CN110400490A - Trajectory predictions method and apparatus - Google Patents

Trajectory predictions method and apparatus Download PDF

Info

Publication number
CN110400490A
CN110400490A CN201910729763.2A CN201910729763A CN110400490A CN 110400490 A CN110400490 A CN 110400490A CN 201910729763 A CN201910729763 A CN 201910729763A CN 110400490 A CN110400490 A CN 110400490A
Authority
CN
China
Prior art keywords
lane
target
information
target vehicle
prediction model
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN201910729763.2A
Other languages
Chinese (zh)
Other versions
CN110400490B (en
Inventor
钱祥隽
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.)
Tencent Technology Shenzhen Co Ltd
Original Assignee
Tencent Technology Shenzhen Co Ltd
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 Tencent Technology Shenzhen Co Ltd filed Critical Tencent Technology Shenzhen Co Ltd
Priority to CN201910729763.2A priority Critical patent/CN110400490B/en
Publication of CN110400490A publication Critical patent/CN110400490A/en
Application granted granted Critical
Publication of CN110400490B publication Critical patent/CN110400490B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/16—Anti-collision systems
    • G08G1/166—Anti-collision systems for active traffic, e.g. moving vehicles, pedestrians, bikes

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Traffic Control Systems (AREA)

Abstract

The embodiment of the invention discloses a kind of trajectory predictions method and apparatus；The available cartographic information of the embodiment of the present invention, the driving information of target vehicle, lane prediction model, motion prediction model；Driving information based on cartographic information and target vehicle determines the driving feature of the association lane and target vehicle of target vehicle relative to association lane；Using lane prediction model, according to the driving feature target lane that prediction target vehicle will drive into association lane；Using motion prediction model, according to motion information of the driving feature prediction target vehicle on target lane；Based on driving information, target lane and motion information, the motion profile of target vehicle is calculated.The motion information of the target lane and vehicle of vehicle on target lane is predicted by different prediction models in embodiments of the present invention, to carry out trajectory calculation.The program can promote the precision of prediction of trajectory predictions as a result,.

Description

Trajectory predictions method and apparatus
Technical field
The present invention relates to computer fields, and in particular to a kind of trajectory predictions method and apparatus.
Background technique
It is universal with Internet of Things, vehicle can by the auxiliary of computer technology come programme path, evade collision accident, Driving efficiency is improved simultaneously.It, can be with for example, car-mounted computer be in addition to that can tell the static-obstacle thing around vehicle The driving trace of other vehicles around prediction, and judge surrounding vehicles whether to itself vehicle according to the driving trace of surrounding vehicles Constitute potential security threat.
However, the prediction accuracy of the method for trajectory predictions is lower at present.
Summary of the invention
The embodiment of the present invention provides a kind of trajectory predictions method and apparatus, can promote the precision of prediction of trajectory predictions.
The embodiment of the present invention provides a kind of trajectory predictions method, comprising:
Obtain cartographic information, the driving information of target vehicle, lane prediction model, motion prediction model, wherein the vehicle Road prediction model, motion prediction model are formed by training sample training；
Driving information based on the cartographic information and target vehicle determines association lane and the mesh of target vehicle Mark driving feature of the vehicle relative to the association lane；
Using lane prediction model, predict that target vehicle will drive into the association lane according to the driving feature Target lane；
Using motion prediction model, believed according to movement of the driving feature prediction target vehicle on the target lane Breath；
Based on the driving information, target lane and motion information, the motion profile of target vehicle is calculated.
In some embodiments, the motion prediction model includes the first motion prediction model, the second motion prediction model, The motion information on the target lane includes the target speed information in predetermined time relative to the target lane With target range information, the target lane includes target lane center；
It is described to use motion prediction model, according to fortune of the driving feature prediction target vehicle on the target lane Dynamic information, comprising:
Using the first motion prediction model, predict target vehicle in predetermined time relative to described according to the driving feature The target speed information in target lane；
Using the second motion prediction model, predict target vehicle in predetermined time relative to described according to the driving feature The target range information of target lane center；
It is described to be based on the driving information, target lane and motion information, calculate the motion profile of target vehicle, comprising:
Based on the driving information, target lane, target speed information and target range information, the fortune of target vehicle is calculated Dynamic rail mark.
