Module 10 – Automotive/ADAS Safety-Critical Testing

Module Project: End-to-End ADAS Pipeline Verification

End-to-end ADAS pipeline verification: ASIL-driven static review plus a scenario-based dynamic test suite (including a fault-injection case) for a camera-to-decision mini-ADAS pipeline.

The module project combines safety-oriented static review with dynamic scenario evidence.

Participants progress through:

Hazard / Safety Context
↓
ASIL Determination
↓
Static Analysis & MISRA-C Review
↓
Sensor Inputs
↓
Sensor Fusion
↓
Detection / Decision
↓
Scenario-Based Testing
↓
Fault Injection
↓
CAN Actuation Validation
↓
ADAS Verification Evidence

This makes the module project significantly more valuable than an isolated tool exercise because it connects multiple verification activities around a small ADAS pipeline.

From General AI/ML Validation to ADAS Verification

Module 9 develops general verification and validation methods for AI/ML systems, including data quality, model outputs, robustness and regression.

Module 10 moves those verification ideas into a specific automotive/ADAS context.

The progression becomes:

AI/ML Validation
↓
Automotive Safety Context
↓
ASIL-Based Verification Thinking
↓
Sensor Fusion & Detection
↓
Driving Scenarios
↓
Fault Behaviour

This relationship should be clearly visible without making Module 10 another generic AI/ML page.

ISO 26262 and ASIL Determination

The module introduces ISO 26262 and focuses practically on ASIL determination using severity, exposure and controllability.

Participants apply these concepts to a sample automotive hazard and use the resulting safety context when triaging technical findings.

Keep the wording precise.

This page should say:

ISO 26262 overview and ASIL determination

rather than claiming:

ISO 26262 certification
ISO 26262 compliance qualification
Certified ISO 26262 training

Automotive Static Analysis and MISRA-C

Module 10 builds on the static-analysis skills developed earlier in the programme but applies them specifically within an automotive context.

Participants:

• Run cppcheck/clang-tidy
• Examine MISRA-C findings
• Review a CAN message ID conflict
• Consider spatial partitioning
• Triage findings by ASIL relevance
• Review an AUTOSAR generated-code example as an extension

This is different from Module 2, which owns broad multi-language static analysis, and Module 7, which owns general embedded static verification.

Freedom from Interference

The curriculum introduces freedom from interference through:

• Spatial partitioning
• Temporal partitioning

and asks participants to review spatial partitioning between mixed-criticality components during Lab 1.

Keep this section concise and tied to the practical exercise rather than expanding it into a separate functional-safety course.

Sensor Fusion Testing

The dynamic half of Module 10 focuses on a sample camera-plus-radar fusion workflow.

Participants validate the sensor fusion output against an expected state estimate before moving downstream into detection and decision behaviour.

This is where Module 10 begins to differentiate strongly from general embedded or AI/ML testing.

Object and Lane Detection Testing

The curriculum includes practical tests for:

Object detection

A labelled scenario including an occlusion case.

Lane detection

A challenging condition such as faded road markings.

These activities give the module explicit ADAS verification content rather than treating sensor fusion only as an algorithm exercise.

Scenario-Based ADAS Testing

Participants define a scenario with explicit pass/fail criteria and run it in simulation.

This is an important concept to expose prominently because it connects:

Driving scenario → expected behaviour → test execution → verification result

Fault Injection and Fail-Safe Behaviour

Normal driving scenarios alone are not sufficient for the module’s verification goals.

Participants inject a simulated sensor dropout and validate fail-safe behaviour.

The source also covers:

• Corrupted data scenarios
• Fail-operational vs fail-safe behaviour
• Regression testing

This creates a clear specialist relationship between ADAS functionality and abnormal/fault conditions.

CAN Actuation and End-to-End Timing

Module 10 does not stop at perception output.

Participants also validate a CAN-based actuation command produced by the decision stage.

An extension task then measures:

end-to-end sensor-to-decision pipeline latency against a timing budget.

This helps connect perception, decision and timing into the wider ADAS verification workflow.

How Module 10 Connects With Other Modules

Module 7: Embedded & Real-Time Systems Testing

Module 7 establishes general embedded verification skills including MISRA-C, QEMU, FreeRTOS, CAN, timing and fault injection.

Module 10 applies related concepts specifically to automotive and ADAS systems.

Module 9: AI/ML Verification & Validation

Module 9 develops general AI/ML validation around data, model behaviour, robustness and regression.

Module 10 adds automotive-specific sensor fusion, perception and scenario-based verification.

Module 12: Cross-Industry Standards Bridge & Capstone

This relationship is especially strong because Module 12 explicitly revisits ISO 26262, ASIL and Module 10 evidence as part of the cross-industry standards mapping.

Frequently Asked Questions

What does the Automotive/ADAS Safety-Critical Testing module cover?

The module covers ISO 26262 and ASIL concepts, automotive MISRA-C, CAN checks, freedom from interference, sensor fusion, object/lane detection, scenario-based testing, fault injection and ADAS regression testing.

Is ISO 26262 covered?

Yes. The module provides an ISO 26262 overview and includes practical ASIL determination based on severity, exposure and controllability.

What is ASIL determination in this module?

Participants determine the ASIL for a sample automotive hazard using severity, exposure and controllability and then use ASIL relevance when reviewing technical findings.

Does the module cover MISRA-C?

Yes. Participants identify an automotive-relevant MISRA-C finding during the static-analysis lab.

Is AUTOSAR covered?

Generated AUTOSAR code review is included as an extension activity.

Does the module cover sensor fusion?

Yes. Participants test a sample camera-plus-radar fusion algorithm against an expected state estimate.

Are object and lane detection tested?

Yes. The module includes object-detection testing with an occlusion case and lane-detection testing under a challenging condition.

Does the module include scenario-based testing?

Yes. Participants define an ADAS scenario with explicit pass/fail criteria and run it in simulation.

Is fault injection included?

Yes. Participants inject a simulated sensor dropout and validate fail-safe behaviour. Corrupted-data scenarios are also included in the module topics.

Is CAN testing covered?

Yes. The static lab includes CAN message ID conflict review, while the dynamic lab validates a CAN-based actuation command.

Does the module cover timing?

Yes. An extension activity measures end-to-end sensor-to-decision pipeline latency against a timing budget.

What is the Module 10 project?

The project is an End-to-End ADAS Pipeline Verification exercise combining ASIL-driven static review with a scenario-based dynamic suite, including fault injection, for a camera-to-decision mini-ADAS pipeline.