Collaborative robot (cobot)
A collaborative robot, or cobot, is an industrial robot arm designed to share a workspace with people, using built-in features such as force sensing, power and force limits, and speed monitoring to reduce the risk of contact. This is often called human-robot collaboration (HRC). Under the 2025 editions of ISO 10218-1 and ISO 10218-2, collaboration is treated as a property of the whole application, not of the robot alone.
In this article
Building visual or physical AI?
Let's talk.
What is a collaborative robot?
Traditional industrial robots run fast behind fences. A collaborative robot, or cobot, is designed to share space with people instead. Cobots typically trade some speed and payload for built-in safety features, such as force and torque sensing that stops the arm on contact, rounded shapes, and limits on speed and power. Many can be taught by hand, by moving the arm through a motion and recording it.
Common cobot payloads run from about 3 kg to around 30 kg, which covers machine tending, packaging, screwdriving, and many pick-and-place jobs. Cobot arms are also common in robot learning research, where labs collect teleoperated demonstrations on them.
Key takeaways
- A cobot is a robot arm designed to work near people, with safety built into how it senses and moves.
- Under the 2025 editions of ISO 10218-1 and ISO 10218-2, collaboration describes the application, not the robot alone.
- Cobots are widely used for machine tending and in robot learning labs.
How it works
A cobot stays safe through a mix of hardware, software, and how the task is set up. Common ways a cobot shares space with people include:
- Power and force limiting: the arm limits the force it can apply, so contact with a person stays below set thresholds.
- Speed and separation monitoring: sensors in the cell track how close people are, and the robot slows or stops as they approach.
- Monitored standstill: the robot holds still while a person works in the shared space. Older standards called this a safety-rated monitored stop, and the 2025 revision treats it as a safety function, not a collaborative method of its own.
- Hand guiding: an operator moves the arm directly, often to teach it a motion.
The 2025 editions of ISO 10218-1 and ISO 10218-2, the core industrial robot safety standards, folded most of the earlier cobot guidance in ISO/TS 15066 into Part 2, and the US standard ANSI/A3 R15.06-2025 follows the same changes. They also treat collaboration as a property of the application. A cobot holding a sharp tool or a heavy part can still need guarding, so each work cell gets its own risk assessment.
Why it matters
Cobots made robot arms affordable and approachable enough for small factories and research labs, which is part of why so much robot learning data comes from them. That data carries the cobot's safety behavior with it. Protective stops, speed limits near people, and episodes cut short by an emergency stop all show up in the log. Separating those from genuine task failures, while keeping both in view, makes a dataset more honest about how the robot really performs.
Frequently asked questions
What is the difference between a cobot and an industrial robot?
A cobot is designed to work near people, with built-in sensing and limits on force and speed. A traditional industrial robot is usually faster and stronger and runs behind guarding. Under current standards, what makes an application collaborative is the full setup, not the robot alone.
Do cobots need safety fencing?
Often not, but it depends on the task. A risk assessment decides what safeguarding a cobot application needs, and sharp tools, heavy parts, or high speeds can still call for guarding.
What are cobots used for?
Machine tending, packaging and palletizing, screwdriving, inspection, and lab automation are common uses. Many robot learning datasets are also collected on cobot arms.
Related terms

Director of Growth
Jesse Mostipak is the Director of Growth at Voxel51, where the work is helping humans find and trust what the brand knows, and teaching the Google knowledge graph and the LLMs answering on their behalf to do the same. That question, how knowledge gets built inside a system, is one Jesse has been chasing for years. Earlier versions of it ran through a New York City high school science classroom, data science and machine learning, and developer relations at Kaggle, Posit (formerly RStudio), and Baseten. The answer doesn't change much depending on whether the learner is a teenager, a software engineer, or a knowledge graph. Jesse holds a Master's in Education from CUNY Hunter College. LinkedIn
See all articles by Jesse Mostipak
Last updated November 19, 2026