CN109489679B

CN109489679B - Arrival time calculation method in navigation path

Info

Publication number: CN109489679B
Application number: CN201811549331.5A
Authority: CN
Inventors: 成佳颖; 王圣齐
Original assignee: Individual
Current assignee: Individual
Priority date: 2018-12-18
Filing date: 2018-12-18
Publication date: 2019-12-17
Anticipated expiration: 2038-12-18
Also published as: CN109489679A

Abstract

The invention discloses a method for calculating arrival time in a navigation path. The method has the advantages that the mass data in the history record are utilized, the speed limit of the road, the number of vehicles, the number of lanes and the speed data of the user are calculated and trained through a machine self-learning mode, the speed of the user, the red light time, the green light time, the number of queued vehicles and the corresponding passing time data are calculated, the interaction among the factors is comprehensively considered, the combined action of the passing time is analyzed, the calculation of the arrival time is optimized, and the calculation precision is improved.

Description

Arrival time calculation method in navigation path

Technical Field

The invention belongs to the technical field of traffic navigation, and particularly relates to a method for calculating arrival time in a navigation path.

Background

with the development of modern society, self-driving traveling becomes an increasingly common phenomenon, and the rapid increase of the number of urban vehicles causes urban road congestion to frequently occur, directly causes a great amount of travel time waste and obviously reduces travel efficiency, and the derived problems of fuel consumption, air pollution, road rage and the like are solved. Seriously reducing the quality of life and causing economic and social problems.

With the rapid development of wireless communication and mobile computing technologies and the rapid development of network mobile terminals such as mobile phones and tablet computers, travel navigation technologies are rapidly developed, such as functions of travel route selection navigation and the like in terms of high school, Tencent and Baidu maps, and the travel navigation technologies become indispensable and beneficial tools for people to travel. In the navigation process, after the path is planned, the estimation of the navigation arrival time of each path is very important, and the method is an important reference basis for a user to select a reliable navigation path. The current planning and the determination of the remaining navigation time are obtained by calculating and summing the passing time of each road. The passing time is mainly comprehensively considered through road conditions such as road distance, user speed per hour, road speed limit, congestion conditions, the number of traffic lights and the like, and because the considered factors are quite a lot, the influence of each factor on the passing time is estimated, a certain error exists in the estimation of a single factor, and the sum of a plurality of factors can cause quite a large error, so that the calculation accuracy is influenced.

Among them, Artificial Neural Network (ANN) is a research hotspot in the field of Artificial intelligence. It abstracts the human brain neuron network from the information processing angle, establishes a certain simple model, and forms different networks by a large number of nodes (or called neurons) according to different connection modes. Each node represents a particular output function, called the excitation function. Each connection between two nodes represents a weighted value, called weight, for the signal passing through the connection. The output of the network is different according to the connection mode of the network, the weight value and the excitation function. Artificial neural networks have successfully solved many practical problems in the fields of pattern recognition, intelligent robotics, automatic control, predictive estimation, biology, medicine, economy, etc., and exhibit good intelligent characteristics. However, no report of application in the field of traffic navigation exists at present.

Disclosure of Invention

in order to solve the technical problem, the invention provides a method for calculating the arrival time in the navigation path.

the complete technical scheme of the invention comprises the following steps:

A method for calculating an arrival time in a navigation path is characterized by comprising the following steps:

(1) Aiming at the current position and the destination, the selected path is divided into road sections, the whole path is divided into different road sections 1-n according to nodes, and the nodes comprise: the traffic light, the lane number change position and one of the road section speed limit change positions, wherein the road section comprises a normal driving road section and a traffic light waiting road section.

(2) calculating the passing time of each road section:

For example, for the normal travel section 1, the passing time t is calculated as follows₁，

Wherein S is the length of road section 1, V₁The speed per hour is calculated for the user.

wherein the calculated speed per hour V of the user₁and calculating by using the trained artificial neural network model. The method specifically comprises the following steps: selecting a driving record in the historical record, extracting road speed limit, vehicle quantity, lane quantity and user speed per hour data of each road section aiming at a certain road section or a plurality of road sections, forming a database after extraction, filling in, smoothing and noise reduction treatment, establishing an artificial neural network system based on the database, inputting the road speed limit, vehicle quantity, lane quantity data and user speed per hour data under the condition, and adopting the artificial neural networkThe system trains the input data to optimize the established artificial neural network system, and utilizes the optimized artificial neural network to input the lane, vehicle quantity and speed limit data at the moment to carry out the user speed per hour V under the road condition₁And (4) calculating.

