Q-learning Based Intelligent Ant Colony Scheduling Algorithm in Heterogeneous System

In view of the variable task structure and complex scheduling environment of heterogeneous computing system, and the low efficiency of existing scheduling algorithms, this paper abstracts the system by establishing a directed acyclic graph and a target system model, and proposes a Q-learning based intelligent ant colony scheduling algorithm. The algorithm adapts the scene to the scheduling environment, dynamically calculates the Q matrix according to the reward function, and the Q matrix is used as the initial pheromone of the ant colony algorithm. According to the pseudo-random proportional rule to select the processor, and the scheduling list is formed by the constraint relationship between the components to complete the task allocation. Analysis by randomly generating task graphs shows that this algorithm is more suitable for heterogeneous computing systems and computationally intensive tasks. Compared with ACO、 QMTS and GA algorithms, this algorithm reduces the scheduling lengths by average of 35.22%、 27.41% and 20.41% respectively, can obtain better scheduling results.