See the interfaces among data, control, hardware and people.
Robotics technology stack · Layer 08
Safety, verification and human oversight
A systems view of risk assessment, protective functions, semantic uncertainty, validation, operating procedures and accountable human control.
Evidence statusEstablished standards for bounded systems; frontier for adaptive general-purpose robotsLast reviewed: 11 August 2026
08
FUURAA thesis
Safety is not a model feature or emergency-stop button; it is a lifecycle argument linking hazards, controls, evidence, people and change management.
Reading method
Turn a technology label into an engineering chain, testable claims and explicit boundaries.
Translate capability into task, latency, failure and recovery.
Separate standards, independent measurement, research and first-party claims.
State what cannot be inferred from a demo, benchmark or interface.
System breakdown
Four interdependent layers determine whether the technology can enter real work.
Each layer shows its role and the failure signal most worth watching.
Hazard and risk model
Intended use, foreseeable misuse, people, tools and environment define credible harmful scenarios.
Protective architecture
Inherent design, guarding, monitored stops, limits and emergency functions reduce risk independently where required.
Verification and validation
Requirements, tests, traceability and residual-risk review show whether controls work in the specified envelope.
Human oversight
Clear authority, interfaces, staffing and escalation make intervention practical rather than nominal.
Engineering evaluation
Five checks turn abstract capability into reviewable system evidence.
Record normal performance, failure, recovery and human cost—not only the best-looking result.
- 01
Map hazards to controls
Every material hazard needs a documented control, owner, test and residual-risk decision.
- 02
Test foreseeable misuse
Include bypass attempts, wrong tools, unexpected people, ambiguous commands and maintenance states.
- 03
Verify independence
Confirm protective layers do not silently depend on the same model, sensor, compute or network path.
- 04
Exercise human intervention
Measure detection, comprehension, decision and stop time under realistic staffing and workload.
- 05
Revalidate material change
Treat model, tool, speed, payload, map, workflow and software updates as possible safety-case changes.
Scope boundaries
State what the evidence supports—and what it does not.
- 01
ISO 10218 applies to industrial robots and explicitly excludes several domains; scope must be checked before citation.
- 02
A collaborative robot component does not make the complete application safe without integration-level assessment.
- 03
Human oversight is ineffective when operators lack time, information, authority or a safe intervention path.
Sources and evidence status
Keep the source, date, evidence identity and reading boundary visible.
This page prioritises standards bodies, public measurement programmes, official project documentation and original research disclosures.
01
ISO · 2025-02
↗02ISO 10218-1:2025 — Safety requirements for industrial robots
Sets robot-level safety requirements and distinguishes them from integration requirements in ISO 10218-2.
NIST · 2016-01-22
↗Performance Evaluation of Human Detection Systems for Robot Safety
Shows how quantitative detection and tracking evidence can inform collaborative-workspace protection.