Research On Resource Allocation Technology For Reinforcement Learning In The Internet Of Things

Machine Type Communication(MTC)is one of the main businesses of cellular networks.An important challenge of Machine-to-Machine(M2M)is that it happens suddenly and when a large number of machine-type communication devices(Machineto-Machine,M2M)Type Communication Devices(MTCDs)requests to access the wireless access network,which will cause the wireless access network to be overloaded.In order to solve the problem of data redundancy and minimize job scheduling time,the optimization of network resource allocation needs to find a more complete solution.System resource management problems usually manifest as difficult online decisionmaking tasks.Some existing basic scheduling algorithms,such as First Come First Service(FCFS),High priority first(HPF),etc.,cannot completely solve the multiresource scheduling problem.The best solution is to design an efficient heuristic algorithm under certain assumed performance guarantees.Based on this,this article conducts research on resource allocation technology in the Internet of Things,specifically:Research and simulation of existing stand-alone resource scheduling algorithms.This paper conducts simulation experiments on several existing resource scheduling algorithms,and the results show that the existing algorithms cannot solve the resource management problem in M2 M communication.Due to the development of communication technology,machine communication has gradually shifted from singlemachine scheduling to multi-machine scheduling,which cannot be done by existing scheduling algorithms.The solution of multi-machine system resource scheduling.In response to the recent development and application of deep reinforcement learning for artificial intelligence problems,a resource system that can be directly learned and managed from experience is simulated and constructed.In this article,research and design a new type of job scheduling algorithm-Deep Reinforcement Q-Learning(Deep RQ)algorithm.The problems in M2 M communication technology will be solved by Deep RQ algorithm.The research results show that the performance of the deep reinforcement learning algorithm is equivalent to that of the most advanced heuristic algorithm,and it adapts to different conditions,so that the convergence of the learning curvature is optimized and improved.After solving the resource management problem and under the condition of good neural network training,further research has been done on the performance of the entire communication system,adjusted and constantly tried the Tensor Flow optimizer in a suitable network environment,made simulations and carried out horizontal Compared with the vertical,the purpose is to reduce the total running time of the system and the overhead of system memory,and to find a more complete solution for the optimization of network resource allocation.