CN115798239A - Vehicle operation road area type identification method - Google Patents
Vehicle operation road area type identification method Download PDFInfo
- Publication number
- CN115798239A CN115798239A CN202211439582.4A CN202211439582A CN115798239A CN 115798239 A CN115798239 A CN 115798239A CN 202211439582 A CN202211439582 A CN 202211439582A CN 115798239 A CN115798239 A CN 115798239A
- Authority
- CN
- China
- Prior art keywords
- area
- torque
- short
- speed
- information
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Landscapes
- Navigation (AREA)
Abstract
The invention discloses a vehicle operation road area type identification method, which comprises the following steps: the method comprises the following steps: dividing a mountain area type, an urban area type, a suburban area type and a high-speed area type in a target area; and the range of the area type is embodied as discrete longitude and latitude point information; step two: judging the area type of each short journey according to the vehicle GPS positioning information and the area type range information; step three: adding labels of the types of the areas to which the vehicles run in each short-stroke characteristic parameter database; and meanwhile, a machine learning database is obtained, a machine learning algorithm is utilized for training, and a vehicle operation road region classification model based on machine learning is constructed and is used for identifying a vehicle operation road region. The method and the device realize acquisition of the region type label of the user data before synthesis of the driving conditions, can be used for constructing the driving conditions of different city partition types, and can calibrate technical parameters timely and accurately to evaluate the performance of the whole vehicle.
Description
Technical Field
The invention relates to the technical field of vehicle operation road area type identification, in particular to a vehicle operation road area type identification method.
Background
The global energy shortage and environmental pollution problem become more serious day by day, and the energy consumption and emission problem of the automobile, which is one of the necessary tools for people going out, have a great influence on the overall situation, so that in order to meet the increasingly strict emission regulation requirements of each country and comply with the strategic requirements of environment-friendly type and sustainable development, each automobile manufacturer pays attention to the improvement of fuel economy of the fuel automobile and the energy consumption of the electric car.
The driving condition is a curve of speed and time describing the driving characteristics of a vehicle, reflects the most representative driving characteristics of the vehicle of a certain vehicle type in a certain area or a certain road type, is mainly used for determining the pollutant emission amount and the fuel consumption amount of the vehicle, is an important basis of vehicle energy consumption and emission detection tests, can also be used for analyzing the damage and the service life of a driving motor of the vehicle, and provides a reference basis for the technical development, evaluation and the like of a new vehicle type. At present, most of automobile testing conditions used in the industry are international or domestic standard conditions, and in fact, social, economic and geographical characteristics of various cities are different and even have great difference. Thus, typical standard conditions are not sufficient to represent actual road driving conditions in different regions or cities. Moreover, road traffic conditions in different geographical areas of the same city are greatly different, simple accumulation is not beneficial to obtaining typical working conditions with strong adaptability, and the driving characteristics of a specific research object cannot be reflected more accurately. In order to make the constructed driving conditions more detailed and representative, the driving conditions of different urban subarea types are constructed according to the geographical position information, so that a novel solution is provided.
The driving conditions of the different region types are the basis of the parameter matching and control strategy optimization of the power assembly of the vehicle research and development design, because the driving conditions of the different region types have different characteristics, and the optimal control parameters and the optimal control modes of the suitable vehicle are also different.
In order to construct a driving condition of a typical city regional type, and to adjust vehicle control parameters in real time according to an actual driving condition type of a vehicle, and calibrate technical parameters to evaluate vehicle performance timely and accurately, a set of method is required to identify the regional type of vehicle operation.
Disclosure of Invention
The invention aims to provide a vehicle operation road area type identification method which can be used for early preparation of typical working condition development and can also be used for real-time working condition identification, so that the vehicle operation condition is improved, and data information is provided for vehicle operation cloud service big data analysis.
In order to achieve the purpose, the technical scheme provided by the invention is as follows:
a vehicle operation road area type identification method is characterized by comprising the following steps:
the method comprises the following steps: dividing the types of mountain areas in the target area according to the geographical position information; dividing a target area into a city area type, a suburb area type and a high-speed area type according to the road type; and the range of the area type is expressed as discrete longitude and latitude point information;
step two: the method comprises the steps of taking user data as input, wherein the user data comprises vehicle running data information and vehicle GPS positioning information, preprocessing the data, dividing short strokes according to a short stroke method, dividing original data into a plurality of kinematic segments, and calculating characteristic parameters of each short stroke to obtain a short stroke characteristic parameter database; judging the area type of each short journey according to the vehicle GPS positioning information and the area type range information;
step three: adding labels of the types of the areas to which the vehicles run in each short-stroke characteristic parameter database; and meanwhile, a machine learning database is obtained, a machine learning algorithm is utilized for training, and a vehicle operation road region classification model based on machine learning is constructed and is used for identifying a vehicle operation road region.
Optionally, the first step specifically includes:
determining a boundary between an urban area and a suburban area according to the urban administrative division range; dividing the boundary of the city and the suburb into a city area, and acquiring GPS (global positioning system) dot information of the boundary of the city area;
dividing all highways in the urban administrative area into high-speed areas, and acquiring GPS (global positioning system) dot-shaped information of the high-speed areas;
calculating the relief degree according to DEM elevation data, performing neighborhood analysis, reclassifying the relief degree to obtain a mountain area range, and obtaining GPS (global positioning system) dot information of a boundary of the mountain area;
the areas except the urban area, the high-speed area and the mountain area are divided into suburban areas, and suburban area boundary GPS point information is obtained.
