Applying virtual system integration and test to validate requirements and verify designs

Increased system complexity, however, creates even larger challenges in the verification and validation processes. To confirm that the embedded system meets quality standards and design requirements, engineers often must conduct time-consuming and resource-intensive design and code reviews, and build increasingly complex test systems. Today, the cost of verification often represents well over 50 percent of the total development cost.

To bring costs down, engineers must find and fix errors closer to the beginning of the development process, when they are easier and cheaper to fix.

Many large design organizations have discovered that a substantial majority of errors they uncover in test and integration at the end of the development process were actually introduced in the requirements themselves or were caused by misinterpreting those requirements in the design. For example, studies of safety-related software development for space systems and studies of electronic systems development show that more than 60 percent of errors found during the system test phase began as errors in the original requirements.

Virtual integration and test
With Model-Based Design, engineers can use systematic system simulation to perform requirements validation and design verification early in the design process —enabling them to detect and eliminate errors made in defining and interpreting requirements. This virtual system integration and test is performed with software and simulation engines; the actual system hardware is not used. To maximize the benefits of this approach the simulations must be systematic —the engineering team must build tests for every requirement and then exercise the system design rigorously against those tests.

Three critical elements are needed for virtual integration and test:
. An executable system specification
. Requirements-based tests
. A test and simulation executive (or environment)

An executable system specification is a system model —or set of models— that can be executed, usually through simulation. It should include models of relevant subsystems (for example, control algorithms, mechanical and electrical components, hydraulics, and so on) that are detailed enough to capture their system-level functional behavior. The system model also should include operating environment models or plant models. These might include actuator and sensor models for control applications, or transmission environment models for communication systems. A complete executable system specification makes it possible to simulate the system and test it against the requirements before implementing any of the components.

Requirements-based tests are used to exercise the executable system specification and thereby validate the requirements and verify the design. The tests are formulated from functional requirements that can be modeled as simulation input signals or sequences and, ideally, expected outputs.