With the development of mobile Ad Hoc ad hoc networks,it has gradually been applied to the military field,and the concept of UAV swarm system has emerged as the times require.The most concerned one is the scheduling problem of communication resources.UAV swarms have characteristics such as constantly changing topology,limited time slot resources,and specific requirements of communication resources by different nodes.Therefore,the centralized network structure cannot meet the requirements of the system well.The distributed network structure can optimize the reuse of limited time slot resources,which is beneficial to improve network throughput efficiency.When the topology changes,the distributed network structure can ensure that the unchanged areas are not affected,and the reasonable resource allocation effectively reduces interference conflicts between networks.Therefore,for the UAV cluster system,the distributed TDMA resource scheduling method is considered to be one of the main technical means to solve the problem.This paper mainly studies the resource scheduling problem in the UAV swarm system.According to the actual characteristics of the system,such as dynamic topology,limited communication resources,transmission conflicts,etc.,the distributed TDMA protocol is designed and improved to enable it to be solved the problem of resource scheduling in the system.Specific work and significance include:Firstly,for the scheduling time-based MAC protocol of fixed time slot,the network throughput and transmission efficiency are low in the distributed network,and the problem of time slot occupation cannot be adaptively implemented.A resource scheduling protocol based on Multi-Arms Bandit is proposed.The protocol utilizes a reinforcement learning model to continuously learn and iterate through state,action,feedback,and state behavior value functions,so that nodes can make time slot occupancy choices based on their own and neighboring node states without interaction.In the protocol,the node has the ability of decision-making learning,continuously gains experience by sensing the surrounding environment,and guides the current action behavior selection according to historical experience.Based on the TDMA time-division multiplexing mechanism,the node can occupy one or more time slots for communication transmission outside the interference range,and design the protocol through the rocker gambling machine model,thereby solving the problem of redundant time slots and enabling the time slot to be more fully utilized to improve network throughput efficiency.Secondly,for the UAV swarm system,in addition to the problem of redundant time slots,it is also necessary to consider the node energy and the traffic demand.The resource scheduling protocol based on the Multi-Arms Bandit is improved.When the node makes learning decisions,the energy and priority factors are added.According to the residual energy of each node and the topological density,the conflict nodes are compared,and the idea of baking algorithm is introduced to back off,so that nodes with low residual energy and high topological density can occupy time slots with fewer cycles to meet communication needs.Compared with other MAC protocols,it is proved that by classifying and comparing the remaining energy and the business,the time slot resources can be allocated more reasonably and the throughput efficiency of the network can be further improved.Finally,in order to further adapt to the problem of the change of the node time slot caused by the dynamic topology of the UAV swarm system and the interference conflict caused by the topology change,consider changing the time slot to variable length and continuously adjusting the time slot,according to the idea of desynchronization,we propose a distributed TDMA protocol.And by adding a timestamp mechanism to solve the stale information problem,and finally the node can dynamically occupy the time slot according to the demand and topology changes.When a node joins or leaves or a network meets and merges,it may cause a sudden occurrence of a transmission conflict because of time division multiplexing.Therefore,a punching and neighboring node monitoring algorithm is proposed to sense the conflict,thereby completing the solution conflict and enabling the node to adaptively changes the slot occupancy.The distributed TDMA protocol based on the desynchronization idea can make the nodes in the distributed network continuously adjust the slot length according to the state requirements,and can adapt to the situation that the node joins or leaves the network to ensure the system has high network throughput and convergence rate. |