In some embodiments, the driving information includes initial position message, initial velocity information；
Based on the driving information, target lane, target speed information and target range information, the fortune of target vehicle is calculated Dynamic rail mark, comprising:
According to the target lane and target range information, determine target vehicle in the target position information of predetermined time；
Target carriage is calculated based on the initial position message, target position information, initial velocity information, target speed information Motion profile.
In some embodiments, described to use lane prediction model, according to the driving feature in the association lane The target lane that prediction target vehicle will drive into, comprising:
Using lane prediction model, the association vehicle is driven into predetermined time according to the driving feature calculation target vehicle Road drives into probability；
Probability is driven into according to described, target lane is determined from the association lane.
In some embodiments, the lane prediction model includes multiple lane prediction submodels, predicts mould using lane Type drives into the association lane in predetermined time according to the driving feature calculation target vehicle and drives into probability, comprising:
Submodel is predicted using lane, the association is driven into predetermined time according to the driving feature calculation target vehicle Sub- probability is driven into lane；
It drives into sub- probability to described and is weighted summation, obtain target vehicle in predetermined time and drive into the association lane Drive into probability.
In some embodiments, cartographic information, the driving information of target vehicle, lane prediction model, motion prediction are obtained Model, before, further includes:
Obtain training sample and initial predicted model, the corresponding multiple sample marks of each of described training sample；
Rejecting processing is carried out to the training sample corresponding sample mark, obtains lane training sample and training Sample；
Initial predicted model is trained using the lane training sample, obtains lane prediction model；
Initial predicted model is trained using the training sample, obtains motion prediction model.
In some embodiments, the sample mark includes lane mark, distance mark, speed mark, the movement instruction Practicing sample includes apart from training sample and speed training sample；
Rejecting processing is carried out to the training sample corresponding sample mark, obtains lane training sample and training Sample, comprising:
The distance mark and speed mark for abandoning the training sample are only remained the lane training sample of lane mark This；
It abandons the lane mark of the training sample and apart from mark, is only remained the speed training sample of speed mark This；
The lane mark and speed mark for abandoning the training sample are only remained the distance training sample of distance mark This.
In some embodiments, the training sample includes speed training sample, apart from training sample, the speed Training sample includes speed training subsample, described to include distance training subsample, the motion prediction mould apart from training sample Type includes speed prediction model, range prediction model；
Initial predicted model is trained using the training sample, obtains motion prediction model, comprising:
Initial predicted model is trained using the speed training subsample, obtains speed prediction model；
Initial predicted model is trained using distance training subsample, obtains range prediction model.
In some embodiments, the driving information based on the cartographic information and target vehicle, determines target vehicle It is associated with the driving feature of lane and target vehicle relative to the association lane, comprising:
According to the cartographic information and the driving information of target vehicle, the current lane of target vehicle is determined；
Topological analysis is carried out according to current lane of the cartographic information to target vehicle, is obtained and the current lane phase Associated association lane；
Driving information based on the association lane and target vehicle calculates target vehicle relative to the association lane Driving feature.
The embodiment of the present invention also provides a kind of trajectory predictions device, comprising:
Acquiring unit, for obtaining the driving information, lane prediction model, motion prediction mould of cartographic information, target vehicle Type, wherein the lane prediction model, motion prediction model are formed by training sample training；
Associative cell determines the pass of target vehicle for the driving information based on the cartographic information and target vehicle Join the driving feature of lane and target vehicle relative to the association lane；
Lane unit predicts mesh in the association lane according to the driving feature for using lane prediction model The target lane that mark vehicle will drive into；
Moving cell predicts target vehicle in the target for using motion prediction model according to the driving feature Motion information on lane；
Trajectory unit calculates the movement rail of target vehicle for being based on the driving information, target lane and motion information Mark.
