CN112629533B - Fine path planning method based on road network rasterization road traffic prediction - Google Patents
Fine path planning method based on road network rasterization road traffic prediction Download PDFInfo
- Publication number
- CN112629533B CN112629533B CN202011252260.XA CN202011252260A CN112629533B CN 112629533 B CN112629533 B CN 112629533B CN 202011252260 A CN202011252260 A CN 202011252260A CN 112629533 B CN112629533 B CN 112629533B
- Authority
- CN
- China
- Prior art keywords
- road
- traffic
- grid
- node
- vehicle
- 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
- G01C21/20—Instruments for performing navigational calculations
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
- G01C21/26—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 specially adapted for navigation in a road network
- G01C21/34—Route searching; Route guidance
- G01C21/3407—Route searching; Route guidance specially adapted for specific applications
- G01C21/343—Calculating itineraries, i.e. routes leading from a starting point to a series of categorical destinations using a global route restraint, round trips, touristic trips
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
- G01C21/26—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 specially adapted for navigation in a road network
- G01C21/34—Route searching; Route guidance
- G01C21/3446—Details of route searching algorithms, e.g. Dijkstra, A*, arc-flags, using precalculated routes
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Abstract
The invention discloses a path planning method based on road network rasterization road traffic prediction, which comprises the following steps: dividing a city into M multiplied by N grid areas based on a road network diagram of the city, and distributing GPS or Beidou data of all vehicles at a certain moment into grids of the city according to longitude, latitude and direction in the road network of the city to obtain density diagrams of the road network diagram of the city in different directions; taking the obtained density map of the road network at a certain moment as the input of the neural network, and predicting the neural network for 10+/-5 minutes in the future by using the road network density data of the first 30+/-15 minutes; and calculating the average running speed of the vehicle of each road after prediction according to the traffic density of the grid area around the road at a certain moment before prediction and the relationship between the traffic density of the road and the average speed. And dividing a grid area according to the current position of a certain vehicle, and selecting the road of the vehicle in the certain grid area according to the predicted road vehicle density and average running speed information of 10+/-5 minutes in the future.
Description
Technical Field
The invention belongs to the technical field of intelligent traffic, and relates to an urban traffic density prediction method based on a neural network and a vehicle path planning method based on traffic prediction.
Background
With the development of intelligent traffic, urban road infrastructure is continuously perfected, and various ways for acquiring various traffic information in cities are rich and various, so that required traffic data can be easily acquired. The traffic data has the characteristics of huge quantity, various types, difficult law expression and the like, and the artificial intelligence and the neural network can well help us find out the hidden law in a large amount of traffic data. The neural network is utilized to find the rules in the traffic condition, the trend of the future traffic condition can be predicted, and more effective and efficient traffic management measures are formulated. This is thus of great potential for traffic prediction using neural networks, and their advantages make it one of the current research hotspots in the field of intelligent traffic technology.
The conventional vehicle path planning algorithm is realized in this way: for vehicles from a start point to a destination, the road closest to the destination is found by traversing nodes and roads of the road network. Conventional vehicle path planning algorithms are able to find the shortest path between the origin and destination, but suffer from a number of drawbacks: the method can not flexibly and intelligently adapt to complex and changeable urban road traffic; the calculated amount is huge when the distance is long; all vehicles are easy to guide to drive on a small number of roads to cause traffic jam; the selected road is not adjustable. Based on intelligent traffic and artificial intelligence development, traditional vehicle path planning algorithms have been gradually improved and replaced by other path planning algorithms to facilitate effective urban traffic management in the context of intelligent traffic and artificial intelligence technology and further accommodate development and application of intelligent traffic and artificial intelligence.
Through a search of the prior art, it was found that Jinglin Li et al and Meng Chen et al published articles entitled "An End-to-End Load Balancer Based on Deep Learning for Vehicular Network Traffic Control (End-to-End load equalizer for vehicular network flow control based on deep learning)" and "PCNN: deep Convolutional Networks for Short-Term Traffic Congestion Prediction (PCNN: deep convolutional network short-term traffic congestion prediction)" in IEEE INTERNET OF THINGS JOURNAL (journal of IEEE Internet of things) and IEEE Transactions on Intelligent Transportation Systems (journal of IEEE Intelligent transportation systems) in 2018, respectively. These articles propose urban traffic flow prediction methods at the regional level and the road level, respectively, based on neural networks. Although the urban traffic flow prediction method at the area level and the road level can meet the needs of part of the urban traffic condition prediction, plays a certain guiding role in specifying traffic management measures, and is difficult to meet the requirements of path planning in future intelligent traffic for the problem of predicting the traffic density of roads at the urban level.
It has also been found by search that in order to further optimize the vehicle path planning algorithm, chang Guo et al in 2018 entitled Real-Time Path Planning in Urban Area via VANET-Assisted Traffic Information Sharing (urban Real-time path planning via VANET assisted traffic information sharing) propose a method for changing the vehicle travel path based on Real-time information. By judging the traffic flow condition of the road ahead in real time, judging whether the road ahead is a congested road section or not and whether the road needs to be replaced for running, the method avoids the vehicle from running on the congested road section to a certain extent. In addition, it has been found by searching that Miao Wang et al published in 2015, "IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY (journal of IEEE vehicle technology)", entitled Real-Time Path Planning Based on Hybrid-VANET-Enhanced Transportation System (Real-time path planning hybrid VANET enhanced transportation System) which prioritizes roads and intersections of different congestion levels based on road network traffic optimization, and vehicles at congested road segments and intersections can be given priority to road replacement rights, which can promote road network traffic optimization to some extent.