Optionally, the boundary between the urban area and the suburban area is a high-speed surrounding or city surrounding line.
Optionally, the second step specifically includes:
classifying the short-stroke data of the user according to the region type; preprocessing original data including vehicle running data information and vehicle GPS positioning information, dividing short strokes according to a short stroke method, dividing the original data into a plurality of kinematic segments to obtain a user short stroke database, and calculating characteristic parameters of each short stroke for subsequent use;
and constructing a function for identifying the type of the short-trip region, wherein the function takes the GPS positioning information of the short trip and the GPS information of the four region types as input, and takes the tag of the short-trip region type as output.
Optionally, the third step specifically includes:
the method comprises the steps of constructing a vehicle operation road region classification model by using a random forest algorithm of machine learning, obtaining a machine learning database which contains user short stroke data, characteristic parameters and labels, training the database by using the random forest algorithm, obtaining the random forest model, reasonably adjusting the proportion of a training set and a testing set of data in the machine learning database, carrying out model parameter adjustment according to an implementation effect, and finally constructing the vehicle operation road region classification model based on machine learning.
Optionally, the characteristic parameters include:
running time, running distance, running speed, acceleration time, deceleration time, uniform speed time, idle time, acceleration proportion, deceleration proportion, uniform speed proportion, idle proportion, maximum speed, average speed, speed standard deviation, maximum acceleration, average acceleration of an acceleration section, standard deviation of positive acceleration, maximum deceleration, average deceleration of a deceleration section, standard deviation of negative deceleration, average acceleration, standard deviation of acceleration, relative positive acceleration, proportion of different speed intervals, acceleration number, deceleration number, constant speed number, parking number, standard deviation of torque, average positive torque, average negative torque, maximum positive torque, maximum negative torque, average torque of an idle section, average torque of a running section, average positive torque of a running section, average negative torque of a running section, time proportion of positive torque, time proportion of negative torque, maximum fluctuation amount when torque is increased, maximum fluctuation amount when torque is decreased, standard deviation of torque fluctuation amount when torque is increased, average fluctuation amount when torque is decreased, time proportion of high-torque interval, time proportion of torque interval time of low torque, absolute maximum yaw angle speed, absolute braking frequency, energy consumption number of braking times, total damage of a motor damage per unit, damage of a motor, damage per unit of a motor, damage per unit of a motor, and damage per unit of a motor damage.
Optionally, the constructing a vehicle operation road region classification model by using a random forest algorithm includes:
importing a machine learning database; randomly dividing sample data into a training set and a test set, wherein the proportion of the training set to the test set is 7:3, solving the importance of the characteristic parameters and sequencing;
the main parameters for designing the classifier are as follows: setting the number of decision trees in the random forest to be 178, setting the maximum tree depth to be 9, setting the maximum leaf node number to be 41, setting the maximum feature number to be 31, setting the minimum sample number of the leaf nodes to be 3, and setting the minimum sample number required for splitting to be 5;
and (4) performing model parameter adjustment to ensure that the accuracy is as high as possible, and finally constructing a vehicle operation road region classification model based on a machine learning algorithm.
The invention has the beneficial effects that:
according to the geographical position information and the road information, the driving conditions of different city subarea types are constructed, so that the constructed driving conditions are more detailed and representative. The operation area type label of the short stroke data is an important condition for determining the running condition of the vehicle and is also a main basis for synthesizing the later characteristic working condition. Based on a short-travel database containing the zone type labels to which vehicles belong, when the running conditions are synthesized, the running conditions of the sub-zones of the typical city can be selectively and respectively synthesized according to the zone types, and the four zones can be selectively merged and spliced according to the mileage length proportion of the four zone types to construct the total running conditions of the typical city. Therefore, an operating area type tag for acquiring user short-trip data before the synthesis of the driving conditions is necessary. When the vehicle runs, the vehicle running condition is identified in real time by using the vehicle running road region classification model, so that the real-time adjustment of the vehicle control parameters is facilitated, and the technical parameters are calibrated timely and accurately to evaluate the vehicle performance. Therefore, the vehicle operation road area type identification method can be used for early preparation of typical working condition development, can also be used for real-time working condition identification, improves the vehicle operation condition and provides data information for vehicle operation cloud service big data analysis.
Drawings
FIG. 1 is a flow chart of a method for identifying a type of a vehicle driving road area according to the present invention;
fig. 2 is a schematic diagram of a road area classification GPS boundary of a typical city in an embodiment of the present invention.
Detailed Description
The present invention will now be described in further detail with reference to specific examples, which are intended to be illustrative, but not limiting, of the invention.
Heretofore, the test conditions of automobiles used in the industry are basically international standard conditions, and the social, economic and geographic characteristics of various cities are different or even greatly different, so that the typical standard conditions are not enough to represent the actual road running conditions of different regions or cities. Moreover, road traffic conditions in different geographical areas of the same city are greatly different, simple accumulation is not beneficial to obtaining typical working conditions with strong adaptability, and the driving characteristics of a specific research object cannot be reflected more accurately. Therefore, according to the geographical position information and the road information, the driving conditions of different city subarea types are constructed, and the constructed driving conditions can be more detailed and representative.
