Research On Personalized Path Planning And Design

The purpose of personalized path planning is to provide a more suitable navigation route for vehicle navigation users.Due to the repeatability of individual travel routes,it is feasible to mine users’ driving rules from vehicle navigation data and plan more personalized routes.This paper uses big data technology to mine vehicular navigation data,studies and implements each function module of personalized path planning,and establishes a reasonable evaluation system according to the characteristics of navigation services,and evaluates personalized strategies by using navigation quality indicators.This paper mainly involves the following modules: map matching,trajectory clustering,trajectory compression,trajectory index,dynamic route recall,index system.The main research contents include:(1)Aiming at the problem of low efficiency of map matching due to the instability of navigation track points,this paper proposes a clustering based adaptive filtering threshold and geometric midpoint mining method to filter outlier abnormal track points,clean and resamble the original track points.On this basis,in order to solve the traditional map Matching mining caused by a single data category Matching the problem of low efficiency,this paper proposes a local ST-Matching + CH map Matching scheme,will process after the trajectory data and other navigation data are combined,form a complete navigation records,it takes only a partial match the users on the network can generate real route.Compared with the ordinary HMM +(A-star)algorithm,the performance time of this method is reduced by 13% and the accuracy is improved by 3.6%.(2)To solve the problem of poor track clustering effect,a clustering method based on word2 vec model is implemented in this paper.The trajectory is vectorized by the trained embedding matrix,and then density clustering is used for the track vector clustering operation.In addition,a similarity computing method based on dynamic programming is implemented to cluster the matched trajectories,which transforms the path similarity problem into character matching problem and greatly reduces the computational complexity.On this basis,the two experiments were carried out to verify their effectiveness.(3)aiming at the problem of low efficiency of path query,this paper presents a static weights based on road feature and high speed stable compression strategy to get to compression of the navigation route after clustering,the method effectively without change the character encoding compression and restore,comparing the compression storage solution,save storage space by an average of 60%.In this paper,an inverted index structure based on B+ tree is proposed to build the index on the clustered and compressed trajectories.Since the B+ tree index is established in advance according to the matching generation sequence of the track,the leaf node stores the features of each road section,and the efficient track retrieval can be realized only by inputting the road section information.Compared with the ordinary B+ tree index,the query efficiency of this index is increased by 35% on average.(4)In view of the personalized route recall problem of real-time road conditions,this paper proposes a recursive original strategy with dynamic weight to recall the compressed trajectory obtained according to the trajectory inversion index.This strategy can effectively improve the hit ratio of inconsistent routes and recall personalized routes that are more consistent with current road conditions.(5)In view of the evaluation problem of personalized route,this paper realizes a navigation evaluation system based on the establishment of data warehouse,establishes a complete index mining process,and evaluates the personalized path planning strategy.Through AB experiment comparison,it is verified that the strategy improves the quality indexes of all kinds of routes.Therefore,the main contribution of this paper lies in fully mining personalized behavior in navigation data,realizing a complete research process from data to personalized route planning,providing users with a better route scheme,which has certain reference significance in the field of personalized navigation route planning.