Research On Path Planning In Discrete Domain Based On Stream Data

The path planning in the discrete domain is a process of locating the starting point and the target point in the complex topological network of paths and then using the path search method to carry out the path optimization planning.As the nodes of topological network tend to grow exponentially,the traditional path optimization algorithm based on graphics has gradually shown fatigue,and it is difficult to adapt to the complex and changeable path topological network.Therefore,it is indispensable to study the direction of path optimization.In this thesis,network topology construction is taken as the entry point to obtain the real-time state of network topology path and dynamically calculate the weight information,so as to complete the whole path optimization process.Specifically,the main research directions of this thesis are as follows:(1)Study the construction method of network topology structure.In this thesis,the network topology characteristics are analyzed,the relationship between nodes is defined according to the graphics method,the hierarchical architecture is introduced to simplify the data storage,and the chain storage structure is used to simplify the node structure.According to the definition of nodes,paths and connectivity strength,the construction process of network topology architecture is realized.(2)Study the state discrimination algorithm based on stream data.Streaming data may contain more extraneous or redundant characteristics than static data.Based on the analysis of flow characteristics,a state discrimination algorithm based on multi-model fusion is proposed in this thesis.Then,the algorithm will automatically select the features with high correlation with state judgment and delete the features with low correlation or even redundant features.Then,multi-feature clustering will be carried out based on the flow feature attributes.The flow feature attributes will be classified according to the similarity principle,and the flow data instances will be divided into several sub-instances with obvious differences.Finally,a real-time classification model is established to make a quick response to the data flowing into the model and complete the classification process of the convective data in a limited time,so as to avoid large-scale data queuing and blocking caused by excessive computational complexity.(3)Study the traditional path planning algorithm,summarize the defects and deficiencies of the traditional method,according to the defects and deficiencies,optimize the spatial structure of the traditional path planning algorithm,reduce the algorithm complexity,propose a linked-dijkstra method,so that it can be applied to the complex network environment.(4)Combined with expressway traffic flow data,a highway network topology structure based on expressway was constructed to obtain the real-time traffic status within the road network and complete the vehicle path guidance process.Experiments on real network topologies show that the proposed method will be applied to complex network topologies.The network topology structure with chain storage is superior to other structures in adaptive performance,and the state discrimination based on multi-model fusion can effectively determine the real-time path state.The linked-dijkstra algorithm is obviously due to the traditional method in complexity.