In summary, the problems of the prior art are: the existing traffic prediction method is generally limited to regional level or road level, road traffic flow of city orders is difficult to predict (2) most of path planning algorithms cannot adapt to various urban traffic conditions (3) and adjust vehicle paths according to road information in real time, and the situation that front congestion is known and optional roads are not available often occurs. The significance of solving the technical problems is that: based on the development of the existing artificial intelligence technology and the development of the intelligent traffic technology, the more efficient and reliable urban traffic prediction method can improve the accuracy of traffic prediction and enrich the information quantity of traffic prediction, provide a new thought for future urban traffic management measures and promote the development of the intelligent traffic field technology and the application of artificial intelligence in the traffic field.
Disclosure of Invention
Aiming at the problems existing in the prior art, the invention aims to provide a gridding fine path planning method based on urban road network gridding road vehicle density prediction.
The technical scheme of the invention is realized in such a way that a path planning method based on road network rasterization road traffic flow prediction is realized, and the urban road network rasterization road traffic flow density prediction method and the rasterization fine path planning method comprise the following steps:
step 1: dividing a city into grid areas with M multiplied by N (M, N is 30-200) based on a road network diagram of the city, and distributing GPS or Beidou data of all vehicles at a certain moment into grids of the city according to longitude, latitude and direction in the road network of the city to obtain density diagrams of the road network diagram of the city in different directions;
step 2: taking the obtained density map of the road network at a certain moment as the input of the neural network, and predicting the neural network for 10+/-5 minutes in the future by using the road network density data of the first 30+/-15 minutes;
step 3: and calculating the average running speed of the vehicle of each road after prediction according to the traffic density of the grid area around the road at a certain moment before prediction and the relationship between the traffic density of the road and the average speed.
Step 4: and dividing a grid area according to the current position of a certain vehicle, and selecting the road of the vehicle in the certain grid area according to the predicted road vehicle density and average running speed information of 10+/-5 minutes in the future.
Step 5: if the driving end point of a certain vehicle is not in the grid area, repeating the step 4 when the vehicle is about to complete the planned journey within 10+/-5 minutes until the end point is in the certain grid area.
Further, the convolutional neural network is utilized to predict the traffic (vehicle) density in the urban level area range, so that the traffic change (vehicle road flow) change condition at a certain time in the future can be effectively predicted.
The urban road network diagram is subjected to grid division, and four sections of southeast, southwest and northwest are divided according to the running direction of the vehicle, so that the urban regional-level traffic prediction can be refined to each road according to the matching of grids and specific roads.
And the vehicle performs path planning, and simultaneously calculates the shortest path of the vehicle and the estimated distance from the edge point of the grid area to the end point by considering the density prediction information of the road in the grid area, so that the vehicle is ensured to be on an effective road which goes to the end point. The invention mainly utilizes the prediction information to grasp the change trend of future traffic.
The traffic density of the road has close relation with the average speed of the road, and the method calculates the average speed of the vehicle of each road by using the traffic density of the road and uses the average speed of the vehicle for path planning.
The vehicle utilizes the predicted traffic information to carry out path planning, so that the selection of a road section which will become congested in the future can be avoided, and the vehicle has the function of traffic guidance and balancing the urban road traffic flow for a long time.
In particular, the vehicle only plans the path in the 10-minute area each time, so that the problems that traffic prediction becomes longer along with time and prediction accuracy index is reduced can be effectively reduced, and meanwhile, the congestion caused by road accident can be reduced. The problem that the accuracy index is reduced when the traffic prediction becomes longer along with time can be effectively reduced, and the method can be more suitable for complex and changeable traffic conditions.
And when the vehicle performs path planning, the shortest path calculated by using the vehicle density prediction information in the grid area and the estimated distance from the edge point of the grid area to the end point are simultaneously considered, so that the vehicle is ensured to be on an effective road going to the end point.
The central control management mechanism for collecting traffic information, carrying out traffic prediction and providing path planning for vehicles by using the traffic big data center comprises the following implementation steps: when a vehicle needs to carry out path planning from a starting point to an end point, dividing a road in a sector area with the starting point as an origin and the radius of a path of 10 minutes as a grid area, and sending an entering moment and the starting point end point to a traffic manager; the traffic big data center continuously predicts future traffic information according to the information collected by urban traffic, and plans the path in the grid for the vehicle according to the predicted information in the grid and the estimated distance from the edge point of the grid to the end point after receiving the path demand information of the vehicle;
the urban road network is typically divided into 100 times 100 grid areas, the running direction of the vehicle is divided into four direction areas of southeast, northwest and northwest, and then GPS data of the vehicle is distributed into the grids according to longitude, latitude and the running direction in a time interval of 10 minutes, so that a density map of running of the vehicle in each direction in the urban road network is obtained:
traffic prediction is performed by adopting a convolutional neural network and a residual neural structure, a traffic flow density map (road network density data of the first 30+/-15 minutes) collected in real time is used as input of the neural network to predict a traffic flow density map of the next time interval, namely 10 minutes later, the density map is matched with each road in the road network, and finally v=v according to the relation between the road density and the average speed of the road max (1-λ/λ max ) The average travel speed of each link is calculated to obtain speed information of each link (average speed of links from i to j at a certain time t):
wherein v is max Is the speed limit of the road, lambda and lambda max The traffic density and the traffic flow of the road respectivelyDensity maximum.
