Research Of Dynamic Route Guidance System Based On Travel Time Combination Prediction Model

Recently years, the quantity of cars are increasing fast,urban congestion problems becomemore and more serious,traditional traffic management means which include driving restrictionand traffic assignment have been difficult to deal with the congestion problem,dynamic routeguidance system of intelligent traffic system(ITS) can play an important role in routes trafficassignment,vehicle guidance and optimal route selection,that is the reason why a number ofcountrys focus to research and apply this technology. Dynamic route guidance systemcompute the optimal route according to the traffic change information.The least travel timeexpense path is the most widely used,so the accurate prediction of travel time for vehiclepassing links is key to the system that has a good guidance ability.Aim at improving theaccuracy of travel time prediction and completing the least travel time expense path algorithmbased on travel time prediction this paper has do some research as the following content:Firstly,utilizing the classical travel time prediction method which include support vectormachine,artificial neural network, multiple linear regression, K nearest neighbors to predicttravel time separately on the obtained data set,comparing the advantage and disadvantage ofeach methord,and then raising the nearest neighbor median model according to the characterof travel time data.Second,against the weakness of single predction model,researching the combinationmethod to combing multiple model for elevating the accuracy of prediction.The method forcombination include linear type such as linear regression,nonlinear type such as neuralnetwork and multiple dimensions,least square linear regression is the most suitable fordynamic route guidance system that need real time processing, experiment has prove thatcombination model prediction effect superior than single model, linear regressioncombination model better than neural network.Raising two method to improve the leastsquare linear regression, cluster binned linear regression and dynamic adjusting weight linearregression,they make a more accurate prediction effect.Third, existing researchs for dynamic optimal route selection algorithm mostly utilize thesimple stratege that compute the optimal route again when meeting a new intersection, it notonly cann’t obtain the optimal route,but also increase the computation.This paper study a leasttravel time consumption dynamic route selection algorithm,which combine the travel timeseries prediction information,based on the dijkstra algorithm and K shortest pathalgorithm,select an optimal route that take the dynamic change of the link travel time intoaccount.