The available cartographic information of the embodiment of the present invention, the driving information of target vehicle, lane prediction model, motion prediction Model, wherein lane prediction model, motion prediction model are formed by training sample training；Based on cartographic information and target carriage Driving information, determine target vehicle association lane and target vehicle relative to association lane driving feature；Using Lane prediction model, according to the driving feature target lane that prediction target vehicle will drive into association lane；Using movement Prediction model, according to motion information of the driving feature prediction target vehicle on target lane；Based on driving information, target lane And motion information, calculate the motion profile of target vehicle.
The target lane for predicting vehicle by different prediction models in embodiments of the present invention and vehicle are in target Motion information on lane, to carry out trajectory calculation.The program can promote the precision of prediction of trajectory predictions as a result,.
Detailed description of the invention
To describe the technical solutions in the embodiments of the present invention more clearly, make required in being described below to embodiment Attached drawing is briefly described, it should be apparent that, drawings in the following description are only some embodiments of the invention, for For those skilled in the art, without creative efforts, it can also be obtained according to these attached drawings other attached Figure.
Fig. 1 a is the schematic diagram of a scenario of trajectory predictions method provided in an embodiment of the present invention；
Fig. 1 b is the flow diagram of trajectory predictions method provided in an embodiment of the present invention；
Fig. 1 c is the map schematic diagram of a layer structure of high-precision map provided in an embodiment of the present invention；
Fig. 1 d is the relation schematic diagram of target vehicle and current lane provided in an embodiment of the present invention；
Fig. 1 e is the structural schematic diagram of Random Forest model provided in an embodiment of the present invention；
Fig. 1 f is the schematic illustration of moving track calculation provided in an embodiment of the present invention；
Fig. 2 is that the embodiment of the present invention provides the trajectory predictions method flow schematic diagram including model predictive process；
Fig. 3 a is the first structural schematic diagram of trajectory predictions device provided in an embodiment of the present invention；
Fig. 3 b is second of structural schematic diagram of trajectory predictions device provided in an embodiment of the present invention；
Fig. 3 c is the third structural schematic diagram of trajectory predictions device provided in an embodiment of the present invention；
Fig. 3 d is the 4th kind of structural schematic diagram of trajectory predictions device provided in an embodiment of the present invention；
Fig. 3 e is the 5th kind of structural schematic diagram of trajectory predictions device provided in an embodiment of the present invention；
Fig. 4 is the structural schematic diagram of the network equipment provided in an embodiment of the present invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on Embodiment in the present invention, those skilled in the art's every other implementation obtained without creative efforts Example, shall fall within the protection scope of the present invention.
The embodiment of the present invention provides a kind of trajectory predictions method and apparatus.
Wherein, which specifically can integrate in the electronic device, which can be terminal, service The equipment such as device.Wherein, terminal can for automatic pilot, smart phone, tablet computer, smart bluetooth equipment, laptop, The equipment such as PC (Personal Computer, PC)；Server can be single server, be also possible to by multiple clothes The server cluster of business device composition.
In some embodiments, which can also be integrated in multiple electronic equipments, for example, trajectory predictions Device can integrate in multiple servers, and trajectory predictions method of the invention is realized by multiple servers.
For example, the electronic equipment can be carried by automatic pilot and by vehicle, automatic pilot can be logical in network Letter, and the sensor by carrying on vehicle obtain the driving information of surrounding vehicles.For example, with reference to Fig. 1 a, trajectory predictions dress It sets and is integrated in automatic pilot, the automatic pilot that this vehicle is equipped with can obtain this vehicle week by the sensor of this vehicle Enclose the driving information of other vehicles.For example, cartographic information, lane prediction model, motion prediction model can be obtained by network, And the driving information of other vehicles of surrounding is obtained (for example, the traveling speed of other vehicles of surrounding by the sensing system of vehicle The distance between degree, type of vehicle and this vehicle, direction of traffic, etc.)；Then, automatic pilot can based on cartographic information with And the driving information of surrounding other vehicles, come the current association lane of other vehicles around determining, and surrounding other vehicle phases For being associated with the driving feature in lane；Again use lane prediction model, according to driving feature association lane in prediction around its The target lane that his vehicle will drive into, using motion prediction model, according to other vehicles around driving feature prediction in target Motion information on lane；It is finally based on driving information, target lane and motion information, calculates the movement rail of other vehicles of surrounding Mark.