In the path planning part, the gridding fine path planning is divided into two aspects of optimal path calculation in a grid and estimated distance calculation from a starting point to an end point through each node, when a certain vehicle t is from the starting point O to the end point D, an elliptical path search space, namely nodes possibly passing through from the starting point O to the end point D (the nodes represent connecting points of different roads and a plurality of nodes exist in a grid area) is established according to the distance between the two points, and the estimated distance from the nodes in the search space to the end point is calculated; the path planning grid is not much the same as the traffic prediction grid, the path planning grid is a small grid area integration within 10 minutes of the journey divided according to the vehicle position, i.e. within 10min vmax, then the nodes are the connection points of the road, i.e. the road is such a node from point a to point B, then the edge points are the nodes where the node is located at the outermost periphery of the large grid and can leave the current grid area from this node.
Each time the vehicle divides the current position as the origin, predicting small grid areas by all road networks within the distance of 10 minutes as large grid areas for vehicle path planning, calculating the shortest time and paths from the current position to the edge points in the grids (the edge points are the edge positions of the grids and the roads leaving the grids from the points) according to the prediction information, then synthesizing the estimated distance from the edge nodes to the end points, judging which edge node leaves the grid areas, dividing the grids again when the vehicle is about to leave the grid areas, and carrying out the path calculation process until the end points are terminated when the end points are in the grids;
the node set of the grid region is first defined as a= { a 1 ,a 2 ,a 3 … and search space b= { B 1 ,b 2 ,b 3 …}
In summary, the path planning problem for each time the vehicle is rasterized according to the present embodiment may be written as:
wherein a is i ,Normal node and edge node in grid area, respectively,/->For the current position to grid area edge node a i Is (are) shortest time->For the estimated distance of the edge node to the end point, α is a weight between 0 and 1, which means that the path planning in the grid according to the prediction information takes a larger weight, because the path planning in the grid is an accurate time calculated according to the prediction information, and the estimated distance is a value calculated according to the static traffic information and has lower accuracy than the calculation according to the prediction information.
Solving the above problem once every time the vehicle divides the grid, when a i Including termination at the end point.
The invention relates to a refined path planning method based on road network rasterization road traffic prediction. And urban traffic density direction division is adopted, and a neural network is utilized to predict traffic density and a vehicle navigation rasterization path planning algorithm. Since the average speed of a road is closely related to the traffic density of the road, the vehicle can stepwise select different roads according to the predicted traffic density of the urban road, thereby finding a travel path with the minimum time cost in each stage. Compared with the traditional traffic navigation method, the method predicts the traffic flow density of the urban road based on the artificial intelligence technology, avoids repeated congestion of vehicles through a rasterized path planning algorithm, can find out a road with shorter time required for leading to a destination, improves the average speed in the driving process, and further achieves the effect of traffic guidance.
The beneficial effects are that: compared with the prior art, the urban road network rasterized road vehicle density prediction method and the rasterized fine path planning method have the advantages that firstly, vehicles can be effectively prevented from running towards a congestion area and a road section direction; secondly, the urban road network grid road traffic density prediction method is one of the outstanding contributions of the patent; and thirdly, the gridding staged path planning is carried out when the vehicle carries out path planning, so that the calculation complexity is reduced, the flexibility of selecting roads by the vehicle is improved, and the robustness of the whole traffic system is greatly improved.
Drawings
FIG. 1 is a rasterized fine path planning scene graph employed by an embodiment of the invention.
FIG. 2 is a block diagram of urban road network rasterized road traffic density prediction in accordance with an embodiment of the present invention.
FIG. 3 is a block diagram of an implementation of the optimal path algorithm in a grid area in accordance with an embodiment of the present invention.
FIG. 4 is a block diagram of an implementation of the node-to-endpoint distance estimation algorithm in accordance with an embodiment of the present invention.
FIG. 5 is a graph showing a comparison of average travel time of a rasterized fine path planning method based on predictive information and a conventional path planning method;
FIG. 6 is a graph showing a comparison of average travel speeds of a rasterized fine path planning method based on predictive information and a conventional path planning method;
fig. 7 is a schematic diagram showing comparison of average calculation time of a rasterized fine path planning method based on prediction information and a conventional path planning method.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the following detailed description of the embodiments of the present invention is given with reference to the accompanying drawings: the embodiment is implemented on the premise of the technical scheme of the invention, and detailed implementation modes and specific operation processes are given. It should be understood that the specific examples described herein are for illustrative purposes only and that the scope of the present invention is not limited to the following examples.
Examples
The embodiment adopts the urban traffic road network scene of fig. 1, and provides a urban road network rasterized road vehicle density prediction method and a rasterized fine path planning method. Firstly, in the traffic scene, a traffic big data center needs to be set up in a city. The traffic big data center is responsible for collecting the whole urban traffic data, collating and analyzing the collected information, predicting the density of road vehicles with urban road grids by utilizing the neural network, and providing service for the vehicles needing path planning according to the predicted traffic information.
The basic object of the present embodiment is to obtain road network traffic prediction information by using an artificial intelligence-based traffic prediction technique, and to use for path selection of vehicles. In order to avoid the degradation of accuracy caused by long-time prediction and reduce the calculation amount of path planning, the traffic big data center only needs to plan the path in the grid area with the vehicle as the center radius and the path of 10 minutes at a time. The central control management mechanism for collecting traffic information, carrying out traffic prediction and providing path planning for vehicles by using the traffic big data center comprises the following implementation steps: when a vehicle needs to carry out path planning from a starting point to an end point, dividing a road in a sector area with the starting point as an origin and the radius of a path of 10 minutes as a grid area, and sending an entering moment and the starting point end point to a traffic manager; the traffic big data center continuously predicts future traffic information according to the information collected by urban traffic, and plans the path in the grid for the vehicle according to the predicted information in the grid and the estimated distance from the edge point of the grid to the end point after receiving the path demand information of the vehicle. The vehicle path planning mechanism is characterized in that the traffic big data center only needs to perform path calculation in the grid when the vehicle needs to perform path planning, the calculation complexity is low, and in addition, the traffic big data center only plans a path for the vehicle for 10 minutes each time, so that the path of the vehicle has adjustability and the predicted traffic information is fully utilized.
