The validation of Autonomous Drive (AD) software needs sufficient testing to ensure the safety of autonomous vehicles. Scenario-based testing in a virtual environment provides a platform to detect AD/ ADAS software bugs and fix them early in the software development life cycle. Executing all the permutations of test cases in the parameter space is computationally expensive. It is time-consuming to discover the function performance boundary. The target-oriented testing in a scenario-based approach generates diverse critical scenarios by searching the whole parameter space in a virtual environment. This thesis follows the target-oriented testing approach, develops a novel algorithm, and provides a systematic and novel method for generating diverse critical scenarios by exploring the whole parameter space in minimal computation. From this, the function performance boundary can be observed, and corner cases can be identified. Also, the performance of our novel algorithm is demonstrated by comparing it with other general optimization approaches. 

This approach has a higher chance of generating critical scenarios than the random parameter selection method and at a reasonable computational cost. It improves the coverage of the test scenarios, verifies the software's stability and quality of the software,  and helps in finding the test cases where the chances of function failure are high. This approach helps to prioritize the test cases that need to be tested in Hardware in Loop (HIL) and cars on a test track. This thesis is carried out at Volvo Car Corporation (VCC) and targets to support the verification and validation of AD/ADAS software. The experiment is performed in a Software In Loop (SIL) environment called CSPAS at VCC. The Autonomous Emergency Braking (AEB) function is considered a use case to develop the test framework. The same test framework approach will be appropriate for generating critical scenarios for the AD function.