Firstly, the operation area type label of the short stroke data is an important condition for determining the vehicle running condition and is also a main basis for synthesizing the later characteristic working condition. Based on a short-travel database containing the zone type labels to which vehicles belong, when the running conditions are synthesized, the running conditions of the sub-zones of the typical city can be selectively and respectively synthesized according to the zone types, and the four zones can be selectively merged and spliced according to the mileage length proportion of the four zone types to construct the total running conditions of the typical city. Therefore, an operating area type tag for acquiring user short-trip data before the synthesis of the driving conditions is necessary.
Secondly, when the vehicle runs, the vehicle running condition is identified in real time by using the vehicle running road region classification model, so that the real-time adjustment of the vehicle control parameters is facilitated, and the technical parameters are calibrated timely and accurately to evaluate the vehicle performance.
Therefore, the vehicle operation road area type identification method can be used for early preparation of typical working condition development, can also be used for real-time working condition identification, improves the vehicle operation condition and provides data information for vehicle operation cloud service big data analysis.
When the vehicle driving condition is constructed, the original data is obtained by adopting an autonomous driving method, but the road traffic conditions in different geographic areas of a typical city are very different, the simple accumulation is not beneficial to obtaining the typical condition with strong adaptability, and the driving characteristics of a specific research object cannot be more accurately reflected. Then, according to the geographical region composition characteristics of the typical city, dividing the typical city into a plurality of regions including urban areas, mountain areas, suburban areas and high speed by using the arcgis software; different from other existing working condition development ideas, the areas are divided according to geographic information and road information, some existing working condition development ideas are named as "urban areas" and "suburban areas" which are types of working conditions and are obtained by clustering short strokes of vehicles which independently run in the whole typical city, for example, in some existing working condition development ideas, road working conditions with certain geographic positions in suburban areas can also be divided into a category of "urban areas", and the category of "urban areas" actually refers to low-speed working conditions with low vehicle speed and frequent start and stop). The driving conditions of the vehicles in the regions divided by the invention have obvious difference characteristics, for example, the low-speed condition in the suburban region category is obviously higher than the low-speed condition in the urban region category. Therefore, when typical city driving condition development is carried out, the driving condition development is carried out after typical city land edge information is specifically divided into a plurality of types, and the refinement of the driving condition is beneficial to more refinement and representativeness of the classification of the working condition.
The invention divides the short-stroke data into different geographical region types, and based on the short-stroke characteristic parameter database containing the region type label of the vehicle operation, when synthesizing the driving condition, (1) the driving condition of the typical city subareas can be selected to be respectively synthesized according to the region types, (2) the four regions can be selected to be merged and spliced according to the mileage length proportion of the four region types, so as to construct the total driving condition of the typical city. This is the work done in step 1, step 2.
And 3, constructing a vehicle running road region classification model by using a machine learning random forest algorithm. The training set database for machine learning is: kinematic feature parameters for short trips of the user + region type tags. The machine learning model carries out learning training according to the correlation between the kinematic characteristic parameters of the short journey of the user and the region type, and the region type can be judged by inputting the kinematic characteristic parameters of the user after training. Therefore, the running condition of the vehicle can be identified in real time, the control parameters of the whole vehicle can be adjusted in real time conveniently, and the technical parameters can be calibrated timely and accurately to evaluate the performance of the whole vehicle.
Selecting and innovating characteristic parameters of the machine learning database: absolute maximum value of yaw angular velocity, velocity and mileage information of adjacent short strokes, absolute maximum value of lateral acceleration, braking times and hundred kilometers of energy consumption.
The invention provides a vehicle running road area type identification method, which mainly comprises the following steps:
step 1, creating a typical road classification GPS boundary database. And importing national map information data into a GIS platform Arcgis, wherein the national map information data comprises provincial administrative districts, county administrative districts, national roads, DEM elevation data and the like. Because each city is divided into suburban areas by high-speed or loop circuits around the city, the high-speed or loop circuits around the city of a typical city are used as the boundary between the city and the suburban area, and then the suburban areas are intersected with the national road bank to obtain the urban road and the suburban road bank; the highway library is a high-speed part collection of roads in provincial regions, including suburban high speed, high speed around cities and the like; the mountain road uses DEM data to obtain the mountain area in typical city in ARCGIS software, so as to obtain the mountain road library; the ordinary road administration grade division uses G, S, X, Y, C, Z to distinguish national road, provincial road, county road, rural road, village road and special road, and mainly covers suburb except high speed, other part roads of mountain area.
Specifically, step 1.1, whether the suburban boundary of each typical city is a city-surrounding high-speed or city-surrounding line is investigated, taking a metropolis as an example, the metropolis takes four loops as the boundary line of the city and the suburban area, data of national provincial administrative districts, county administrative districts and national roads are imported into Arcgis, and other provinces except the province of the typical city are deleted to obtain vector data of the province of the typical city. Intersecting the provincial and prefectural administrative district data of the typical city with the county-level administrative district data of the country to obtain the provincial and prefectural administrative district of the typical city consisting of the county-level administrative districts. And deleting redundant counties according to the existing standard administrative divisions, thereby obtaining a typical urban administrative district consisting of county-level administrative districts.
And carrying out surface line switching operation on the typical urban administrative district to obtain the peripheral contour line of the typical urban administrative district, dotting on the contour line at intervals of 100 meters, and obtaining the GPS point information of the peripheral contour line of the typical urban administrative district through calculation.