In this embodiment, the urban road network rasterized road vehicle density prediction method and the rasterized fine path planning method may be divided into two parts, i.e., traffic prediction and path planning.
In the traffic prediction part, a certain urban road network is typically divided into 100-100 grid areas, the running direction of a vehicle is divided into four direction areas of southeast, southwest and northwest, and then GPS data of the vehicle is distributed into the grids according to longitude, latitude and the running direction by taking 10 minutes as a time interval, so that a density map of running of the vehicle in each direction in the urban road network is obtained:
where (m, n) is the mesh region of the mth row and the nth column,is the information of vehicle k at time t, including position information and speed information> Meaning that the vehicle t moment is within the grid area (m, n) and the direction of travel is within the direction interval s i And (3) inner part. Thus in the direction of travel s i The density of vehicles within the grid region (m, n) of (c) can be x t (s i M, n). Finally, the urban road network traffic density distribution condition of each time interval can be obtained:
{X t |t＝t 0 ,t 0 -1…}
next we build a neural network to make predictions of traffic. In the invention, the convolutional neural network and the residual neural structure are adopted for traffic prediction, the convolutional neural network is a neural network with powerful functions, the training complexity of the neural network can be greatly reduced through sharing parameters and the like, and the residual neural structure has a good effect in training a deep network. A layer of convolutional neural network may be represented by the following formula:
X l+1 ＝f l (W l *X l +b l )
wherein X represents a convolution operation l Is the input of the neural network and the output of the upper layer neural network, X l+1 Then is the output of the layer of neural network, W l And b l Is the layer of neural netLinear operation parameter of complex, f l Is a nonlinear operational function of the neural network of this layer, and we used the ReLu function in this study.
Whereas the residual neural structure consists of a three-layer convolutional neural network, and its inputs and outputs are connected, so its formula is as follows:
and->Input and output of residual neural structure, respectively, < >>Refers to a specific construction of a residual neural structure that includes 3 convolutional neural networks and a ReLu function.
And then training the neural network according to the historical traffic information and the error function, wherein the training aim is to minimize the value of the error function, and the error function is as follows:
wherein x is t (s i M, n) andthe true value and the predicted value of the traffic information are respectively, and S is the number of direction intervals, for example, s=4 in the case of four direction intervals of the southwest and northwest.
Finally, the traffic flow density map collected in real time is used as the input of the neural network to predict the traffic flow density map of the next time interval, the density map is matched with each road in the road network, and finally, the traffic flow density map is matched with each road according to the road densityRelationship v=v of average road speed max (1-λ/λ max ) The average travel speed of each link is calculated to obtain speed information of each link (average speed of links from i to j at a certain time t):
V(i,j,t)
the effect of urban road gridding traffic flow density prediction is shown in fig. 2.
In the path planning section, the gridding fine path planning can be divided into two aspects of optimal path calculation in a grid and calculation of estimated distance from a starting point to an end point through each node, when a certain vehicle t is from the starting point O to the end point D, an elliptical path search space, namely, a node (the node represents a connecting point of different roads and a plurality of nodes exist in one grid area) possibly passing through from the starting point O to the end point D is established according to the distance between the two points, and the estimated distance from the node in the search space to the end point is calculated. Every time the vehicle divides the current position as the origin, all road networks with the radius of 10 minutes are used for predicting small grid areas as large grid areas for vehicle path planning, the shortest time and the paths from the current position to the edge points in the grids (the edge points are the edge positions of the grids and the roads leaving the grids from the edge points) are calculated according to the prediction information, the estimated distance from the edge nodes to the end points is synthesized, the grid areas are judged to be left from the edge nodes, when the vehicle is about to leave the grid areas, the grids are divided again, and the path calculation process is carried out until the end points are terminated when the end points are in the grids. Here we first define the node set of the grid region as a= { a 1 ,a 2 ,a 3 … and search space b= { B 1 ,b 2 ,b 3 …}
In summary, the path planning problem for each time the vehicle is rasterized according to the present embodiment may be written as:
wherein a is i ,Normal node and edge node in grid area, respectively,/->For the current position to grid area edge node a i Is (are) shortest time->For the estimated distance of the edge node to the end point, α is a weight between 0 and 1, which means that the path planning in the grid according to the prediction information takes a larger weight, because the path planning in the grid is an accurate time calculated according to the prediction information, and the estimated distance is a value calculated according to the static traffic information and has lower accuracy than the calculation according to the prediction information.
The path planning grid is not much the same as the traffic prediction grid, the path planning grid is a small grid area integration within 10 minutes of the journey divided according to the vehicle position, i.e. within 10min vmax, then the nodes are the connection points of the road, i.e. the road is such a node from point a to point B, then the edge points are the nodes where the node is located at the outermost periphery of the large grid and can leave the current grid area from this node.
Solving the above problem once every time the vehicle divides the grid, when a i Including termination at the end point.
Solving the above problem includes two-aspect calculation, the calculation of the estimated distance from each node to the end point and the calculation of the optimal path in the grid, and the solving process is shown in fig. 3 and fig. 4 respectively.
