CN110610268A - Trajectory prediction method for unmanned driving - Google Patents

Trajectory prediction method for unmanned driving Download PDF

Info

Publication number
CN110610268A
CN110610268A CN201910858029.6A CN201910858029A CN110610268A CN 110610268 A CN110610268 A CN 110610268A CN 201910858029 A CN201910858029 A CN 201910858029A CN 110610268 A CN110610268 A CN 110610268A
Authority
CN
China
Prior art keywords
intersection
state transition
transition matrix
maximum value
track
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.)
Pending
Application number
CN201910858029.6A
Other languages
Chinese (zh)
Inventor
高岭
王泽天
高全力
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xian Polytechnic University
Original Assignee
Xian Polytechnic University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Xian Polytechnic University filed Critical Xian Polytechnic University
Priority to CN201910858029.6A priority Critical patent/CN110610268A/en
Publication of CN110610268A publication Critical patent/CN110610268A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/29—Graphical models, e.g. Bayesian networks
    • G06F18/295—Markov models or related models, e.g. semi-Markov models; Markov random fields; Networks embedding Markov models
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G06Q50/40—

Abstract

A track prediction method for unmanned driving is characterized in that track prediction is carried out on historical track data by using a Markov algorithm, unreasonable prediction results are found out by comparing predicted values with real values, and partial prediction results are corrected according to specific road condition information in the driving process of a vehicle.

