Skip to content
Tutorials Center
My learning

Topics

Lesson 34 of 38 ยท Applications and Safety

Robot Safety, Standards and Ethics

Learn how to make robots safe: hazards, risk assessment, e-stops, speed and separation monitoring, watchdogs, standards such as ISO 10218, and ethics.

A robot is a machine that can move with force, on its own timing, sometimes near people. A bug in a web page shows a wrong number. A bug in a robot can break a wrist, a window, or a bone. Safety is therefore not a final polish step; it is part of the design from the first sketch.

This lesson introduces the vocabulary and the main techniques: how to find hazards, how to reduce risk, how to build a protective stop that actually works, and how to think about the ethical side of robots.

This is a learning overview, not a substitute for the safety standards, a risk assessment by qualified people, or the laws of your country. Never test a robot that can hurt someone without an accessible emergency stop and a person ready to use it.

Hazards and risk

A hazard is a source of possible harm: a spinning blade, a heavy arm swinging through a person's space, stored energy in a spring, a hot surface, high voltage, a pinch point between two links, a battery fire, a drone falling. Risk combines how severe the harm could be with how likely it is.

A basic risk assessment follows four steps:

  1. Identify hazards in every mode: normal operation, setup, maintenance, fault, misuse.
  2. Estimate severity and probability for each.
  3. Reduce the risk, in this order of preference: eliminate by design (lighter arm, lower speed, rounded corners), guard (fences, light curtains, safety scanners), inform (signs, training, procedures). Warnings are the weakest measure.
  4. Verify that the measures work, then document and review whenever something changes.

Reduce risk in this order

Design it out Less speed, force, mass, sharpness.
Guard it Fences, scanners, interlocks.
Inform it Labels, training, procedures.

Stronger measures first. Signs alone are not safety.

1 / 2

Emergency stop, protective stop and safe torque off

  • An emergency stop (E-stop) is a red, mushroom-shaped, latching button that removes power to the motors, regardless of software. It must work even if the computer has crashed. It is wired through hardware (a safety relay), not just read by a program.
  • A protective stop is a controlled stop triggered by a safety function (a person entering a zone) that can restart after the condition clears.
  • Safe Torque Off (STO) is a drive feature that disables torque in a way certified for safety.

Stop categories describe how the stop happens: remove power immediately (the arm may coast), or brake in a controlled way and then remove power. Controlled stops are gentler; the right choice depends on the machine.

A key point: software that "decides to stop" is not enough. The safety function must live in separate, simple, certified hardware or a safety-rated controller that does not depend on the application code.

Stopping takes time and distance

When a safety sensor detects a person, the robot does not stop instantly: the sensor, the controller and the brakes each take time. The distance you must leave is roughly:

S = v * (t_react + t_brake) + d_brake + d_intrusion

where v is the robot's speed toward the person, t_react the sensor and controller delay, t_brake the time brakes take to engage, d_brake the stopping distance, and d_intrusion an extra margin for how far a person can reach into the zone before detection.

Python Minimum safe distance for a mobile robot

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

This is why safety scanners on mobile robots define zones: a large warning zone that slows the robot, and a smaller protective zone that stops it, with their sizes tied to the speed.

Speed and separation monitoring

Collaborative systems often use speed and separation monitoring (SSM): the closer a person is, the slower the robot must move; inside the minimum separation distance it stops. Another approach is power and force limiting (PFL), where the robot is designed so that contact forces stay below injury thresholds, which is how many cobots work without fences. A third is hand guiding, where a person moves the arm by hand at a safe speed.

Python Speed and separation monitoring controller

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

In the output the speed limit falls as the person approaches and the watchdog forces a stop when the sensor data goes stale, even though the last known distance looked safe. The limit recovers when fresh data shows the person leaving. A real system adds filtering, protective stop latching, and independent redundant sensing.

Functional safety and standards

Safety functions must themselves be reliable. Functional safety standards define how to design them and how reliable they must be, using levels (for example Performance Level a to e in ISO 13849, or SIL 1 to 3 in IEC 61508 and IEC 62061). Higher levels need redundancy, diagnostics and testing.

Names worth knowing (always check the current editions and your region's laws):

Standard Covers
ISO 10218 Safety requirements for industrial robots and robot systems
ISO/TS 15066 Collaborative operation, including force and pressure limits
ISO 13482 Personal care robots (mobile servants, physical assistants)
ISO 3691-4 Driverless industrial trucks and their systems
ISO 13849 / IEC 62061 Design of safety-related control systems
IEC 60204-1 Electrical equipment of machines, including E-stop wiring
IEC 60601 Medical electrical equipment
CE marking, UL, FDA Regional product compliance and approval

Drones and road vehicles have their own aviation and traffic rules. Using a standard is not just paperwork: it records decades of accident experience.

Software safety habits

  • Fail safe. The default on any loss (power, link, sensor, heartbeat) is a safe stop, not "keep going."
  • Limit everything. Maximum speed, acceleration, torque, joint range, workspace. Enforce limits in the lowest layer, not just the planner.
  • Test in simulation, then slowly. Run a new program at low speed with the E-stop in hand, in an empty area.
  • Check sensors and localization health. If confidence drops, slow down or stop. See localization.
  • Never rely on a single sensor for a protective function.
  • Cybersecurity is safety. A robot that can be hijacked remotely is a hazard: update software, use authentication, and separate control networks.
  • Log decisions so incidents can be understood.

Ethics and society

Robots raise questions beyond engineering.

  • Responsibility. When an autonomous machine causes harm, who is accountable: the maker, the operator, the owner? Clear responsibility and traceable decisions matter.
  • Privacy. Robots with cameras and microphones in homes and offices collect sensitive data. Collect the minimum, protect it, and be transparent.
  • Bias and fairness. Perception systems trained on narrow data can fail for some people or conditions. Test broadly.
  • Work and society. Automation changes jobs. Responsible deployment includes retraining and honest communication.
  • Deception and attachment. Humanlike robots can mislead people about what they understand or feel.
  • Dual use and weapons. The same technology serves rescue and war; many engineers and societies argue that lethal force decisions must stay under meaningful human control.
  • Transparency. People near a robot should be able to tell what it is doing and how to stop it.

Isaac Asimov's "Three Laws of Robotics" are a famous idea from fiction, not an engineering method. Real systems need specific, testable requirements, not slogans.

Check your understanding

Quick check ยท 4 questions

of 4 answered You scored / 4 ยท perfect!

  1. 1. Which risk-reduction measure should come first?

    Eliminating the hazard by design beats guarding, which beats information. Signs are the weakest control.
  2. 2. Why must an E-stop not depend only on the robot's application software?

    The stop path must remain effective when the main computer fails, so it uses independent safety hardware.
  3. 3. A robot's speed doubles. What happens to its braking distance at the same deceleration?

    Braking distance is v squared divided by 2a, so doubling v multiplies it by four.
  4. 4. What does a watchdog do in a safety design?

    A missing heartbeat or sensor message is treated as a fault, producing a safe stop.

Summary

  • Safety starts at design: identify hazards, estimate risk, and eliminate or guard before warning.
  • E-stops and safety functions must be independent of ordinary software; stopping needs time and distance that grow with speed squared.
  • Speed and separation monitoring, force limiting, and watchdogs make shared workspaces possible.
  • Standards (ISO 10218, ISO/TS 15066, ISO 13849 and others) capture accumulated experience; follow the current local rules.
  • Ethics includes responsibility, privacy, fairness, employment, and human control over lethal force.

Next: Project: A Grid Delivery Robot