Functional Safety Certification for Bin-Picking Robots: ISO/IEC TR 5469 Explained

Date: 2027-06-28
Location: Online
Host: Zach L

About this event

A Verification and Validation Clinic for Quality Engineers pursuing functional safety certification. A live 30-minute verification and validation clinic on ISO/IEC TR 5469 (AI and Functional Safety) for bin-picking robots, built for quality engineers pursuing functional safety certification. What We'll Cover: What ISO/IEC TR 5469 (AI and Functional Safety) requires and how its scope applies to bin-picking robots Mapping ISO/IEC TR 5469 clauses to the certification evidence assessors expect System-specific hazards and safety functions typical of bin-picking robots Common findings that delay certification, and how to avoid them A practical readiness roadmap for quality engineers preparing for assessment Related topics: ISO/IEC TR 5469 · Bin-Picking Robots · functional safety certification ·

Topics

Registration

View event page

About ISO/IEC TR 5469 (AI and Functional Safety)

This event covers the compliance, engineering, and verification requirements defined by ISO/IEC TR 5469.

ISO/IEC TR 5469 addresses the use of artificial intelligence, particularly machine learning, inside or alongside safety-related functions, cataloguing the properties of AI technology that challenge classical functional safety and the methods available to mitigate them.

What it regulates

Classification of AI usage in safety applications, desirable properties (robustness, transparency, predictability), lifecycle considerations for data and models, and architectural patterns for constraining AI behavior.

Who must comply

Teams introducing ML-based perception, diagnostics or control into systems governed by IEC 61508 or its derivatives, and assessors evaluating whether AI-based elements can be justified in a safety argument.

Key requirements of ISO/IEC TR 5469

Three-dimensional classification: AI technology class, usage level in the safety function, and automation level

Analysis of properties such as degree of transparency, robustness to distribution shift, and specifiability

Mitigation approaches: architectural supervision (safety envelopes, monitors), diverse redundancy, constrained/interpretable models

Data lifecycle considerations: representativeness, labeling quality, and drift

Key concepts: ISO/IEC TR 5469

Usage level

Whether AI is in the safety function itself, supporting it, or only in non-safety roles — driving the required rigor.

Safety envelope / supervisor

A deterministic layer that bounds or overrides AI outputs, keeping system behavior within verified safe limits.

Distribution shift

Degradation when operational data departs from training data, a core AI-specific failure mechanism.

Explainability/transparency

The degree to which model behavior can be understood and analyzed, affecting verifiability.

Out-of-distribution detection

Runtime detection that inputs fall outside the model's competence, triggering safe fallback.

Frequently asked questions: ISO/IEC TR 5469

How does TR 5469 relate to ISO 8800 and ISO/IEC 42001?

TR 5469 is the cross-sector functional-safety view of AI; ISO 8800 applies AI safety specifically to road vehicles building on ISO 26262/21448; ISO/IEC 42001 governs the organizational AI management system within which such engineering work is managed.

Can machine learning be used in a SIL-rated function today?

Only with strong architectural constraints: current practice keeps the certified safety behavior in deterministic logic (monitors, envelopes, plausibility checks) while ML contributes performance, or uses ML in lower-risk supporting roles with rigorous data and validation practices.