For example, for the traffic light passing time of the traffic light section 2, the red light time, the green light time, the number of queued vehicles and the corresponding passing time data of each traffic light can be extracted for a certain traffic light or a plurality of traffic lights by selecting the driving record in the history record, a database is formed after extraction, filling and smooth noise reduction processing, an artificial neural network system is established based on the database, the red light time, the green light time, the number of queued vehicles and the corresponding passing time data of each traffic light are input, the input data are trained by the artificial neural network system to optimize the established artificial neural network system, the red light time, the green light time and the number of queued vehicles at the moment are input by the optimized artificial neural network, and the passing time t is carried out₂And (4) calculating.

(3) And (4) adding the passing time of each road section in the step (3) to obtain the arrival time in the navigation path.

And evaluating the fitting precision of the artificial neural network through an average square error (MSE) index. The formula is as follows:

Wherein N is₀Is the number of sets of output data, Q is the number of sets of training data, d is experimental data, and y is neural network output data. The goal of the training is relevance>9, MSE less than 0.001.

compared with the prior art, the invention has the advantages that: the method has the advantages that the mass data in the history record are utilized, the speed limit of the road, the number of vehicles, the number of lanes and the speed data of the user are calculated and trained through a machine self-learning mode, the speed of the user, the red light time, the green light time, the number of queued vehicles and the corresponding passing time data are calculated, the interaction among the factors is comprehensively considered, the combined action of the passing time is analyzed, the calculation of the arrival time is optimized, and the calculation precision is improved.

Drawings

FIG. 1 is a flow chart of the disclosed method.

Detailed Description

The invention is further described with reference to the following figures and detailed description.

A method for calculating arrival time in a navigation path comprises the following steps:

(1) aiming at the current position and the destination, the selected path is divided into road sections, the whole path is divided into different road sections 1-n according to nodes, and the nodes comprise: one of traffic lights, lane number change positions, and road section speed limit change positions.

Table 1 road segment division example

As shown in the above table, the whole route is divided into five sections, wherein the number of lanes of the section 1 is 4, the speed limit is 80Km/h, and the section 1 is isolated from the non-motor lane at the roadside by a rail or an isolation belt, which is indicated by 1. Road segment 2 is a traffic light. The number of lanes in the road section 3 is 3, the speed limit is 60Km/h, and the lanes are isolated from non-motor lanes beside roads by railings or isolation belts. And the road section 4 is a traffic indicator lamp. The number of lanes in the road section 5 is 2, the speed limit is 40Km/h, and the road section is isolated from non-motor lanes at the roadside without railings or isolation belts, and is represented by 0.

(2) And calculating the passing time of each road section, and adding the passing times to obtain the overall passing time.

Wherein for the road section 1, the passing time t is calculated in the following way₁，

Wherein the calculated speed per hour V of the user₁And calculating by using the trained artificial neural network model. Because the driving habits of all users are different, some users are used to drive at a high speed, the speed per hour is basically close to the speed limit of the road, and some users are more stable and drive at the speed per hour which is obviously lower than the speed limit of the road. Therefore, it is important to determine the basic driving habits of the users, especially for determining the passing time. Meanwhile, the influence of the number of vehicles running on the current road section on the passing time is very obvious, the influence on the speed per hour of the user is not great under the condition that the number of the vehicles is less, but when the number of the vehicles exceeds a certain threshold value, the driving habits of the user can be obviously influenced, for example, some users can adopt a more conservative driving mode when the number of the vehicles is increased, and meanwhile, the responses of different users to the number of the vehicles are different. However, under the premise of the same number of vehicles, the number of lanes set on the current road section has the same obvious influence on the passing time, and then the driving habits of the user are also influenced. Therefore, the influence of the driving habits, the number of lanes and the number of vehicles on the speed of the user is combined, and the three factors are also mutually influenced, so that the passing time calculation is carried out in a single estimation mode and the time is increased. However, no effective method for the combined action of the three interactive influence factors on the passing time exists at present.

Because a large amount of user travel history records are collected in the using process of the navigation tool, the invention adopts an artificial neural network model to train through the collected big data, and further predicts the passing time of the user.