And step 1.2, on the basis of the step 1.1, intersecting the data of the roads in the typical city and the national city to obtain all the road information in the typical city. Copying all road information in a typical city to obtain a boundary between the city and the suburb, opening an attribute table of the road information, wherein the name attribute in the attribute table indicates the name of the road, screening out a four-ring line, and deleting other unselected roads to obtain the suburb boundary. And (5) observing whether the loop is notched, and if so, completing the loop into a complete coil. And (4) dotting points on suburb boundary lines at intervals of 100 meters, and obtaining GPS point information of the suburb boundary lines of typical cities through calculation. And on the basis of obtaining the boundary line, turning the line element to obtain a typical urban area, copying and cutting the typical urban area to obtain suburbs and mountain areas.
And 1.3, acquiring typical urban highway information, and performing intersection operation on data of typical suburban and mountain areas and national roads on the basis of the steps 1.1 and 1.2 to obtain typical suburban roads and change the suburban roads into expressways, so that subsequent operation is facilitated. Opening an attribute table of road information, screening 'fclass' = 'motorway' OR 'fclass' = 'motorway _ link' according to attributes to obtain highways (including roads around city highways, loop highways and the like), highway overpasses, ramps and other roads, switching, selecting and deleting the rest of the roads to obtain a highway library, dotting the highways at intervals of 100 meters, and obtaining GPS point information of the highways through calculation. And meanwhile, deleting the expressway in the urban and suburban road bank.
Step 1.4, obtaining the range of a mountain area, wherein DEM data realizes the digital simulation of the ground terrain through limited terrain elevation data, dividing the range of the mountain area according to national DEM data, firstly, obtaining a plurality of DEM data near a typical city according to longitude and latitude information of the typical city, importing the same map layer into Arcgis,
the whole integrated combined DEM data is obtained through grid embedding, so that the range of the elevation data of the mountainous area can be unified, and the unified processing is facilitated. The grid data structure is a simple and most intuitive spatial data structure, which is also called a network structure or an image element structure, and refers to a method for dividing an area into grid arrays with equal size and compact arrangement, wherein each grid is used as an image element, is defined by a row and column number and comprises a code representing the attribute type or information value of the grid.
The relief is then calculated. 1: the 250 ten thousand European international relief maps, the Chinese scale international legend guideline, the Chinese 1. The relief degree, also called relief degree, local relief degree and relative height, refers to the height difference between the highest point and the lowest point in a certain area, and is expressed as formula (1-1):
R＝H max -H min (1-1)；
wherein R is the relief of the topography in the analysis area, H max 、H min Maximum and minimum elevation values within the analysis region, respectively.
And carrying out rectangular neighborhood focus statistics on the whole DEM data, wherein neighborhood analysis is used for calculating the area around each pixel of the statistical data. And defining a square neighborhood with the height and width of 23 x 23 pixels, and performing focus statistics on the rectangular block to obtain the MAXIMUM MAXIMUM and the MINIMUM MINIMUM, wherein MAXIMUM is the MAXIMUM value in the calculation neighborhood. MINIMUM is the MINIMUM value to compute the pel within the neighborhood. A map algebra and grid calculator in a Spatial analysis tool is used for calculating a maximum to a minimum, and the output is a topographic relief map. And re-classifying the waviness by using a re-classification tool. The boundary stretching value of the mountain area is approximately checked by using the recognition function, the boundary line is divided into two levels up and down again, and a range diagram of the mountain area can be obtained. Since the "natural discontinuity" category is based on natural groupings inherent in the data. Classification intervals will be identified, the similarity values can be grouped most appropriately, and the differences between classes can be maximized.
Firstly, performing grid surface turning operation on the result of reclassification of the grid to obtain vector data, intersecting the vector data with the vector data of the suburban area range, deleting low data, and cutting out the mountain area range of a typical city.
The tile portions are processed. According to the obtained mountain area and the Chinese map, the east mountain area part is fluctuated due to the river, but the east mountain area part should belong to the mountain area part as a whole, so the east mountain area part and the east mountain area part are combined into a whole area by using a tool.
Starting an editor to select a map spot area in mountain area vector data, selecting a drawing tool → drawing synthesis → a gathering surface, inputting the map spot area into a mountain area, and newly building a file geographic database gdb in a folder during output to store the map spot area as the mountain area gathering surface; the rest part selects an automatic generation surface for creating elements in editing, and holes in the large block and surrounding small areas are connected into a whole block and are combined; and (4) cutting and copying the polymerized surface and the automatic generation surface. The obtained area is mountain area vector data, the obtained mountain area range surface is converted into a mountain area range line, mountain area range GPS points are obtained by dotting at intervals of 100 meters, the image blocks are named according to the blocks in the attribute table information, and the selected part and the excavated part are marked by 0 and 1.
Selecting an intersection negation tool, negating a mountain area in a suburb area to obtain a common highway area, carrying out surface-to-line operation on the common highway area, dotting points on the contour line of the common highway area at intervals of 100 meters, obtaining GPS point information of the peripheral contour line of the typical urban common highway area through calculation, naming the GPS point information in an information table, and marking a selection range or a removal range by using 0 and 1.
Since then, the GPS information base of four areas, namely urban area, mountain area, suburban area and high-speed area, has been built, as shown in fig. 2.
And 2, classifying the short-stroke data of the user according to the region type. The method comprises the steps of taking a large amount of user data as input original data, wherein the original data comprises vehicle running information such as GPS positioning information, speed, acceleration, soc, high-voltage electrifying signals, motor torque and rotating speed and the like. After the original data are preprocessed, according to the conventional early stage steps of constructing the running working condition, the user data are divided into a plurality of short-stroke running effective segments based on the high-voltage electrifying signals. And obtaining a user short-stroke database, and calculating the characteristic parameters of each short stroke for subsequent use.