For the optimal path calculation in the grid, referring to dijistra algorithm, calculating the shortest time from the starting point to the edge point in the grid, selecting the node with the shortest time from the starting point from unselected nodes in the grid each time, and updating the shortest time from the starting point to the neighbor nodes of the node:
wherein a is i For the selected node(s),l is the neighbor node of the node i,j Is the distance between two points, V (i, j, T (a) i ) At T (a) i ) And obtaining the average running speed of the road between the two nodes according to traffic prediction at the moment.
Then the calculated node a i And (3) selecting the node with the shortest time required by the next departure point as the selected node, and repeatedly selecting the node with the shortest time required by the next departure point until all the nodes are iterated or the value of the node with the shortest time required by the departure point is infinity, thus obtaining the shortest time T from the departure point to each edge point.
For the calculation of the estimated distance from each node to the end point, for the nodes in the search space, we first adopt a layering method to layer the whole node set according to the number of nodes between the nodes and the end point, and then calculate the estimated distance from the nodes of each layer to the end point. Specifically, from the end point D, layering the nodes in the node set B, if there are at least n nodes between the nodes and the end point D, the nodes belong to the n-th layer node, and then from the 0-th layer node, calculating the estimated distance between the nodes and the end point in the j-th layer each time according to static road information, wherein the estimated distance is as follows:
wherein the method comprises the steps ofFor the node in layer j +.>Is->And finally obtaining the estimated distance Q from each point to the end point D according to the formula at the neighbor node in the j-1 layer.
And finally, solving and solving the specific running path of the vehicle in the grid area according to the shortest time T from the starting point to the edge node based on the traffic prediction information and the estimated distance D from the edge node to the end point based on the static traffic information.
In order to make this embodiment more intuitive and compare the performance of the traffic prediction information-based rasterized fine path planning algorithm with that of the conventional path planning algorithm, fig. 5,6, and 7 show a comparison diagram of average travel time, speed, and calculation time of the traffic prediction-based rasterized fine path planning algorithm and the conventional shortest distance path planning algorithm, and the shortest path algorithm that relies on real-time information. The traditional shortest path planning algorithm only considers the physical distance of the selected road, selects the road with the shortest total distance, and calculates the path with the shortest total travel time required according to the current traffic condition by means of the shortest path algorithm of real-time information. It can be obviously found that the travel time of the traffic prediction-based rasterization fine path planning algorithm is obviously smaller than that of the traditional shortest distance path planning algorithm, and is close to or even slightly smaller than that of the shortest path algorithm depending on real-time information. The travel average speed of the rasterized fine path planning algorithm based on traffic prediction is greater than that of the traditional path planning algorithm and the shortest path algorithm depending on real-time information, and the travel distance is slightly greater than that of the traditional shortest path planning. The method is characterized in that the traditional path planning is to select the path with the shortest distance without considering real-time traffic information, the shortest path algorithm depending on the real-time information considers the current traffic information, but cannot completely adapt to the traffic condition of instantaneous change, and the grid fine path planning algorithm based on traffic prediction can effectively predict the blocking condition of the road, avoid selecting the road with congestion or the road to be congested, find the road with a slightly longer possible driving distance but low congestion degree, and achieve the effects of reducing the travel time and improving the average travel speed. In addition, the traffic prediction-based rasterization fine path planning algorithm has obvious advantages in the vicinity of 5-8 points in the morning every day, because the time period belongs to a transition interval from a trip valley to a trip peak, the traffic density of roads in the time period changes rapidly, the general path algorithm cannot adapt to the urban traffic which changes rapidly, and the traffic prediction-based rasterization fine path planning algorithm can well find the roads with shorter required time in the urban traffic network which changes rapidly by a traffic information prediction method.
The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather is intended to cover all modifications, equivalents, and alternatives falling within the spirit and principles of the invention.
Claims (3)
1. The path planning method based on the road network rasterized road traffic prediction is characterized by comprising the following steps of: the road network rasterization road traffic density prediction method and the rasterization fine path planning method comprise the following steps:
step 1: dividing a city into M multiplied by N grid areas based on a road network diagram of the city, wherein M, N is 30-200, and distributing GPS or Beidou data of all vehicles at a certain moment into grids of the city according to longitude, latitude and direction in the road network of the city to obtain density diagrams of the road network diagram of the city in different directions;
step 2: taking the obtained density map of the road network at a certain moment as the input of the neural network, and predicting the neural network for 10+/-5 minutes in the future by using the road network density data of the first 30+/-15 minutes;
step 3: calculating the average running speed of the vehicle of each road after prediction according to the traffic density of the grid area around the road at a certain moment before prediction and the relationship between the traffic density of the road and the average speed;