Description

Trajectory prediction method for unmanned driving
Technical Field
The patent belongs to the technical field of track prediction in urban traffic, and particularly relates to a track prediction method for unmanned driving.
Background
With the rapid development of the internet and the rapid advance of technology, the automobile field has been intelligentized for hundreds of years and has also come to an unprecedented development stage. Security technologies are also becoming mature, such as: electronic stability systems, active braking systems, lane departure and hold systems, and the like. However, traffic accidents caused by misoperation of drivers still account for more than 93%. Moreover, the reaction time and subjective thinking of the driver are existed after all, which causes that the optimal decision is difficult to be made when the danger occurs, thereby becoming the main factor of the accident.
Therefore, the development of unmanned driving technology is imperative to reduce the occurrence of traffic accidents and realize zero casualties of road traffic. By combining the intelligent network connection with the unmanned automobile, a huge intelligent traffic network is formed to exert intellectualization to the maximum extent so as to improve the road traffic capacity, save energy, reduce congestion and improve the living environment.
And predicting the position information of the vehicle in the next period by mining the historical track data. According to the prediction information, the vehicle can make a decision in advance, and the mode of changing the travel time or the travel route is changed to avoid the traffic jam which possibly occurs, so that the problem of urban traffic jam is relieved to the maximum extent.
Content of patent
In order to overcome the defects of the prior art, the patent aims to provide a track prediction method for unmanned driving, wherein probability distribution is obtained through a Markov algorithm according to historical track data of a vehicle, and a driving route of the vehicle is predicted; and correcting the probability distribution according to the scene information of vehicle running, and increasing the prediction accuracy.
In order to achieve the purpose, the technical scheme adopted by the patent is as follows:
a trajectory prediction algorithm for unmanned driving, comprising the steps of:
step 1, acquiring an operation instruction for triggering a Markov algorithm;
step 2, collecting historical track data of the vehicle and scene information of the vehicle, and generating a data source required by a prediction algorithm;
step 3, calculating the state transition probability by using a Markov algorithm, and establishing a state transition matrix so as to obtain a predicted track;
step 4, comparing the predicted track of the vehicle obtained in the step 3 with the actual vehicle track, and if an error exists, improving a Markov algorithm, correcting an unreasonable prediction result and improving the track prediction accuracy; it was found that only 65% of the prediction accuracy could be achieved; because the Markov algorithm selects the maximum value every time, when the difference between the maximum value and the second large value is too small, the maximum value intersection is selected to be accidental compared with the second large value intersection in the actual driving process;
the step 3 is specifically carried out: setting n intersections of a road, wherein the Markov state transition matrix is an n-n matrix, and the ith row and the jth column elements P in the one-step state transition matrix PijThe probability that the intersection passes through an intersection j after passing through an intersection i is represented, i is more than or equal to 1, j is less than or equal to n, and the calculation formula is shown as the formula (1):
wherein N isijThe method is characterized in that the number of times that the historical track data firstly passes through the intersection i and then passes through the intersection j is shown, i and j must be adjacent intersections, a state transition matrix P is obtained in one step, a state transition matrix P in k steps is further obtained, and the state transition matrix is expressed as a formula (2):
P(k)＝Pk (2)
the one-step state transition matrix P obtained according to equation (1) is as shown in equation (3):
the row number in the matrix represents the current known track intersection formed by historical data with the total number of n intersections, and the column number represents the next intersection to pass through in the future; scanning the matrix according to the current position to obtain a row number, and selecting a row of data corresponding to the row number for comparison; taking the train number with the maximum probability as the next future intersection, then substituting the intersection as the current position into the state transition matrix again for iterative solution, and finally obtaining a series of future vehicle tracks;
k-step state transition matrix P obtained according to formula (2)(k)As shown in formula (4):
in the formula, please explain P each parameterijIs the probability of passing through intersection j after passing through intersection i, NijFor the number of times of passing through the intersection i first and then the intersection j in the historical track data, i and j must be adjacent intersections, P is a one-step state transition matrix, and P is(k)For a state transition moment of k stepsArray, Pk ijThe probability of passing through the intersection i after k steps and then passing through the intersection j is represented by a row number in the matrix, wherein the row number represents a current known track intersection formed by historical data with the total number of n intersections, and a column number represents the next intersection to pass through in the future;
the improved Markov algorithm in step 4 is improved by setting a threshold value delta between the maximum value and the second maximum value in the state transition matrix;
if the difference between the two values is larger than delta, the next intersection where the vehicle passes in the future is considered to be equal to the maximum value of the state transition matrix; when the traffic congestion is smaller than the delta, different methods are adopted according to whether the traffic congestion exists at the intersection with the maximum value; two different situations can be distinguished:
1) if the traffic at the maximum intersection is not congested, the patent considers that the next intersection through which the vehicle passes is still equal to the maximum value in the state transition matrix.
2) If the traffic jam occurs at the intersection with the maximum value; the vehicle will select the second largest intersection in the state transition matrix.
Compared with the prior art, the invention has the beneficial effects that:
the improved Markov prediction algorithm is provided by analyzing the actual driving condition. During the track prediction, the contingency problem of the Markov algorithm during the selection of the prediction result is solved, the specific road condition information in the vehicle driving process is effectively fused, and the accuracy of the track prediction is improved.
Drawings
FIG. 1 is a flow chart of the present invention.
Detailed Description
The invention is further illustrated by the following examples in conjunction with the accompanying drawings:
the core of the invention is: by comparing the prediction result obtained by the Markov algorithm with the actual result and combining specific road condition information, some unreasonable predictions are found in the prediction process, partial prediction results are corrected, and the prediction accuracy is further improved.
The specific algorithm flow is as follows:
referring to fig. 1, a trajectory prediction method for unmanned driving includes the steps of:
step 1, acquiring an operation instruction for triggering a Markov algorithm;
step 2, collecting historical track data of the vehicle and scene information of the vehicle, and generating a data source required by a prediction algorithm;
step 3, calculating the state transition probability by using a Markov algorithm, and establishing a state transition matrix so as to obtain a predicted track; the specific method comprises the following steps:
setting n intersections of a road, wherein the Markov state transition matrix is an n-n matrix, and the ith row and the jth column elements P in the one-step state transition matrix PijThe probability that the road passes through a road junction i (i is more than or equal to 1) and then passes through a road junction j (j is less than or equal to n) is represented, and the calculation formula is as shown in the formula (1):
wherein N isijThe number of times that the historical track data firstly passes through the intersection i and then passes through the intersection j is shown, i and j must be adjacent intersections, and therefore a one-step state transition matrix P is obtained, and a k-step state transition matrix is further obtained, and is expressed as a formula (2):
P(k)＝Pk (2)
the one-step state transition matrix P obtained according to equation (1) is as shown in equation (3):
the row number in the matrix represents the current known track intersection formed by historical data with the total number of n intersections, and the column number represents the next intersection to pass through in the future; scanning the matrix according to the current position to obtain a row number, and selecting a row of data corresponding to the row number for comparison; taking the train number with the maximum probability as the next future intersection, then substituting the intersection as the current position into the state transition matrix again for iterative solution, and finally obtaining a series of future vehicle tracks;
k-step state transition matrix obtained according to formula (2)P(k)As shown in formula (4):
and 4, comparing the predicted track of the vehicle obtained in the step 3 with the actual track of the vehicle, and finding that the prediction precision can only reach 65%. Since the markov algorithm selects the maximum value each time, when the difference between the maximum value and the second maximum value is too small, it is accidental to select the maximum value intersection in actual driving compared with the second maximum value intersection.
The present embodiment solves this problem by setting a threshold δ in the state transition matrix between the maximum value and the second largest value. If the difference between the two values is greater than δ, the next intersection through which the vehicle will pass in the future is considered to be equal to the maximum value of the state transition matrix. And when the traffic congestion is smaller than the delta, different methods are adopted according to whether the intersection with the maximum value is congested or not. Two different situations can be distinguished: 1) when the traffic is not congested at the maximum intersection; 2) when the traffic is congested at the maximum intersection.
1) If the traffic is not congested, the patent believes that the next intersection through which the vehicle passes is still equal to the maximum value in the state transition matrix.
2) If the traffic is congested, the vehicle will select the intersection with the second largest value in the travel state transition matrix.