And constructing a function for identifying the type of the short-distance belonging area, wherein the function takes the short-distance GPS positioning information and the GPS information of each area (including urban area, high-speed area, mountain area and suburban area) of the city as input, and takes the short-distance belonging area category label as output. The basic process is as follows:
since the short journey of the user is a segment of running segment which may span multiple area types, the proportion of the area type to which the short journey belongs needs to be judged according to the longitude and latitude information of the short journey, and the short journey is considered to belong to the area type when the longitude and latitude points of the short journey exceed a certain proportion and belong to a certain area type.
The short journey belonging to a typical city administrative district is screened out, the short journey belonging to the city district is firstly judged, because the short journey is located in the city district, the short journey does not belong to a high-speed area, therefore, the calculation amount can be reduced, short journey GPS point information and GPS point information of suburb boundary are taken as input, firstly, whether the GPS point information of the suburb boundary is closed or not is judged, if the GPS point information is not closed, the suburb boundary is closed, secondly, GPS points belonging to the city district area are judged by using an inpigon () function of Matlab, the proportion of the GPS points belonging to the city district area in the short journey to the total GPS points is calculated, and if more than 80% of the GPS points belong to the city district area, the short journey segment is considered to belong to the city district area category.
The short-stroke judgment belonging to the high-speed area is distinguished, the judgment method is that the closest point to each short-stroke GPS point is found in the high-speed library GPS point set, then the distance between the high-speed library GPS point and the closest short-stroke GPS point is calculated according to the longitude and latitude, if the distance is less than 50 m, the GPS point is considered to belong to the high-speed area, meanwhile, the proportion of the GPS points belonging to the high-speed area in the short stroke to the total GPS points is calculated, and if more than 60% of the GPS points belong to the high-speed area, the short-stroke segment is considered to belong to the high-speed area category.
The suburb is different from the urban area in that the suburb is not a convex space, so that the suburb range in a suburb large outline needs to be considered, the suburb part and the suburb part are divided, short-stroke GPS point information and GPS point information of suburb contour lines are taken as input, the first step is to judge whether the GPS point information of the suburb contour lines is closed or not, if the suburb contour lines are not closed, the GPS point information is closed, the second step is to judge the GPS point belonging to the suburb area by using an Inpolygon () function of Matlab, the proportion of the GPS point belonging to the suburb area in the short stroke to the total GPS point is calculated, and if more than 80% of the GPS points belong to the suburb area, the short-stroke segment is considered to belong to the suburb area category.
The method comprises the steps of taking short-stroke GPS point information and GPS point information of mountain area contour lines as input, judging whether the GPS point information of the mountain area contour lines is closed or not in the first step, closing the GPS point information if the GPS point information of the mountain area contour lines is not closed, judging the GPS points belonging to the mountain area by using an inpolygon () function of Matlab, calculating the proportion of the GPS points belonging to the mountain area in the short stroke to the total GPS points, and if more than 80% of the GPS points belong to the mountain area, determining that the short stroke segment belongs to the mountain area category.
And obtaining the region types passed by the user in the short journey, adding labels of the region types to which the vehicle runs in each short journey characteristic parameter database for subsequent synthesis working conditions, and obtaining a machine learning database for the step 3.
And 3, constructing a vehicle running road region classification model by using a machine learning random forest algorithm. And classifying and identifying the kinematic characteristic parameter data of the short running distance of the vehicle based on the characteristic parameter indexes by adopting a random forest algorithm according to the advantages of random forest classification purposes.
And training the database by using a random forest algorithm to obtain a random forest model. The random forest model is a classification recognition model for the operation road region type machine learning of a user, and the decision trees are not related to each other. The random forest model in the embodiment of the invention takes CART decision trees (decision tree algorithm taking reduction of GINI coefficient as division standard) with non-binary tree structures as weak learners, and each decision tree is generated according to a classification regression algorithm. The GINI coefficient, which may represent the frequency with which randomly selected data points in the data set may be misclassified, is calculated as follows:
1、p k representing the probability that the selected sample belongs to the k class, the probability that this sample is misclassified is 1-p k ；
2. There are K classes in the sample set, and a randomly selected sample may belong to any one of the K classes, so that the classes are summed, e.g. there are K classes in the sample set D, the number of K-th classes being C k Gini index of (a):
on the basis of using the decision tree, the random forest randomly selects a part of sample features on nodes, supposing Nsub, and then selects an optimal feature from the Nsub sample features randomly selected to divide left and right subtrees of the decision tree. This further enhances the generalization capability of the model. The specific process of constructing the vehicle operation road region classification model by using the random forest algorithm is as follows:
firstly, a machine learning database is imported, the machine learning database in step 2 comprises a plurality of user short-stroke data, and characteristic parameters of the user short-stroke data comprise running time, running distance, running speed, acceleration time, deceleration time, uniform speed time, idle time, acceleration proportion, deceleration proportion, uniform speed proportion, idle proportion, maximum speed, average speed, speed standard deviation, maximum acceleration, average acceleration of an acceleration section, standard difference of positive acceleration, maximum deceleration, average deceleration of a deceleration section, standard difference of negative deceleration, average acceleration, standard difference of acceleration, relative positive acceleration, proportion of different speed intervals, acceleration number, deceleration number, uniform speed number, parking number, standard difference of torque, average positive torque, average negative torque, maximum positive torque, maximum negative torque, average torque of an idle section, average positive torque of a running section, average negative torque of a running section, proportion of positive torque time momentum, proportion of negative torque time, proportion of torque time, maximum fluctuation time when torque is increased, maximum fluctuation time duration when torque is decreased, average fluctuation kilometer when torque is increased, average fluctuation kilometer when torque is decreased, average fluctuation time proportion of a yaw torque is increased, maximum damage time interval of a motor damage of a motor, total damage of a unit, mileage mark, damage of a motor, total damage area before and damage of a unit, a motor damage area of a motor damage of a motor.