step 4: dividing a grid area according to the current position of a certain vehicle, and selecting a road of the vehicle in the certain grid area according to predicted road vehicle density and average running speed information of 10+/-5 minutes in the future;
step 5: if the driving end point of a certain vehicle is not in the grid area, repeating the step 4 when the vehicle is about to complete the planned journey within 10+/-5 minutes until the end point is in the certain grid area;
the central control management mechanism for collecting traffic information, carrying out traffic prediction and providing path planning for vehicles by using the traffic big data center comprises the following implementation steps: when a vehicle needs to carry out path planning from a starting point to an end point, dividing a road in a sector area with the starting point as an origin and the radius of a path of 10 minutes as a grid area, and sending an entering moment and the starting point end point to a traffic manager; the traffic big data center continuously predicts future traffic information according to the information collected by urban traffic, and plans the path in the grid for the vehicle according to the predicted information in the grid and the estimated distance from the edge point of the grid to the end point after receiving the path demand information of the vehicle;
the urban road network is typically divided into 100 times 100 grid areas, the running direction of the vehicle is divided into four direction areas of southeast, northwest and northwest, and then GPS data of the vehicle is distributed into the grids according to longitude, latitude and the running direction in a time interval of 10 minutes, so that a density map of running of the vehicle in each direction in the urban road network is obtained:
traffic prediction is carried out by adopting a convolutional neural network and a residual neural structure, a traffic flow density map collected in real time, namely road network density data of the first 30+/-15 minutes, is used as input of the neural network to predict the traffic flow density map of the next time interval, namely 10 minutes later, the density map is matched with each road in the road network, and finally v=v according to the relation between the road density and the average speed of the road max (1-λ/λ max ) Calculating the average running speed of each road to obtain speed information of each road, wherein the speed information refers to the average speed of the road from i to j at a certain moment t:
wherein v is max Is the speed limit of the road, lambda and lambda max Respectively the maximum value of the traffic density and the traffic density of the road;
in a path planning part, a grid fine path is divided into two aspects of optimal path calculation in a grid and estimated distance calculation from a starting point to an end point through each node, when a certain vehicle t is from the starting point O to the end point D, an elliptical path search space, namely a node which possibly passes from the starting point O to the end point D, is established according to the distance between the two points, the node represents the connecting point of different roads, and a plurality of nodes exist in one grid area; calculating an estimated distance from the node in the search space to the end point;
each time the vehicle divides the current position as an origin, predicting small grid areas by using all road networks within a path with the radius of 10 minutes as large grid areas for vehicle path planning, calculating the shortest time and the path from the current position to edge points in the grids according to the prediction information, wherein the edge points are the edge positions of the grids, the roads leaving the grids from the points exist, then synthesizing the estimated distance from the edge nodes to the end points, judging which edge node leaves the grid areas, dividing the grids again when the vehicle is about to leave the grid areas, and performing a path calculation process until the end points are terminated when the end points are in the grids;
the node set defining the grid region is a= { a 1 ，a 2 ，a 3 … and search space b= { B 1 ，b 2 ，b 3 …}；
In summary, each time the vehicle performs the rasterized path planning problem is written as:
wherein the method comprises the steps ofNormal node and edge node in grid area, respectively,/->For the current position to grid area edge node a i Is (are) shortest time->For the estimated distance of the edge node to the end point, α is a weight between 0,1, which means that the path planning in the grid based on the prediction information takes a larger weight, since the path planning in the grid is based on the predictionThe exact time of the information calculation, while the estimated distance is a value calculated from the static traffic information, lower than the accuracy calculated from the predicted information; solving the above problem once every time the vehicle divides the grid, when a i Including termination at the end point.
2. The urban road network rasterized road traffic density prediction method and the rasterized fine path planning method according to claim 1, characterized in that: the vehicle only plans the path in the 10-minute area each time, so that the problems that traffic prediction becomes longer along with time and the prediction accuracy index is reduced are effectively solved, and meanwhile, the congestion caused by road accident can be reduced.
3. The urban road network rasterized road traffic density prediction method and the rasterized fine path planning method according to claim 1, characterized in that: solving the above problem includes two-way calculation, calculation of estimated distance from each node to the end point and calculation of optimal path in the grid, for the calculation of optimal path in the grid, referring to dijistra algorithm, calculating the shortest time from the start point to the edge point in the grid, selecting the node with the shortest time from the unselected node in the grid, and updating the shortest time from the start point to the neighbor node:
wherein a is i For the selected node(s),l is the neighbor node of the node i，j Is the distance between two points, V (i, j, T (a) i ) At T (a) i ) The average running speed of the road between the two nodes is obtained according to traffic prediction at the moment;
