Energy Efficient Real-Time DVS based on Genetic Algorithm

Abstract

This paper proposes a novel real-time dynamic voltage scheduling algorithm(GA-DVS) based on genetic algorithm for periodically real-time task set. Based on a mathematical system model in the real situation, the GA-DVS algorithm is different from classical DVS algorithms, some critical parts of which are specially designed, such as encoding, the fitness function, the crossover/mutation/repair operator and the termination condition; GA-DVS searches from multiple initial points, mutates during the search process and uses the repair operator to guarantee the convergence of the algorithm. GA-DVS can give optimal solution for the hard real-time task on CPUs with N adjustable frequencies and voltages in most cases. Finally, experimental results demonstrate the efficiency of the GA-DVS algorithm, which can achieve a good tradeoff between time cost and precision and search effectively in the solution space of the NP-complete problem.