基于隐马尔可夫模型的车辆轨迹预测算法的研究

发布时间：2018-03-05 00:29

本文选题：轨迹预测　切入点：隐马尔科夫模型(HMM)　出处：《南京邮电大学》2017年硕士论文　论文类型：学位论文

【摘要】：在智能交通系统(ITS)、智能导航、移动电子商务、物流配送中,实时、精确、可靠的车辆轨迹预测具有极高的应用价值。车辆轨迹预测不仅可以提供精准的基于位置的服务,而且可以提前监测和预判交通状况,进而为用户推荐最佳路线。本文首先建立道路网络模型,挖掘车辆历史轨迹的双层隐藏状态,并根据历史轨迹确定隐马尔可夫模型(HMM)参数,然后采用Viterbi算法确定车辆已行驶轨迹对应的双层隐藏状态序列。最后提出一种基于双层隐藏状态的隐马尔可夫模型车辆轨迹预测算法(DHMTP),预测车辆将要行驶的完整轨迹和车辆行驶轨迹在未来k个阶段的近邻单元的位置信息。由于采集到的数据相对有限,当前的训练集数据无法包含所有的车辆行驶路径。在测试该模型预测车辆路径的准确性时,我们所用到的测试集包括训练集中不存在的路径或者坐标点。另外,在我们先前已经建立的用于车辆路径预测的隐马尔可夫模型中还存在一类“零概率”问题。出现“零概率”时有两种可能,一种是在训练集中某种坐标点出现的次数为零,另外一种是某些车辆行车路径或坐标点在训练集中不存在,但是在测试集中存在。因此,本文采用基于K-means++聚类算法扩充训练集,采用平滑技术解决“零概率”问题,以此优化隐马尔科夫模型,进而提出Possible_Track算法预测车辆轨迹路径。最后,在研究基于隐马尔可夫模型的车辆轨迹预测方法的基础上,我们提出了基于HMM的路径推荐系统框架。在短时间内,可以提醒处在拥堵十字路口的驾驶员车辆前行的安全性;在长时间内,预测可能会发生交通堵塞的区域,及时做出调度,引导交通并提醒驾驶员及时做出路线调整。
[Abstract]:In the intelligent transportation system (its), intelligent navigation, mobile electronic commerce, logistics distribution, real-time, accurate and reliable vehicle trajectory prediction has high application value. Vehicle trajectory prediction can not only provide accurate location-based services. Moreover, the traffic condition can be monitored and forecasted in advance, and then the best route can be recommended. Firstly, the road network model is established to mine the double hidden state of the historical track of the vehicle, and the HMMM parameters are determined according to the historical track. Finally, a hidden Markov model vehicle trajectory prediction algorithm based on double hidden state is proposed to predict the integrity of vehicle driving. Location information of the nearest neighbor unit in the next k stages of the trajectory and the vehicle trajectory. Due to the relatively limited data collected, The current training set data cannot contain all vehicle paths. In testing the accuracy of the model for predicting vehicle paths, the test set we use includes paths or coordinate points that do not exist in the training set. There is also a class of "zero probability" problem in the hidden Markov model that we have established previously for vehicle path prediction. There are two possibilities when "zero probability" occurs, one is that the number of occurrences of a certain coordinate point in the training set is zero. The other is that some vehicle paths or coordinate points do not exist in the training set, but exist in the test set. Therefore, this paper uses K-means clustering algorithm to expand the training set, and uses smoothing technology to solve the "zero probability" problem. In this way, the hidden Markov model is optimized, and then the Possible_Track algorithm is proposed to predict the vehicle trajectory. Finally, based on the research of the vehicle trajectory prediction method based on the hidden Markov model, We propose a path recommendation system framework based on HMM. In a short period of time, we can remind drivers at congested crossroads of the safety of moving ahead; in the long run, we can predict the possible traffic jam areas and make timely scheduling. Guide traffic and remind drivers to make route adjustments in time.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U495;O211.62

【相似文献】