then the calculated node a i Listed as selected node, and the selection of the node with the shortest time required for the next departure point is repeatedSelecting, until all nodes are iterated or the value of the node with the shortest time required for leaving the starting point is infinity, and obtaining the shortest time T from the starting point to each edge point;
for the calculation of the estimated distance from each node to the end point, for the nodes in the search space, firstly layering the whole node set according to the number of nodes between the nodes and the end point by adopting a layering method, and then calculating the estimated distance from the nodes of each layer to the end point: layering nodes in the node set B from the end point D, if at least n nodes exist between the nodes and the end point D, the nodes belong to the node of the nth layer, and then according to static road information, calculating the estimated distance between the nodes in the jth layer and the end point each time from the node of the 0 th layer, wherein the estimated distance is represented by the following formula:
wherein the method comprises the steps ofFor the node in layer j +.>Is->The neighbor nodes in the j-1 layer finally obtain the estimated distance Q from each point to the end point D according to the formula;
and finally, solving and solving the specific running path of the vehicle in the grid area according to the shortest time T from the starting point to the edge node based on the traffic prediction information and the estimated distance D from the edge node to the end point based on the static traffic information.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011252260.XA CN112629533B (en) | 2020-11-11 | 2020-11-11 | Fine path planning method based on road network rasterization road traffic prediction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011252260.XA CN112629533B (en) | 2020-11-11 | 2020-11-11 | Fine path planning method based on road network rasterization road traffic prediction |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112629533A CN112629533A (en) | 2021-04-09 |
| CN112629533B true CN112629533B (en) | 2023-07-25 |
Family
ID=75303051
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011252260.XA Active CN112629533B (en) | 2020-11-11 | 2020-11-11 | Fine path planning method based on road network rasterization road traffic prediction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112629533B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113449402B (en) * | 2021-06-22 | 2022-08-05 | 武汉大学 | Road network efficiency gain prediction method after broken road is opened |
| CN113651061B (en) * | 2021-08-18 | 2023-04-14 | 上海德启信息科技有限公司 | Monitoring method and monitoring system for logistics circulation state |
| CN113848897B (en) * | 2021-09-17 | 2023-12-12 | 宁波大学 | Method, system and related products for path planning for unmanned surface vessels |
| CN114463977B (en) * | 2022-02-10 | 2023-06-23 | 北京工业大学 | Path planning method based on vehicle-road cooperation multi-source data fusion traffic flow prediction |
| CN115662114A (en) * | 2022-10-08 | 2023-01-31 | 广州玩鑫信息科技有限公司 | Intelligent traffic system for relieving congestion based on big data and operation method thereof |
| CN115630988B (en) * | 2022-12-22 | 2023-10-27 | 北京大学深圳研究生院 | Method and device for measuring and calculating land comprehensive traffic accessibility |
| CN116132998B (en) * | 2023-03-30 | 2023-07-25 | 江西师范大学 | Urban edge server deployment method based on intersection centrality |
Citations (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104567906A (en) * | 2015-01-14 | 2015-04-29 | 合肥革绿信息科技有限公司 | Beidou-based urban road network vehicle path planning method and device |
| CN105258704A (en) * | 2014-06-16 | 2016-01-20 | 中国科学院沈阳自动化研究所 | Multi-scale space-time hot point path detection method based on rapid road network modeling |
| CN107084732A (en) * | 2011-02-03 | 2017-08-22 | 通腾发展德国公司 | The method for producing expected average gait of march |
| CN108106623A (en) * | 2017-09-08 | 2018-06-01 | 同济大学 | A kind of unmanned vehicle paths planning method based on flow field |
| CN108877268A (en) * | 2018-08-07 | 2018-11-23 | 南京大学 | One kind is towards unpiloted no traffic lights crossroad intelligent dispatching method |
| WO2019048070A1 (en) * | 2017-09-06 | 2019-03-14 | Swiss Reinsurance Company Ltd. | Electronic system for dynamic, quasi-realtime measuring and identifying driver maneuvers solely based on mobile phone telemetry, and a corresponding method thereof |
| CN109959388A (en) * | 2019-04-09 | 2019-07-02 | 南京大学 | A kind of intelligent transportation fining paths planning method based on grid extended model |
| CN110009906A (en) * | 2019-03-25 | 2019-07-12 | 上海交通大学 | Dynamic path planning method based on traffic forecast |
| CN110046742A (en) * | 2019-02-28 | 2019-07-23 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Perceive planing method of the fluid on feasible path |
| CN110111575A (en) * | 2019-05-16 | 2019-08-09 | 北京航空航天大学 | A kind of Forecast of Urban Traffic Flow network analysis method based on Complex Networks Theory |
| CN110299005A (en) * | 2019-06-10 | 2019-10-01 | 浙江大学 | A kind of city large-scale road network traffic speed prediction technique based on Deep integrating study |
| CN110691406A (en) * | 2019-10-10 | 2020-01-14 | 南京大学 | D2D frequency spectrum efficient sharing method for internet-connected unmanned safe communication |
| CN111197991A (en) * | 2020-01-15 | 2020-05-26 | 西安电子科技大学 | Method for predicting optimal driving path of vehicle based on deep neural network |
| CN111223301A (en) * | 2020-03-11 | 2020-06-02 | 北京理工大学 | Traffic flow prediction method based on graph attention convolution network |
| CN111337044A (en) * | 2020-03-24 | 2020-06-26 | 北京交通发展研究院 | Urban road path planning method based on traffic weight |
| CN111811525A (en) * | 2020-06-09 | 2020-10-23 | 广东国地规划科技股份有限公司 | Road network generation method and system based on remote sensing image and floating car track |
| JP2020180870A (en) * | 2019-04-25 | 2020-11-05 | 山内 和博 | Automobile and automobile driving support system |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10612930B2 (en) * | 2016-07-26 | 2020-04-07 | Apple Inc. | Normalization of device probe data for path identification |