The absolute maximum value of the yaw angular velocity, the velocity and mileage information of adjacent short strokes, the absolute maximum value of the lateral acceleration, the braking times, the hundred-kilometer energy consumption and other characteristic parameters are selected, so that the identification accuracy of the constructed classification model is facilitated.
Randomly dividing sample data into a training set and a testing set, wherein the proportion of the training set to the testing set is 7: and 3, solving the importance of the characteristic parameters and sequencing. The main parameters for designing the classifier are as follows:
the number of decision trees in the random forest was set to 178, the maximum tree depth was set to 9, the maximum number of leaf nodes was set to 41, the maximum feature number was set to 31, the minimum number of samples of leaf nodes was set to 3, and the minimum number of samples required for splitting was set to 5.
The following is a method of constructing a random forest:
1. one sample with the volume of N is drawn N times with the replacement, 1 sample is drawn each time, and finally N samples are formed. The selected N samples are used for training a decision tree as the samples at the root node of the decision tree.
2. When each sample has M attributes, when each node of the decision tree needs to be split, M attributes are randomly selected from the M attributes, and the condition that M is less than or equal to M is met. Then, a certain policy (for example, information gain) is applied to select 1 attribute from the m attributes as the splitting attribute of the node.
3. Each node in the decision tree formation process is split according to step 2 (if the attribute selected by the node next time is the attribute used by the parent node in splitting, the node has already reached the leaf node, and the splitting is not required to be continued) until the splitting can not be performed any more. Pruning is not performed in the whole decision tree forming process.
4. And (4) establishing a large number of decision trees according to the steps 1-3, thus forming a random forest.
And reasonably adjusting the proportion of the training set and the test set of the data in the machine learning database, and adjusting parameters of the model to enable the accuracy to be as high as possible. And finally, constructing a vehicle running road region classification model based on a machine learning algorithm.
Through verification, the identification accuracy of the constructed vehicle running road region classification model is higher, and the identification effect is shown in table 1.
TABLE 1
| Urban area | Gao Su | Suburb | Mountain area | |
| Test set accuracy | 92.85％ | 94.01％ | 93.90％ | 92.21％ |
| Testing recall rates | 93.97％ | 94.53％ | 94.96％ | 90.63％ |
In the embodiment of the invention, the short-stroke characteristic parameter database containing the zone type label to which the vehicle operation belongs is constructed in the steps 1 and 2, and based on the short-stroke characteristic parameter database, when the running working condition is synthesized, the running working conditions of the subareas of the typical city can be respectively synthesized according to the zone types, and the four zones can be merged and spliced according to the mileage length proportion of the four zone types to construct the total running working condition of the typical city. And 3, constructing a vehicle running road region classification model based on machine learning, inputting kinematic characteristic parameters and running data information of the vehicle to obtain the region type of the vehicle running, adjusting the control parameters of the whole vehicle based on the region type, and timely and accurately calibrating technical parameters to evaluate the performance of the whole vehicle.
The invention is not the best known technology.
The above embodiments are merely illustrative of the technical ideas and features of the present invention, and the purpose thereof is to enable those skilled in the art to understand the contents of the present invention and implement the present invention, and not to limit the protection scope of the present invention. All equivalent changes and modifications made according to the spirit of the present invention should be covered within the protection scope of the present invention.
Claims (7)
1. A vehicle operation road area type identification method is characterized by comprising the following steps:
the method comprises the following steps: dividing the types of mountain areas in the target area according to the geographical position information; dividing a target area into a city area type, a suburb area type and a high-speed area type according to the road type; and the range of the area type is embodied as discrete longitude and latitude point information;
step two: the method comprises the steps of taking user data as input, wherein the user data comprises vehicle running data information and vehicle GPS positioning information, preprocessing the data, dividing short strokes according to a short stroke method, dividing original data into a plurality of kinematic segments, and calculating characteristic parameters of each short stroke to obtain a short stroke characteristic parameter database; judging the area type of each short journey according to the GPS positioning information and the area type range information of the vehicle;
step three: adding labels of the types of the areas to which the vehicles run in each short-stroke characteristic parameter database; and meanwhile, a machine learning database is obtained, a machine learning algorithm is utilized for training, and a vehicle operation road region classification model based on machine learning is constructed and is used for identifying a vehicle operation road region.
2. The method for identifying the type of the vehicle operation road region as claimed in claim 1, wherein the first step specifically comprises:
determining a boundary between an urban area and a suburban area according to the urban administrative division range; dividing the boundary of the city and the suburb into a city area, and acquiring GPS dotted information of the boundary of the city area;
dividing all highways in the urban administrative area into high-speed areas, and acquiring GPS (global positioning system) dot-shaped information of the high-speed areas;
calculating the relief degree according to DEM elevation data, performing neighborhood analysis, reclassifying the relief degree to obtain a mountain area range, and obtaining GPS (global positioning system) dot information of a boundary of the mountain area;
the areas except the urban area, the high-speed area and the mountain area are divided into suburban areas, and suburban area boundary GPS point information is obtained.