It is described in detail separately below.It should be noted that the serial number of following embodiment is not as preferably suitable to embodiment The restriction of sequence.
In the present embodiment, a kind of trajectory predictions method is provided, as shown in Figure 1 b, the specific stream of the trajectory predictions method Journey can be such that
101, cartographic information, the driving information of target vehicle, lane prediction model, motion prediction model are obtained.
Wherein, cartographic information, which can be, depicts the spatial informations such as road in real world, traffic condition, administrative region Image information is also possible to the map datum of customized virtual world.Cartographic information can be used for ground traffic control, vehicle Navigation, vehicle driving route planning etc..
In some embodiments, cartographic information may include high-precision map, which may include static height Smart map layer and the high-precision map layer of dynamic.
Wherein, may include lane layer, road component layer, road attribute layer etc. in static high-precision map layer includes static letter The map layer of breath.Specifically, lane layer in may include lane specific information, as lane line, lane center, lane width, The information such as curvature, the gradient, course, lane rule.It may include the roads such as traffic mark board, pavement marker portion in road component layer Part, for example, recording traffic signal lamp exact position and height etc..
Wherein, the high-precision map layer of dynamic may include congestion in road layer, condition of construction layer, traffic accident layer, traffic control Layer, day gas-bearing formation etc. include the map layer of dynamic information.For example, being may include in condition of construction layer such as trimming, road markings The information such as line wears and repaints, traffic marking changes.
For example, as illustrated in figure 1 c, providing a kind of map schematic diagram of a layer structure of high-precision map, including static state is high-precisionly Figure layer and the high-precision map layer of dynamic, wherein including lane layer and road component layer, the high-precision map of dynamic in static high-precision map layer Layer includes a day gas-bearing formation.
In embodiments of the present invention, it can will be equipped with trajectory predictions arrangement vehicle and be known as this vehicle, target vehicle is except this Other vehicles other than vehicle, around this vehicle in certain distance.The distance can also be set by user setting by technical staff It sets, it can also be with this vehicle sensor sensing distance dependent.
The driving information of target vehicle refer to target vehicle in the process of moving can by the information detected by this vehicle, than Such as position location, travel speed, driving direction, vehicle license, type of vehicle.
Lane prediction model can be a kind of for predicting target vehicle in lane where future time instance (for example, after 3 seconds) Mathematical model, similar, motion prediction model is a kind of for predicting target vehicle in the prediction of future time instance motion information Mathematical model.
Wherein, the mode of cartographic information, the driving information of target vehicle, lane prediction model, motion prediction model is obtained Multiplicity, cartographic information, the driving information of target vehicle, lane prediction model, motion prediction model can obtain in the same way It takes, can also obtain in different ways.
For example, cartographic information, lane prediction model, motion prediction model can be read from local memory, and pass through biography The driving information of sensor system acquisition target vehicle.
For another example, lane prediction model and motion prediction model can be obtained by network, then is read from local memory Cartographic information, and pass through the driving information, etc. of sensing system acquisition target vehicle.
102, the driving information based on cartographic information and target vehicle determines association lane and the mesh of target vehicle Mark driving feature of the vehicle relative to association lane.
Lane as where vehicle is possible to keep the current time vehicle in future time instance, it is also possible to lane change to the left, Lane change to the right, or even connect change twice, therefore, before carrying out trajectory predictions, need the row based on cartographic information and target vehicle It sails information sifting and goes out the vehicle and be possible to the lane driven into future time instance, these lanes are denoted as association lane.
Association lane, which refers to, is currently located the relevant lane in lane to target vehicle.Specifically, association lane refers to target The lane that vehicle can drive into from current lane.