Claims (4)

1. A trajectory prediction method for unmanned driving, comprising the steps of:
step 1, acquiring an operation instruction for triggering a Markov algorithm;
step 2, collecting historical track data of the vehicle and scene information of the vehicle, and generating a data source required by a prediction algorithm;
step 3, calculating the state transition probability by using a Markov algorithm, and establishing a state transition matrix to obtain a predicted track;
and 4, comparing the predicted track obtained in the step 3 with the actual running track, and if an error exists, improving the Markov algorithm, correcting an unreasonable prediction result and improving the track prediction accuracy.
2. The trajectory prediction method for unmanned driving according to claim 1, wherein the step 3 is implemented by:
setting n intersections of a road, wherein the Markov state transition matrix is an n-n matrix, and the ith row and the jth column elements P in the one-step state transition matrix PijThe probability that the intersection passes through an intersection j after passing through an intersection i is represented, i is more than or equal to 1, j is less than or equal to n, and the calculation formula is shown as the formula (1):
wherein N isijThe method is characterized in that the number of times that the historical track data firstly passes through the intersection i and then passes through the intersection j is shown, i and j must be adjacent intersections, a state transition matrix P is obtained in one step, a state transition matrix P in k steps is further obtained, and the state transition matrix is expressed as a formula (2):
P(k)＝Pk。 (2)
the one-step state transition matrix P obtained according to equation (1) is as shown in equation (3):
the row number in the matrix represents the current known track intersection formed by historical data with the total number of n intersections, and the column number represents the next intersection to pass through in the future; scanning the matrix according to the current position to obtain a row number, and selecting a row of data corresponding to the row number for comparison; taking the train number with the maximum probability as the next future intersection, then substituting the intersection as the current position into the state transition matrix again for iterative solution, and finally obtaining a series of future vehicle tracks;
k-step state transition matrix P obtained according to formula (2)(k)As shown in formula (4):
in the formula, please explain P each parameterijIs the probability of passing through intersection j after passing through intersection i, NijFor the number of times of passing through the intersection i first and then the intersection j in the historical track data, i and j must be adjacent intersections, P is a one-step state transition matrix, and P is(k)For a k-step state transition matrix, Pk ijIn order to obtain the probability of passing through the intersection j after passing through k steps, the row number in the matrix represents the current known track intersection formed by the historical data with the total number of n intersections, and the column number represents the next intersection to pass through in the future.
3. The method of claim 1, wherein the predicted trajectory obtained from step 3 in step 4 has an error from the actual driving trajectory because the Markov algorithm selects the maximum value each time, but when the difference between the maximum value and the second maximum value is too small, the maximum intersection is accidentally selected during the actual driving process compared with the second maximum intersection.
4. A trajectory prediction method for unmanned vehicles according to claim 1, wherein the improved markov algorithm of step 4 is improved by setting a threshold δ in the state transition matrix between a maximum value and a second maximum value;
if the difference between the two values is larger than delta, the next intersection where the vehicle passes in the future is considered to be equal to the maximum value of the state transition matrix; when the traffic congestion is smaller than the delta, different methods are adopted according to whether the traffic congestion exists at the intersection with the maximum value; two different situations can be distinguished:
1) if the traffic at the maximum intersection is not congested, the patent considers that the next intersection through which the vehicle passes is still equal to the maximum value in the state transition matrix.
2) If the traffic jam occurs at the intersection with the maximum value; the vehicle will select the second largest intersection in the state transition matrix.
CN201910858029.6A 2019-09-11 2019-09-11 Trajectory prediction method for unmanned driving Pending CN110610268A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910858029.6A CN110610268A (en) 2019-09-11 2019-09-11 Trajectory prediction method for unmanned driving

