Time Weighted Schedular

Sammanfattning: The In-Car Digital (ICD) department is developing Infotainment solutions for Volvo Cars R&D. The current, award-winning, Infotainment solution is a complex setup with multiple ECUs (Electronic Control Units). The ICD section is utilizing HIL (Hardware in the loop) environments for Infotainment domain-level testing. ICD units currently have a limited number of HIL environments and are requesting a method to divide the number of selected executable tests into a time-limited schedular.  This thesis aims to divide the number of selected executable tests into a time-weighted schedular by evaluating and comparing the run time of three different scheduling algorithms for optimizing the execution run time of the total amount of tests.  The first algorithm used is the Shortest Job First. Here two methods of the Shortest Job First algorithm are used namely Pre-emptive and Non-Preemptive. The test cases with the shortest execution time are allotted a HIL first as they arrive, whereas test cases with longer execution times are placed last and given the lowest priority.  The second algorithm used is the Genetic algorithm. A random population is generated by randomly assigning test cases to HILs. Then each solution is evaluated for its fitness. A new population is created by selecting two parent individuals from the population according to their fitness. The two best individual solutions are picked and chosen for crossover. After the crossover, the new offspring is placed in the new population. Finally, the mutation is performed. These steps are repeated until an optimum solution is found. The third algorithm used is the Wrangler algorithm. Wrangler predicts test cases causing stranglers and avoids these test cases. If the test case’s confidence measure exceeds the threshold, the test case is delayed. If not, a default scheduling policy is used if the prediction does not exceed the threshold. These performances of the algorithms were analyzed based on their runtime. Each algorithm was evaluated using 5000 test cases generated randomly with different run times. The first set of data was generated by grouping the test results based on the number of available HILs concerning their run time. The Genetic algorithm was found to stand out when the number of available HILs was low. The Shortest Job algorithm performed better with a higher number of available HILs. The second set of data was generated by grouping the test results based on the test case distribution concerning its run time. This did not show any trend that any of the algorithms perform better with a higher proportion of large test cases or a large proportion of small test cases. The Genetic algorithm can be the first choice with a limited number of HILs. As more HILs are procured, they can be replaced by Shortest Job First algorithm.