Collaborative Robotics is the engineering discipline concerned with designing, deploying, and operating robotic systems — particularly collaborative robots (cobots) — that share workspace and tasks with human workers without requiring physical barriers, relying instead on force-torque sensing, speed-and-separation monitoring, and power-and-force limiting to maintain safety under ISO 10218 and ISO/TS 15066. Unlike traditional industrial robots that operate in guarded cages, cobots are designed with compliant joints, rounded profiles, and real-time collision-detection to enable direct physical cooperation with humans on assembly, inspection, and logistics tasks. The discipline integrates mechanical design, control theory, human-robot interaction research, and regulatory compliance to enable safe human-robot co-presence in shared workspaces.
Overview
- Collaborative Robotics emerged as a distinct engineering discipline in the mid-1990s, formalised by J. Edward Colgate and Michael Peshkin at Northwestern University who coined the term “cobot” in 1996. The concept was a direct response to the rigidity of classical Industrial Robot deployments, which required expensive safety guarding, extensive integration, and high volumes to justify investment.
- The core insight was that a robot designed to be inherently safe through mechanical compliance and sensing — rather than through physical isolation — could operate alongside human workers, combining robotic repeatability and strength with human dexterity and judgement.
- Cobots have become a mainstream category within Manufacturing Automation, occupying deployments in small and medium enterprises, research laboratories, healthcare facilities, and logistics centres.
Key Components
- Force-Torque Sensing
- Force Torque Sensor units measure contact forces and torques at joints or the wrist, enabling the controller to detect unexpected collisions and yield compliantly.
- Joint-level torque sensing provides whole-arm collision detection without additional external hardware.
- Compliant Control Architectures
- Impedance Control and admittance control regulate the dynamic relationship between force and motion, allowing the robot to feel soft in response to external disturbances.
- Position-based impedance control is widely used for assembly tasks requiring gentle insertion.
- Safety-Rated Monitoring Systems
- Cobot Safety Levels define four ISO-specified collaboration modes: safety-rated monitored stop, hand-guiding, speed-and-separation monitoring, and power-and-force limiting.
- Safety-rated monitored stop halts the robot when a human enters the collaboration zone; hand-guiding allows direct physical teaching.
- Speed-and-separation monitoring dynamically reduces robot velocity as a human approaches, using Computer Vision or time-of-flight sensors to estimate proximity.
- Power-and-force limiting ensures contact forces and pressures remain below biomechanical injury thresholds at all times.
- End-Effector Tooling
- End Effector design is critical: cobot-compatible grippers, suction cups, and tool changers must themselves be compliant or collision-safe.
- Quick-change tooling systems allow a single cobot arm to serve multiple tasks within a shift.
- Perception Systems
- Computer Vision — RGB-D cameras, structured light, and stereo vision — provides workspace awareness, part localisation, and human proximity detection.
- Tactile Sensing at the fingertips or palm enables fine manipulation and grasp quality assessment.
- Motion Planning
- Motion Planning algorithms generate collision-free trajectories that respect both workspace geometry and human presence zones.
- Online replanning capability is essential for reactive collaboration where human position changes continuously.
Collaboration Modes (ISO/TS 15066)
- Safety-Rated Monitored Stop (SRMS): Robot is stationary while human works in the shared zone; resumes automatically when zone is clear.
- Hand Guiding: Human physically guides the robot through a desired trajectory; often used for programming by demonstration.
- Speed and Separation Monitoring (SSM): Robot velocity scales inversely with estimated human-robot distance; requires reliable proximity sensing.
- Power and Force Limiting (PFL): Robot moves freely but contact forces are always limited; enables true simultaneous co-working.
- Risk assessment under ISO 10218-2 and TS 15066 determines which mode is appropriate for each task.
Applications and Use Cases
- Assembly and Screwdriving
- Cobots excel at screw insertion, torque tightening, and snap-fit assembly — tasks requiring moderate force with fine positional accuracy.
- Human operators handle part feeding, orientation, and exception handling; the cobot handles the repetitive, ergonomically stressful insertion.
- Quality Inspection
- Cobot-mounted cameras and structured-light sensors conduct dimensional inspection, surface-defect detection, and barcode scanning alongside human visual checks.
