Research On High Time-effective Robust Topology Control Algorithm

Robust topology control optimizes the network topology by adjusting the transmitting power of nodes and other means,which can ensure that the network is still connected and communication is smooth when the nodes fail due to energy depletion or interference attack.It is of great significance in the application of cooperative operation,disaster relief,emergency communication and so on.The existing research on topology control is mainly based on graph theory and probability analysis.The computational complexity of the algorithm increases exponentially with the increase of the number of network nodes and robustness.However,with the increasing scale of communication network and the high dynamic change of network environment,the existing topology control algorithms need a lot of time for complex calculation,resulting in the problem of inaccurate decision or decision lagging behind the change of network,which makes the network performance decline sharply with the increase of network scale.Therefore,improving the accuracy and timeliness of topology control algorithm is the key factor to ensure the network performance.In order to improve the timeliness of the topology control algorithm,this paper proposes a k-point connectivity topology control scheme suitable for distributed execution(that is,any k-1 node fails at the same time and the network can still remain connected).Each node is constructed based on the local topology detection results.Local k-point connectivity topology,and power control is performed through information exchange to make the entire network satisfy k-point connectivity.Specifically,each node uses the k-point connected topology control algorithm based on deep reinforcement learning to construct a local topology.This algorithm designs a reward function that considers both robustness and power consumption.If the robustness requirement is met,positive feedback is given considering the power consumption at the same time,otherwise,negative feedback is given as a penalty according to the difference between the actual connectivity and the target k,which guides the agent to find the relationship between the transmit power,network connectivity and power consumption.,so that the network uses less power consumption to satisfy the network k-point connectivity.The simulation results show that: compared with the traditional topology control algorithm,the algorithm can ensure that the decision-making time is always maintained at the millisecond level,and at the same time,it uses lower energy consumption to ensure the connectivity of k-points in the network.However,the above methods have high requirements on the computing power of nodes,and the training time required to obtain the optimal topology control scheme is long,which is difficult to implement in nodes with limited computing power.Aiming at the implementation of the intelligent robust topology control scheme proposed in this paper in the actual system,we further propose a knowledge transfer method for robust topology control,which enables nodes with limited computing power and volume to quickly obtain knowledge and make decisions through knowledge transfer.Therefore,we propose a network knowledge representation method,which represents data-driven tacit knowledge through deep neural network and process knowledge through effective paths between entities,so as to realize the structured representation of network knowledge.Furthermore,through the effective reasoning and transmission of network knowledge,the efficient migration of network knowledge is realized,so that nodes with limited computing power can make topology control decisions quickly.Through the example test,it is verified that the knowledge representation can store the topology control knowledge and guide the construction of k-point connected topology.Knowledge transfer can enable agents without knowledge to quickly obtain topology control knowledge suitable for their environment,and the decision-making time is reduced by 52.4%.