It should be noted that association lane can be target vehicle and be currently located lane.
In some embodiments, step 102 can specifically include following steps:
(1) driving information of information and target vehicle according to the map, determines the current lane of target vehicle.
Wherein, the driving information of target vehicle be this vehicle by collected vehicle-to-target vehicle of sensing system it Between relative positional relationship information target vehicle can be derived in height by combining position of this vehicle on high-precision map Exact position on precision map determines target vehicle in high-precision according to exact position of the target vehicle on high-precision map Locating current lane on map.
For example, in the present embodiment, with reference to shown in Fig. 1 d, it is locating current on high-precision map to provide target vehicle The schematic illustration in lane, it is known that position coordinates of this vehicle on high-precision map are (x=0, y=0), collect target vehicle The distance between this vehicle is d=3 meters, and the position angle between target vehicle and this vehicle is θ=60 °, then can calculate mesh Vehicle and relative distance of this vehicle in controlled map reference axis are marked, i.e. target vehicle is d* at a distance from this vehicle is between X-axis Sin θ, the distance between Y-axis are d*cos θ, then position coordinates of the target vehicle on high-precision map are (x=1.5, y=- 2.6), it is known that, which falls into the regional scope (1 < x < 2, -20 < y < 20) in lane 2, current where judgement target vehicle Lane is lane 2.
(2) information carries out topological analysis to the current lane of target vehicle according to the map, obtains associated with current lane Association lane.
After determining the lane that target vehicle is currently located, topological analysis can be carried out to the current lane of target vehicle, Association associated with current lane lane is obtained, specifically, the lane topological relation being currently located by obtaining target vehicle, It is available that the associated association lane in lane is currently located with target vehicle according to the topological relation.
Wherein, which may include the relationships, in some embodiments, map such as abutting, be associated with, include and being connected to Information may include the topological relation between all roads.
(3) driving information based on association lane and target vehicle calculates row of the target vehicle relative to association lane Vehicle feature.
In above-mentioned steps, the corresponding one or more associations lane of each target vehicle can be obtained, target carriage is passed through Driving information can calculate target vehicle relative to association lane driving feature.
Wherein, feature of driving a vehicle refers to physical features when vehicle driving, for example, row of the target vehicle relative to association lane Vehicle feature can be target vehicle be associated with the distance between lane, target vehicle and the relative velocity, the target carriage that are associated with lane With the relative distance etc. that is associated on lane between barrier.
For example, target vehicle may include target vehicle and be associated with obstacle on lane relative to the driving feature in association lane Relative velocity between object, wherein barrier refers to may cause the object hindered, the barrier on association lane to target vehicle Hinder object to can be stationary body, be also possible to dynamic object, for example, barrier can be traffic lights, motor vehicle, non-maneuver Vehicle, greenbelt, etc..
Specifically, target vehicle may include being associated with the land occupation of barrier in lane relative to the driving feature in association lane Relative velocity between barrier and target vehicle in area, size, type, motion state, and association lane, it is opposite away from From, etc..
103, using lane prediction model, according to the driving feature mesh that prediction target vehicle will drive into association lane Mark lane.
Wherein, the type of lane prediction model has a variety of, for example, lane prediction model can be neural network model, Such as convolutional neural networks model (Convolutional Neural Networks, CNN), deep neural network model (Deep Neural Networks, DNN), Recognition with Recurrent Neural Network model (Recurrent Neural Networks, CNN), etc..
For example, in some embodiments, lane prediction model can be the deep neural network based on random forests algorithm Model.
Wherein, random forests algorithm refers to using more decision trees a kind of algorithm for being trained and predicting to sample. It may include multiple decision trees in Random Forest model, for example, the Random Forest model with reference to shown in Fig. 1 e, including two The classification that can be exported by partial decision tree-model of classification of decision tree, Random Forest model output determines, numerical value are as follows:
Wherein, P1(c | f) is the output of left side decision tree in Fig. 1 e, Pn(c | f) is the output of right side decision tree in Fig. 1 e, P (c | f) it is Random Forest model output.
