TUNING THE COST FUNCTION IN A GENETIC/HEURISTIC APPROACH TO THE HARD REAL-TIME MULTITASK-MULTIPROCESSOR ASSIGNMENT PROBLEM

摘要

The problem of scheduling a set of independent preemptible real-time tasks in a multiprocessor system belongs to a general class of NP-complete. It has been attacked with heuristic methods and genetic algorithms separately. In this paper, a hybrid genetic/heuristic method is proposed to solve the assignment problem. We show comparisons with Tindell's method and Rammarithan's method as benchmarks to evaluate the performance of our method.