 study published in Scientific Reports by Yuchen Ji and colleagues states plainly that "existing HRC systems lack the transfer and generalization capability for environment perception and task reasoning." The authors detail two concrete manifestations: current methods rely on specialized models to perceive scenes and require retraining when facing unseen objects, and they only address predefined tasks, without support for undefined task reasoning. In practice, that means an HRC cell tuned to one part and one sequence will not gracefully absorb a new product variant or an unplanned intervention. Budget for that rigidity — or for the retraining and integration work it implies — when you choose a more interactive pattern.

From Fenced Cells to Shared Workspaces: Choosing the Right Human-Robot Collaboration Architecture for Your Line

Key Architectural Patterns for HRC: A Comparative Framework

Stage 1: Separated operation with fixed guarding

Stage 2: Occasionally shared workspace with virtual guarding

Stage 3: Frequently shared workspace with speed and separation monitoring

Toward simultaneous collaboration and autonomy

HRC Collaboration Model Selection Matrix

Selecting the Right HRC Architecture for Your Task

Implementing HRC: A Decision Framework

Sources

Tags

Related articles

Choosing Between LiDAR, Visual, and Multi-Modal SLAM for Your Industrial Mobile Robot

Choosing an RTOS for Robotics: How to Classify Your Tasks Before You Pick a Kernel

How Digital Twins Simulate, Optimize, and Predict Robotic Manufacturing Performance

Parallel vs. Compliant Grippers: How to Choose the Right End-Effector for Precision Assembly