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910858029.6A CN110610268A (en) 2019-09-11 2019-09-11 Trajectory prediction method for unmanned driving

Publications (1)

Publication Number Publication Date
CN110610268A true CN110610268A (en) 2019-12-24

Family

ID=68892731

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910858029.6A Pending CN110610268A (en) 2019-09-11 2019-09-11 Trajectory prediction method for unmanned driving

Country Status (1)

Country Link
CN (1) CN110610268A (en)

Citations (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090306881A1 (en) * 2008-06-06 2009-12-10 Toyota Motor Engineering & Manufacturing North America, Inc. Detecting principal directions of unknown environments
US20120143488A1 (en) * 2009-08-31 2012-06-07 Toyota Motor Europe Nv/Sa Vehicle or traffic control method and system
US20130024060A1 (en) * 2011-07-21 2013-01-24 Saturna Green Systems Inc. Vehicle communication, analysis and operation system
US20130226365A1 (en) * 2012-02-23 2013-08-29 Ford Global Technologies, Llc Vehicle drive matching system and method
US20140114556A1 (en) * 2012-10-23 2014-04-24 University Of Southern California Traffic prediction using real-world transportation data
US8880326B1 (en) * 2013-02-20 2014-11-04 Google Inc. Methods and systems for determining a cyclical fleet plan satisfying a recurring set of coverage requirements
CN105424052A (en) * 2014-09-16 2016-03-23 福特全球技术公司 Stochastic range
CN105956704A (en) * 2016-05-04 2016-09-21 大连理工大学 Destination identification method for plug-in type hybrid vehicle
CN108022026A (en) * 2018-01-25 2018-05-11 戴姆勒股份公司 A kind of traffic security early warning method of traffic control, device and the system including the device
CN108346284A (en) * 2018-01-29 2018-07-31 河海大学 A kind of uncertain road network track of vehicle prediction technique based on Markov model
US10147020B1 (en) * 2017-07-28 2018-12-04 Cognism Limited System and method for computational disambiguation and prediction of dynamic hierarchical data structures
US20190051174A1 (en) * 2017-08-11 2019-02-14 Lyft, Inc. Travel path and location predictions
CN109752019A (en) * 2019-02-26 2019-05-14 西安工程大学 Optimum transportation route planing method based on Bayesian network
US20190222595A1 (en) * 2018-01-18 2019-07-18 General Electric Company Decision system and method for separating faults from attacks