| US20190004524A1 (en) * | 2016-08-31 | 2019-01-03 | Faraday&Future Inc. | System and method for planning a vehicle path |
| US11755018B2 (en) * | 2018-11-16 | 2023-09-12 | Uatc, Llc | End-to-end interpretable motion planner for autonomous vehicles |
-
2020
- 2020-11-11 CN CN202011252260.XA patent/CN112629533B/en active Active
Patent Citations (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107084732A (en) * | 2011-02-03 | 2017-08-22 | 通腾发展德国公司 | The method for producing expected average gait of march |
| CN105258704A (en) * | 2014-06-16 | 2016-01-20 | 中国科学院沈阳自动化研究所 | Multi-scale space-time hot point path detection method based on rapid road network modeling |
| CN104567906A (en) * | 2015-01-14 | 2015-04-29 | 合肥革绿信息科技有限公司 | Beidou-based urban road network vehicle path planning method and device |
| WO2019048070A1 (en) * | 2017-09-06 | 2019-03-14 | Swiss Reinsurance Company Ltd. | Electronic system for dynamic, quasi-realtime measuring and identifying driver maneuvers solely based on mobile phone telemetry, and a corresponding method thereof |
| CN108106623A (en) * | 2017-09-08 | 2018-06-01 | 同济大学 | A kind of unmanned vehicle paths planning method based on flow field |
| CN108877268A (en) * | 2018-08-07 | 2018-11-23 | 南京大学 | One kind is towards unpiloted no traffic lights crossroad intelligent dispatching method |
| CN110046742A (en) * | 2019-02-28 | 2019-07-23 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Perceive planing method of the fluid on feasible path |
| CN110009906A (en) * | 2019-03-25 | 2019-07-12 | 上海交通大学 | Dynamic path planning method based on traffic forecast |
| CN109959388A (en) * | 2019-04-09 | 2019-07-02 | 南京大学 | A kind of intelligent transportation fining paths planning method based on grid extended model |
| JP2020180870A (en) * | 2019-04-25 | 2020-11-05 | 山内 和博 | Automobile and automobile driving support system |
| CN110111575A (en) * | 2019-05-16 | 2019-08-09 | 北京航空航天大学 | A kind of Forecast of Urban Traffic Flow network analysis method based on Complex Networks Theory |
| CN110299005A (en) * | 2019-06-10 | 2019-10-01 | 浙江大学 | A kind of city large-scale road network traffic speed prediction technique based on Deep integrating study |
| CN110691406A (en) * | 2019-10-10 | 2020-01-14 | 南京大学 | D2D frequency spectrum efficient sharing method for internet-connected unmanned safe communication |
| CN111197991A (en) * | 2020-01-15 | 2020-05-26 | 西安电子科技大学 | Method for predicting optimal driving path of vehicle based on deep neural network |
| CN111223301A (en) * | 2020-03-11 | 2020-06-02 | 北京理工大学 | Traffic flow prediction method based on graph attention convolution network |
| CN111337044A (en) * | 2020-03-24 | 2020-06-26 | 北京交通发展研究院 | Urban road path planning method based on traffic weight |
| CN111811525A (en) * | 2020-06-09 | 2020-10-23 | 广东国地规划科技股份有限公司 | Road network generation method and system based on remote sensing image and floating car track |
Non-Patent Citations (7)
| Title |
|---|
| Big data driven vehicular networks;Cheng, N., Lyu, F., Chen, J., Xu, W., Zhou, H., Zhang, S., & Shen, X;《IEEE Network》;第32卷(第6期);160-167 * |
| Deep learning sensor fusion for autonomous vehicle perception and localization: A review;Fayyad, J., Jaradat, M. A., Gruyer, D., & Najjaran, H;《Sensors》;第20卷(第15期);1-4 * |
| Toward collision-free and efficient coordination for automated vehicles at unsignalized intersection;Qian B, Zhou H, Lyu F, et al;《IEEE Internet of Things Journal》;第6卷(第6期);10408-10420 * |
| 城市道路未来车速预测模型研究;杨马英;楼挺;李一飞;;《小型微型计算机系统》(第12期);190-194 * |
| 基于典型经验路径库的路径规划算法;谢海莹;;《交通运输研究》;第2卷(第01期);17-19 * |
| 空天地一体化网络技术：探索与展望;焦椤方,许云霆,张天祺等;《物联网学报》;第4卷(第3期);3-6 * |
| 车辆导航系统中基于街区分块的分层路网路径规划;张照生;杨殿阁;张德鑫;连小珉;;《中国机械工程》(第23期);3255-3257 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112629533A (en) | 2021-04-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112629533B (en) | Fine path planning method based on road network rasterization road traffic prediction | |
| CN107944605B (en) | Dynamic traffic path planning method based on data prediction | |
| CN109670277B (en) | Travel time prediction method based on multi-mode data fusion and multi-model integration | |
| Lin et al. | Efficient network-wide model-based predictive control for urban traffic networks | |
| CN109959388B (en) | Intelligent traffic refined path planning method based on grid expansion model | |
| CN103996289B (en) | A kind of flow-speeds match model and Travel Time Estimation Method and system | |
| CN101639978B (en) | Method capable of dynamically partitioning traffic control subregion | |
| CN103927890B (en) | A kind of Trunk Road Coordination signal control method based on dynamic O-D Matrix Estimation | |
| CN113947900B (en) | Intelligent network-connected expressway ramp cooperative control system | |
| CN113538910B (en) | Self-adaptive full-chain urban area network signal control optimization method | |
| CN111080018B (en) | Intelligent network-connected automobile speed prediction method based on road traffic environment | |
| CN109269516B (en) | Dynamic path induction method based on multi-target Sarsa learning | |
| CN113299107B (en) | Multi-target fusion intersection dynamic vehicle internet speed guiding control method | |
| CN105489000A (en) | Night-shift bus stop and path selection method | |
| CN108985506A (en) | Planning driving path recommended method, prediction technique, acquisition methods and its device | |
| CN112863182B (en) | Cross-modal data prediction method based on transfer learning | |
| CN110570672A (en) | regional traffic signal lamp control method based on graph neural network | |
| Kong et al. | Urban arterial traffic two-direction green wave intelligent coordination control technique and its application | |
| CN115409256B (en) | Route recommendation method for avoiding congestion area based on travel time prediction | |
| CN102436598A (en) | Optimum vehicle path mining method based on traffic image information | |
| CN112967493A (en) | Neural network-based prediction method for vehicle passing intersection travel time | |
| Kanoh et al. | Short-term traffic prediction using fuzzy c-means and cellular automata in a wide-area road network | |
| CN114694382B (en) | Dynamic one-way traffic control system based on Internet of vehicles environment | |
| CN113763713B (en) | Road network bottleneck road section excavation and management method and device | |
| Wang et al. | Real-time urban regional route planning model for connected vehicles based on V2X communication |
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 |