3. The method as claimed in claim 2, wherein the boundary between the urban area and the suburban area is a high speed city-around line or a city-around line.
4. The vehicle operation road region type identification method according to claim 1, 2 or 3, wherein the second step specifically comprises:
classifying the short-stroke data of the user according to the region type; preprocessing original data including vehicle running data information and vehicle GPS positioning information, dividing short strokes according to a short stroke method, dividing the original data into a plurality of kinematic segments to obtain a user short stroke database, and calculating characteristic parameters of each short stroke for subsequent use;
and constructing a function for identifying the type of the short-trip belonging area, wherein the function takes the short-trip GPS positioning information and the GPS information of the four area types as input and takes the short-trip belonging area type label as output.
5. The vehicle operation road region type identification method according to claim 1, 2 or 3, wherein the third step specifically comprises:
the method comprises the steps of constructing a vehicle operation road region classification model by using a random forest algorithm of machine learning, obtaining a machine learning database which contains user short-travel data, characteristic parameters and labels, training the database by using the random forest algorithm, obtaining the random forest model, reasonably adjusting the proportion of a training set and a testing set of data in the machine learning database, performing model parameter adjustment according to an implementation effect, and finally constructing the vehicle operation road region classification model based on machine learning.
6. The method as claimed in claim 5, wherein the characteristic parameters include:
running time, running distance, running speed, acceleration time, deceleration time, uniform speed time, idle time, acceleration proportion, deceleration proportion, uniform speed proportion, idle proportion, maximum speed, average speed, speed standard deviation, maximum acceleration, average acceleration of an acceleration section, standard deviation of positive acceleration, maximum deceleration, average deceleration of a deceleration section, standard deviation of negative deceleration, average acceleration, standard deviation of acceleration, relative positive acceleration, proportion of different speed intervals, acceleration number, deceleration number, constant speed number, parking number, standard deviation of torque, average positive torque, average negative torque, maximum positive torque, maximum negative torque, average torque of an idle section, average torque of a running section, average positive torque of a running section, average negative torque of a running section, time proportion of positive torque, time proportion of negative torque, maximum fluctuation amount when torque is increased, maximum fluctuation amount when torque is decreased, standard deviation of torque fluctuation amount when torque is increased, average fluctuation amount when torque is decreased, time proportion of high-torque interval, time proportion of torque interval time of low torque, absolute maximum yaw angle speed, absolute braking frequency, energy consumption number of braking times, total damage of a motor damage per unit, damage of a motor, damage per unit of a motor, damage per unit of a motor, and damage per unit of a motor damage.
7. The vehicle operation road region type identification method according to claim 5, wherein the step of constructing the vehicle operation road region classification model by using the random forest algorithm comprises the following steps:
importing a machine learning database; randomly dividing sample data into a training set and a test set, wherein the proportion of the training set to the test set is 7:3, solving the importance of the characteristic parameters and sequencing;
the parameters of the design classifier are as follows: setting the number of decision trees in the random forest to be 178, setting the maximum tree depth to be 9, setting the maximum leaf node number to be 41, setting the maximum feature number to be 31, setting the minimum sample number of the leaf nodes to be 3, and setting the minimum sample number required for splitting to be 5;
and (4) performing model parameter adjustment to enable the accuracy to be as high as possible, and finally constructing a vehicle operation road region classification model based on a machine learning algorithm.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211439582.4A CN115798239B (en) | 2022-11-17 | 2022-11-17 | Vehicle running road area type identification method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211439582.4A CN115798239B (en) | 2022-11-17 | 2022-11-17 | Vehicle running road area type identification method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115798239A true CN115798239A (en) | 2023-03-14 |
| CN115798239B CN115798239B (en) | 2023-09-22 |
Family
ID=85438490
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211439582.4A Active CN115798239B (en) | 2022-11-17 | 2022-11-17 | Vehicle running road area type identification method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115798239B (en) |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110246051A1 (en) * | 2010-03-31 | 2011-10-06 | Denso International America, Inc. | Method of displaying traffic information and displaying traffic camera view for vehicle systems |
| US8892347B1 (en) * | 2013-06-01 | 2014-11-18 | Savari, Inc. | System and method for node adaptive filtering and congestion control for safety and mobility applications toward automated vehicles system |
| CN105359200A (en) * | 2013-06-26 | 2016-02-24 | 宝马股份公司 | Method for processing measurement data of a vehicle in order to determine the start of a search for a parking space |
| CN106790367A (en) * | 2016-11-15 | 2017-05-31 | 山东省科学院自动化研究所 | The vehicle safety hidden danger early warning of big data treatment and accident reproduction system and method |