Patent Citations (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090306881A1 (en) * 2008-06-06 2009-12-10 Toyota Motor Engineering & Manufacturing North America, Inc. Detecting principal directions of unknown environments
US20120143488A1 (en) * 2009-08-31 2012-06-07 Toyota Motor Europe Nv/Sa Vehicle or traffic control method and system
US20130024060A1 (en) * 2011-07-21 2013-01-24 Saturna Green Systems Inc. Vehicle communication, analysis and operation system
US20130226365A1 (en) * 2012-02-23 2013-08-29 Ford Global Technologies, Llc Vehicle drive matching system and method
US20140114556A1 (en) * 2012-10-23 2014-04-24 University Of Southern California Traffic prediction using real-world transportation data
US8880326B1 (en) * 2013-02-20 2014-11-04 Google Inc. Methods and systems for determining a cyclical fleet plan satisfying a recurring set of coverage requirements
CN105424052A (en) * 2014-09-16 2016-03-23 福特全球技术公司 Stochastic range
CN105956704A (en) * 2016-05-04 2016-09-21 大连理工大学 Destination identification method for plug-in type hybrid vehicle
US10147020B1 (en) * 2017-07-28 2018-12-04 Cognism Limited System and method for computational disambiguation and prediction of dynamic hierarchical data structures
US20190051174A1 (en) * 2017-08-11 2019-02-14 Lyft, Inc. Travel path and location predictions
US20190222595A1 (en) * 2018-01-18 2019-07-18 General Electric Company Decision system and method for separating faults from attacks
CN108022026A (en) * 2018-01-25 2018-05-11 戴姆勒股份公司 A kind of traffic security early warning method of traffic control, device and the system including the device
CN108346284A (en) * 2018-01-29 2018-07-31 河海大学 A kind of uncertain road network track of vehicle prediction technique based on Markov model
CN109752019A (en) * 2019-02-26 2019-05-14 西安工程大学 Optimum transportation route planing method based on Bayesian network

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
ANONYMOUS: "《IP.COM非专利全文库》", 28 March 2012, IP.COM *
彭曲等: "基于马尔可夫链的轨迹预测", 《计算机科学》 *
毛莺池等: "不确定性车辆路口的轨迹预测", 《计算机科学》 *

Similar Documents

Publication Publication Date Title
WO2021077725A1 (en) System and method for predicting motion state of surrounding vehicle based on driving intention
Murphey et al. Driver's style classification using jerk analysis
CN111267846B (en) Game theory-based peripheral vehicle interaction behavior prediction method
CN110843789A (en) Vehicle lane change intention prediction method based on time sequence convolution network
CN112614373B (en) BiLSTM-based weekly vehicle lane change intention prediction method
Li et al. Development and evaluation of two learning-based personalized driver models for pure pursuit path-tracking behaviors
CN112071059A (en) Intelligent vehicle track changing collaborative planning method based on instantaneous risk assessment
CN114212110B (en) Obstacle trajectory prediction method and device, electronic equipment and storage medium
Li et al. Modeling vehicle merging position selection behaviors based on a finite mixture of linear regression models
CN116259185B (en) Vehicle behavior decision method and device fusing prediction algorithm in parking lot scene
CN115662131B (en) Multi-lane collaborative lane changing method for road accident section in network environment
CN112758105B (en) Automatic driving fleet following running control method, device and system
CN110610268A (en) Trajectory prediction method for unmanned driving
CN111027859B (en) Driving risk prevention method and system based on motor vehicle state monitoring data mining
Dong et al. Determining E-bike drivers’ decision-making mechanisms during signal change interval using the hidden Markov driving model
CN115204755B (en) Service area access rate measuring method and device, electronic equipment and readable storage medium
CN111915888B (en) Method for calculating complexity of traffic participants in automatic driving test scene
CN115454082A (en) Vehicle obstacle avoidance method and system, computer readable storage medium and electronic device
CN111968397B (en) Automatic driving behavior determination method
CN113844446A (en) Long-short-range fused vehicle track prediction method
CN114267191A (en) Control system, method, medium, equipment and application for relieving traffic jam of driver
CN115071758B (en) Man-machine common driving control right switching method based on reinforcement learning
Liu et al. A Centralized Cooperative Lane-Changing Strategy for Merging into CAV Platoon
CN117408086B (en) Mixed traffic flow simulation method based on driver response time
CN114446053B (en) Uncontrolled intersection safety evaluation method for intelligent networked vehicle track error

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
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20191224