- Machine Tending
- Cobots load and unload CNC machines, injection moulders, and presses, eliminating the most ergonomically harmful task in machining cells.
- Polishing and Surface Finishing
- Impedance-controlled cobots maintain constant contact force during polishing, producing consistent surface finish on complex geometries.
- Pick-and-Place and Kitting
- In warehouse and fulfilment contexts, cobots handle items with Computer Vision-guided grasping; humans handle novel or fragile items.
- Healthcare and Rehabilitation
- Assistive Robotics for physical rehabilitation employs cobot-class arms to support and guide limb movement in stroke rehabilitation.
- Surgical assistance cobots (e.g., Intuitive Surgical’s systems) operate at the boundary of collaborative and autonomous operation.
- Agriculture
- Cobot arms mounted on autonomous ground vehicles perform selective harvesting, pruning, and inspection in horticultural settings.
- Construction
- Lightweight cobot arms assist with drilling, fastening, and material handling in environments too constrained or variable for traditional automation.
AI Integration and Learning
- Imitation Learning (programming by demonstration) allows operators to teach tasks by physically guiding the robot, recording joint trajectories, and generalising from demonstrations.
- Foundation models for robotics (e.g., RT-2, OpenVLA, ACT) trained on large datasets of human manipulation trajectories are enabling zero-shot and few-shot task generalisation, substantially reducing deployment effort.
- Machine Learning-driven perception enables cobots to handle part variability that would defeat rule-based grasping: deformable objects, reflective surfaces, and cluttered bins.
- Reinforcement learning from human feedback (RLHF) applied to manipulation tasks allows cobots to refine grasping policies from operator corrections.
- Digital Twin simulations of cobot cells allow task validation, collision checking, and control tuning before physical deployment, reducing commissioning time.
Standards and Governance
- ISO 10218-1 and -2 — core standard defining requirements for industrial robot safety; Part 1 covers the robot itself, Part 2 covers integration and installation. Mandatory for CE marking in the EU and widely referenced globally.
- ISO/TS 15066:2016 — Technical Specification extending ISO 10218 to collaborative robot applications; defines the four collaboration modes and provides biomechanical injury threshold data for power-and-force limiting calculations.
- RIA TR R15.806 — US technical report providing guidance on collaborative robot safety, closely aligned with ISO/TS 15066.
- IEC 62061 and ISO 13849 — functional safety standards governing the design of safety-rated control systems (performance levels and safety integrity levels) that underpin cobot monitoring functions.
- The revision of ISO 10218 (ongoing as of 2025) is incorporating mobile collaborative robots and AI-driven adaptive robots into its scope, reflecting the convergence of cobots with Autonomous Mobile Robot platforms.
Key Vendors and Ecosystem
- Universal Robots (UR3e, UR5e, UR10e, UR20) — dominant market share, first to commercialise the cobot concept at accessible price points.
- FANUC CRX series, KUKA LBR iisy, ABB YuMi and GoFa — incumbent industrial robot makers with collaborative product lines.
- Techman Robot, Doosan Robotics, Kassow Robots — specialist cobot manufacturers.
- Rethink Robotics (Sawyer) pioneered the integrated vision and compliant arm paradigm, though the company was acquired and restructured.
- An ecosystem of End Effector suppliers (Robotiq, Schunk, OnRobot), vision systems, and integration software has formed around the UR+ and similar partner programmes.
Limitations and Open Problems
- Payload and speed trade-offs: cobots operate at lower speeds and payloads than industrial robots to stay within power-and-force limits; tasks requiring high force or high speed still require traditional guarded automation.
- Dexterous manipulation remains unsolved for unstructured environments; current cobots struggle with deformable objects, fine assembly with micron tolerances, and bimanual coordination.
- Certification burden for AI-driven adaptations: when Machine Learning changes robot behaviour at runtime, re-validation under ISO 10218 / ISO/TS 15066 is required, creating a compliance bottleneck for adaptive systems.
- Human factors: worker acceptance depends on trust, transparency of robot intent, and adequate training — poorly designed deployments generate stress rather than relief.