Trusted Clustering Routing Protocol Based on Deep Reinforcement Learning
Addressing the security issues caused by malicious nodes acting as cluster heads in clustering routing protocols,as well as the challenges of slow convergence and substantial volatility encountered in deep reinforcement learning-based routing pro-tocols,a clustering routing protocol was proposed based on trust mechanism and deep reinforcement learning algorithm Soft Actor-Critic(SAC)was proposed.This protocol integrates a trust mechanism and leverages the advanced deep reinforcement learning algorithm,Soft Actor-Critic(SAC).The protocol employed an enhanced label propagation algorithm to efficiently cluster the net-work.Then,a trust-based cluster-head election mechanism was utilized to carefully elect trustworthy cluster heads from within the cluster,and a master-slave cluster-head mechanism was adopted,effectively safeguarding against cluster heads transforming into malicious nodes.At last,the SAC algorithm was leveraged to make dynamic routing decisions,with the elected cluster heads acting as agents.Experimental results demonstrate that the protocol has better performance and convergence than RTRPT,SCR-TBE,DQN,D3QN and PPO routing protocols.Its packet loss rate,average delay and network throughput are the best.In multiple test scenarios,the performance of the protocol was improved by 3.97％and 22.39％compared with the PPO scheme.
NETL
NSTL
万方数据