| CN106997662A (en) * | 2017-04-26 | 2017-08-01 | 长安大学 | A kind of city bus operating mode construction method |
| EP3321854A2 (en) * | 2016-11-15 | 2018-05-16 | Panasonic Intellectual Property Corporation of America | Identification method, identification apparatus, classifier creating method, and classifier creating apparatus |
| CN108198425A (en) * | 2018-02-10 | 2018-06-22 | 长安大学 | A kind of construction method of Electric Vehicles Driving Cycle |
| CN109923575A (en) * | 2016-11-07 | 2019-06-21 | 瑞士再保险有限公司 | Absolute and/or relative risk possibility the device and method of automatic traffic and driving mode identification and position measurement of correlation car accident |
| CN112633722A (en) * | 2020-12-29 | 2021-04-09 | 交通运输部公路科学研究所 | Vehicle-mounted road safety risk assessment system and method |
| CN113538896A (en) * | 2021-04-20 | 2021-10-22 | 清华大学 | Critical driving condition analysis method and vehicle safety driving prompt system |
| CN113688558A (en) * | 2021-06-18 | 2021-11-23 | 长安大学 | Automobile driving condition construction method and system based on large database samples |
-
2022
- 2022-11-17 CN CN202211439582.4A patent/CN115798239B/en active Active
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110246051A1 (en) * | 2010-03-31 | 2011-10-06 | Denso International America, Inc. | Method of displaying traffic information and displaying traffic camera view for vehicle systems |
| US8892347B1 (en) * | 2013-06-01 | 2014-11-18 | Savari, Inc. | System and method for node adaptive filtering and congestion control for safety and mobility applications toward automated vehicles system |
| CN105359200A (en) * | 2013-06-26 | 2016-02-24 | 宝马股份公司 | Method for processing measurement data of a vehicle in order to determine the start of a search for a parking space |
| CN109923575A (en) * | 2016-11-07 | 2019-06-21 | 瑞士再保险有限公司 | Absolute and/or relative risk possibility the device and method of automatic traffic and driving mode identification and position measurement of correlation car accident |
| CN109952592A (en) * | 2016-11-07 | 2019-06-28 | 瑞士再保险有限公司 | The absolute and relative risk system and method for automated location measurement of correlation and prediction automobile risk |
| CN106790367A (en) * | 2016-11-15 | 2017-05-31 | 山东省科学院自动化研究所 | The vehicle safety hidden danger early warning of big data treatment and accident reproduction system and method |
| EP3321854A2 (en) * | 2016-11-15 | 2018-05-16 | Panasonic Intellectual Property Corporation of America | Identification method, identification apparatus, classifier creating method, and classifier creating apparatus |
| CN106997662A (en) * | 2017-04-26 | 2017-08-01 | 长安大学 | A kind of city bus operating mode construction method |
| CN108198425A (en) * | 2018-02-10 | 2018-06-22 | 长安大学 | A kind of construction method of Electric Vehicles Driving Cycle |
| CN112633722A (en) * | 2020-12-29 | 2021-04-09 | 交通运输部公路科学研究所 | Vehicle-mounted road safety risk assessment system and method |
| CN113538896A (en) * | 2021-04-20 | 2021-10-22 | 清华大学 | Critical driving condition analysis method and vehicle safety driving prompt system |
| CN113688558A (en) * | 2021-06-18 | 2021-11-23 | 长安大学 | Automobile driving condition construction method and system based on large database samples |
Non-Patent Citations (1)
| Title |
|---|
| 李卓;陈雨人;: "基于决策树方法的偏远地区山区公路驾驶安全性研究", 华东交通大学学报, no. 02 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115798239B (en) | 2023-09-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108427965B (en) | Hot spot area mining method based on road network clustering | |
| CN110298500B (en) | Urban traffic track data set generation method based on taxi data and urban road network | |
| CN108920481B (en) | Road network reconstruction method and system based on mobile phone positioning data | |
| US11403938B2 (en) | Method for determining traffic metrics of a road network | |
| WO2023050955A1 (en) | Urban functional zone identification method based on function mixing degree and ensemble learning | |
| CN110503139A (en) | A kind of urban area tail gas pollution prediction technique | |
| CN107067781B (en) | GIS road black spot map generation method for advanced driving assistance system application | |
| CN111696369A (en) | Whole-city road time-division vehicle type traffic flow prediction method based on multi-source geographic space big data | |
| CN110188797A (en) | A kind of intelligent automobile method for rapidly testing based on Bayes's optimization | |
| CN110889794A (en) | Statistical method for municipal road land acquisition houses | |
| Zhang et al. | Using street view images to identify road noise barriers with ensemble classification model and geospatial analysis | |
| CN105574313B (en) | A kind of determination method and apparatus of category of roads | |
| CN111242223B (en) | Street space quality evaluation method based on streetscape image multi-feature fusion | |
| CN116542709A (en) | Electric vehicle charging station planning analysis method based on traffic situation awareness | |
| CN106126637A (en) | A kind of vehicles classification recognition methods and device | |
| CN109800903A (en) | A kind of profit route planning method based on taxi track data | |
| CN115620165B (en) | Method, device, equipment and medium for evaluating slow-moving system facilities of urban built-up area | |
| CN115798239B (en) | Vehicle running road area type identification method | |
| CN114170441B (en) | Roadside tree automatic extraction method based on geographic national condition data and image classification | |
| EP3922947A2 (en) | Traffic analytics system for defining road networks | |
| Di Nicola et al. | A new approach for roughness representation within urban dispersion models | |
| CN112905856A (en) | Method for constructing high-speed traffic data set with space-time dependence | |
| CN114387501A (en) | Remote sensing intelligent identification method and device for deposition of flood-washing fan | |
| CN118568526B (en) | Comprehensive transportation channel identification method based on cluster analysis | |
| EP3913551A1 (en) | Method for defining road networks |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |