To propose a verification and validation philosophy specifically targeted at autonomous systems in support of the system design and verification process

 

Contrary to a traditional system, in which input A should provide expected results B, an adaptive and evolving system needs different acceptance criteria.

Besides this, the traditional process for doing the verification and validation is not sufficient anymore and needs rethinking. For example; it might no longer be possible to run-through all the possible test-scenarios before launch.

 

A solution can be found in a Verification and Validation philosophy in which only "basic" examination of the system is performed up to ?Acceptance? Review, with an acceptable ?open? risk and having further ?periodic? or ?continuous? examinations (Verification Validation) after launch. As such there is continuous testing and at the same time further improvement (learning) of the system after deployment/launch.

 

For doing this, different (types of) datasets are necessary, e.g. Training data, Learnt data, Test data, Expected data and Operational data.

Next, the activity analyzes how Simulation Tools can be used to possibly generate these data sets, possibly augmenting real-world data.

 

The verification model and the revised verification and validation philosophy will be of guidance and input for the testing and verification strategies of the domain-specific implementations of autonomy.

 

Follow-on Activities:

- Usage of System-level Simulation Facilities (e.g. Digital Twin) for supporting the verification, validation and operations of systems containing autonomy.

- Usage of Artificial Intelligence for supporting the verification and validation by generating/augmenting the data-sets.