基于PerDQN的时间敏感网络保护带调度算法OA

When the data frame is excessive long,the effective bandwidth utilization of the network will reduce due to redundant guard band.In view of this,a PerDQN-TAS method based on deep reinforcement learning PerDQN is proposed.The classical priority constrained knapsack problem(PCKP)is transformed into Markov decision processe(MDP).By introducing two new reward values(weighted sum of packet length and packet priority),the PerDQN-TAS model is trained to learn the optimal behavior to be taken in different states,so as to improve the cumulative priority while making full use of the available bandwidth in the guard band.Experimental results show that in comparison with the existing TAS(time-aware shaper)guard band algorithm,the bandwidth utilization of PerDQN-TAS consistently remains above 91%,which gradually increases with the initial remaining capacity of the guard band and ultimately reaches 98.1%.Simultaneously,the cumulative priority continues to rise and reaches 8.98 eventually.To sum up,the proposed algorithm improves the bandwidth utilization significantly.