Research On Intelligent Path Planning Technology Based On Deep Reinforcement Learning

In recent years,with the explosive growth of network data traffic and the increasing fluctuation frequency,the communication network has brought severe challenges.The main reason is that the current communication network is still running on the routing framework designed many years ago.Traditional dynamic routing protocols transmit traffic by calculating the shortest path.When the traffic fluctuates violently,some links may be selected by mltiple traffic transmission paths at the same time,which will cause network congestion,network throughput decline and transmission delay increase.This routing mechanism is unable to adjust the traffic transmission path dynamically according to the real-time state of the network,nor can it learn from the previous experience.Therefore,an intelligent network path planning scheme becomes very important.Accompanied by the rapid development of artificial intelligence technology in recent years,DRL,which combines the model representation ability of deep learning with the decision-making ability of reinforcement learning,has been applied to many fields and achieved great success.The software defined network which separates the data plane from the control plane also makes the network management more convenient,which provides a new idea for the intelligent path planning scheme.This paper studies the problem of network path plannning,and proposes an intelligent path planning scheme based on DRL algorithm.The main contents of this paper are as follows:1.Network congestion and traffic load balancing are studied first,and a networ-k path planning scheme based on DRL algorithm is proposed.In order to improve the network performance,the modeling of network path planning problem is carried out,and on this basis,the input and output design of algorithm model and optimization function design are completed.2.In order to verify the effectiveness of the path planning algorithm based on DRL,in the simulation network enviroment,the average network delay and the maximum link bandwidth utilization are taken as the objective function to train the algorithm model under different load intensities,and the convergence is observed,which shows that the algorithm is feasible and effective.Then the algorithm is compared with the comparison algorithm,and the average network delay and the maximum link bandwidth utilization can be significantly reduced.3.An intelligent path planning system is designed based on DRL and SDN architecture.Detailed description of the overall system architecture design and the design of each sub module,and the system can train the algorithm model based on DRL on the physical network,and complete the network path planning task on the basis of the model.4.For comparing the performance,the intelligent path planning system and the comparison algorithm are run on the physical network,and the network throughput and network bandwidth loss rate are taken as indicators under different load intensity.The experimental results show that the algorithm proposed in this paper can improve the network throughput and network performance.Test with data traffic of different distribution,and the experimental results show that the algorithm has certain generalization ability.