In most circumstances, there is a need for a bigger amount of computational power, which calls for the implementation of innovative and cutting-edge allocation strategies. It is imperative that we do not ignore the problem of coming up with efficient compression techniques that can be implemented on front haul lines. It is vitally necessary to monitor and evaluate the influence that latency has on the performance of the upper layers of the fronthaul. This is an imperative necessity. In addition, optimal resource allocation within the setting of constrained fronthaul should be the subject of further research. The loss of packets that can take place on fronthaul networks is another potentially fascinating topic that could be brought up in conversation. The fact that the fronthaul network is so diversified shouldn't come as much of a shock to anyone. It has a delay as well as a variety of connection capabilities, and because of both of these factors, the fronthaul needs to be re-configured in order to respond appropriately to the traffic load and the topology of the network. In this paper, we use Federated Learning (FL) model to optimize the resource allocation between device to Device (D2D) in 5G network. The model is designed to optimize the resources for all the users in the network , especially in the monitoring of Agricultural farms. The results of simulation shows that the proposed method has reduced delay, throughput and reduced packet loss than the other methods.