Selecting a driving record in a user history record, selecting a plurality of road sections, extracting road speed limit, vehicle quantity, lane quantity and user speed per hour data of each road section, forming a database comprising 1000 records after extraction, filling and smooth noise reduction, establishing an artificial neural network system based on the database, inputting the road speed limit, vehicle quantity, lane quantity data and user speed per hour data under the condition, wherein 800 pieces of data are used as training samples, and the artificial neural network system is used for training the input data to establish the artificial neural networkOptimizing by a network system, adopting 200 data to verify samples, inputting the lane, the number of vehicles and the speed limit data at the moment by utilizing the optimized artificial neural network, and carrying out the speed per hour V of the user under certain road conditions₁And (6) predicting.

For the traffic light passing time of the road section 2, the red light time, the green light time, the number of queued vehicles and corresponding passing time data of each traffic light can be extracted for a certain traffic light or a plurality of traffic lights by selecting the driving record in the history record, a database is formed after extraction, filling and smooth noise reduction treatment, the data format is shown in table 2, an artificial neural network system is established based on the database, the red light time, the green light time, the number of queued vehicles and corresponding passing time data of each traffic light are input, the input data are trained by the artificial neural network system to optimize the established artificial neural network system, the red light time, the green light time and the number of queued vehicles at the moment are input by the optimized artificial neural network, and the passing time t is carried out₂And (4) calculating.

TABLE 2 intersection data Format

Crossing	Time red light(s)	Green light time(s)	Number of vehicles in line	Transit time(s)
					1	135	45	10	27
2	45	45	20	124
					3	120	20	5	65
4	45	45	11	20

Further, the fitting precision of the artificial neural network is evaluated through an average square error (MSE) index. The formula is as follows:

wherein N is₀Is the number of sets of output data, Q is the number of sets of training data, d is experimental data, and y is neural network output data. The goal of the training is relevance>9, MSE less than 0.001. According to research, the number of the hidden layer neurons is changed from 3 to 15 in sequence, the network is trained, when the number of the hidden layer neurons is found to be 8, the MSE of the artificial neural network is at the lowest point, the correlation coefficient (Regression) is at the high point, and the network meets the training target and has the best fitting precision.

The calculated time for the remaining road segments is similar to that described above.

And calculating the passing time of each road section, and adding the passing times to obtain the overall arrival time.

the above description is only a preferred embodiment of the present invention, and is not intended to limit the present invention, and all simple modifications, changes and equivalent structural changes made to the above embodiment according to the technical spirit of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A method for calculating an arrival time in a navigation path is characterized by comprising the following steps:

(1) Aiming at the current position and the destination, carrying out road section division on the selected path, and dividing the whole path into different road sections according to nodes, wherein the nodes comprise traffic lights, lane number change positions and road section speed limit change positions, and the road sections comprise normal driving road sections and traffic light waiting road sections;

(2) Calculating the passing time of each road section:

For a normal driving section, the passing time t is calculated in the following way₁，

Wherein S is the road length, V₁Calculating a speed per hour for the user;

Wherein the calculated speed per hour V of the user₁calculating by using the artificial neural network system after training optimization; the method specifically comprises the following steps: selecting a driving record in a trip history record of a user, extracting road speed limit, vehicle quantity, lane quantity and user speed per hour data of each road section aiming at a plurality of road sections, forming a database after extraction, filling and smooth noise reduction processing, and establishing an artificial neural network system based on the database; inputting road speed limit, vehicle quantity, lane quantity data of each road section and corresponding user speed per hour data, and training the input data by adopting an artificial neural network system so as to optimize the established artificial neural network system; using an optimized artificial neural network systemThe data of the road speed limit, the vehicle quantity and the lane quantity at the moment are input to obtain the calculated speed per hour V of the user under the road condition₁；

For the passing time of the traffic light waiting road section, extracting the red light time, the green light time, the number of queued vehicles and corresponding passing time data of each traffic light aiming at a plurality of traffic lights by selecting the driving records in the historical records, forming a database after extraction, filling up and smooth noise reduction treatment, and establishing an artificial neural network system based on the database; inputting red light time and green light time of each traffic light, the number of queued vehicles and corresponding passing time data, and training the input data by adopting an artificial neural network system so as to optimize the established artificial neural network system; inputting the red light time, the green light time and the vehicle number data of queuing by using the optimized artificial neural network system, and carrying out the passing time t₂Calculating;

(3) Adding the passing time of each road section in the step (2) to obtain the arrival time in the navigation path;

In the step (2), evaluating the fitting precision of the artificial neural network system through an average square error (MSE) index; the formula is as follows:

Wherein N is₀is the number of groups of output data, Q is the number of groups of training data, d is experimental data, and y is neural network output data; the goal of the training is relevance>9, MSE less than 0.001.