Location Allocation Of Canteens For Taxi Drivers Based On Trajectory Big Data

With the implementation of the restrictive policy in the first-tier cities such as“Beijing,Shanghai,Guangzhou”,vehicles restriction,expensive parking,and limited parking spaces make many people choose the latter between buying private cars and choosing taxis.The demand for taxis is the strongest in the morning and evening commuter time,which is also the dining time of taxi drivers.Taxi drivers are often unable to eat in time because they are busy with taking passengers.Gastrointestinal diseases caused by irregular eating habits have become occupational diseases for taxi drivers.In view of the irregular diet of taxi drivers,most of the existing researches provide solutions from the driver's own point of view,such as the habit of eating on time.However,part of the reason for this problem is that there is little dining place for taxi drivers to stop eating,but there is no research to propose best location for creating new dining places.Today,a large number of taxis carry GPS receivers,and the generated massive trajectory data not only provides accurate vehicle ID,but also state information such as vehicle status,vehicle GPS position,vehicle GPS time,etc.These data have good time continuity and carry large amount of information.Massive trajectory data has a large amount of information.Taxi trajectory big data contains the taxi drivers'demand and provides quantitative basis for creating new canteen to solve the problem of taxi drivers'dining.However,there are two technical difficulties in using trajectory big data for site selection:First,it is necessary to quickly extract position and state information from massive trajectory data,and the index structure of traditional spatial data?such as R-tree?cannot meet the requirements;Second,the location problem is a high complexity optimization problem,and the speed and accuracy of the existing algorithms still need to be improved.In summary,this study proposes a scheme for locating canteen using trajectory big data,and proposes a solution to the existing technical difficulties.First of all,this paper proposes an indexing method for vehicle trajectory big data.This index queries a specific time period,vehicle position and status at different time granularities,Also,this structure could quickly store trajectory big data.Then,this article counts the time series of dining demand.This paper verifies predictability by hypothesis testing and analyzing prediction deviation.Finally,this paper proposes a sub-population collaboration and particle reinitiate strategy.This strategy enhances the global search capabilities of particle swarm optimization.This algorithm avoid convergence to local best,and thus obtains canteen location with better fitness values.In order to verify the effectiveness of the scheme,this study proposed a location scheme based on the taxi trajectory data of about 13,000 taxis covering the whole of Shanghai in August 2016?31 days?.The conclusions of the paper include two parts:one is the predictability of the demand time series within the inner ring and the outer ring of Shanghai;the other part is the site selection plan of the meal and the average time required for each taxi driver to go to the meal.The research results of this paper can provide a quantitative basis for the government departments to carry out the layout of dining facilities and parking space planning,thus solving the problem of drivers'difficulty in dining.The innovations of this article include:?1?Based on the real dining demand data,this paper proposes an optimal scheme of canteen's location quantitatively to solve taxi drivers'difficulty to have meals.?2?This study designs an indexing scheme concerning data characteristics?sorted by time,low frequency of deletion?and application targets?vehicle location and status query at specific time?.The indexing method resamples the original data by time granularity.The creation time,space,and query time are 68%,62%,and 2%of the B+index of the MySQL database,respectively.?3?The traditional particle swarm optimization algorithm is improved and applied to the solution of location optimization problem.This paper proposes sub-population collaboration and particle restart strategy.This strategy enhances the global search capabilities of particle swarm optimization,avoids premature convergence to local best,guarantees convergence speed,and obtains a better location solution with better fitness value.Experiments show that the improved algorithm fitness value is better than the existing method.After the improvement,the algorithm converges to the global optimal solution more times than the existing method^[61].