Specifically, in some embodiments, step 103 may include steps of:
(1) lane prediction model is used, association lane is driven into predetermined time according to driving feature calculation target vehicle Drive into probability.
For example, in some embodiments, in order to reduce over-fitting, the efficiency for improving processing high dimensional feature, adapt to it is a large amount of more The prediction data of sample, lane prediction model can be the deep neural network model based on random forests algorithm.At this point, using should Lane prediction model handles driving feature, and available target vehicle drives into the more of an association lane in predetermined time It is a to drive into sub- probability, it sub- probability is driven into these is weighted summation and can find out target vehicle and drive into association vehicle in predetermined time Road drives into probability.
(2) according to probability is driven into, target lane is determined from association lane.
For example, in some embodiments, lane prediction model may include multiple lanes prediction submodels (such as decision tree Model), using lane prediction model, driving into generally for association lane is driven into predetermined time according to driving feature calculation target vehicle Rate can specifically include following steps:
(a) submodel is predicted using lane, association lane is driven into predetermined time according to driving feature calculation target vehicle Drive into sub- probability；
(b) summation is weighted to driving into sub- probability, obtain target vehicle in predetermined time and drive into driving into for association lane Probability.
Association lane is driven into according to target vehicle and drives into probability, and one or more target carriages can be determined from association lane Road.
Wherein, to determine that the method in one or more target lanes has from association lane a variety of, for example, in some implementations In example, to the descending sequence of probability is driven into, the association lane corresponding to probability of driving into of preceding preset quantity is denoted as target Lane；For example, preset quantity is 3, to driving into the descending sequence of probability, association vehicle corresponding to probability is driven by first 3 Road is denoted as target lane.
In some embodiments, it can also will drive into probability to be compared with predetermined probabilities range, predetermined probabilities will be belonged to The association lane corresponding to probability of driving into of range is denoted as target lane.For example, predetermined probabilities range is [0.8,1], then it will be general Drive into probability corresponding to institute relevant lane of the rate score more than or equal to 0.8 is denoted as target lane.
Wherein, preset quantity and predetermined probabilities range can be obtained by reading local memory, can also pass through network It is obtained from server, it can also be by user setting, etc..
104, for using motion prediction model, believed according to movement of the driving feature prediction target vehicle on target lane Breath.
Wherein, motion information may include travel speed, driving direction, phase of the target vehicle when driving on target lane It adjusts the distance, the information such as position location.
Wherein, the type of motion prediction model has a variety of, for example, motion prediction model can be also possible to neural network Model, such as convolutional neural networks model, deep neural network model, Recognition with Recurrent Neural Network model, etc..
Similarly, motion prediction model is also possible to the deep neural network model based on random forests algorithm.
In some embodiments, motion prediction model includes the first motion prediction model, the second motion prediction model, movement Information includes target speed information and target range information, and target lane may include target lane center, and target vehicle exists Motion information on target lane can refer to target speed information and mesh of the target vehicle in predetermined time relative to target lane Subject distance information, step 104 may include steps of:
(1) the first motion prediction model is used, according to driving feature prediction target vehicle in predetermined time relative to target The target speed information in lane；
(2) the second motion prediction model is used, according to driving feature prediction target vehicle in predetermined time relative to target The target range information of lane center.
Wherein, target lane center refers to the lane center from target lane.Specifically, lane center refer to from The origin-to-destination in lane, the line being connected in sequence by the central point between the shoulder of lane.It can be in high-precision map Store the lane center information in each lane.
105, it is based on driving information, target lane and motion information, calculates the motion profile of target vehicle.
In some embodiments, obtain target vehicle predetermined time relative to target lane target speed information, After predetermined time is relative to the target range information in target lane, step 105 can specifically can be based on row target vehicle Information, target lane, target speed information and target range information are sailed, the motion profile of target vehicle is calculated.
In some embodiments, the driving information of target vehicle includes initial position message, initial velocity information, wherein Initial position message can be currently located the distance between the lane middle line in lane, initial velocity with feeling the pulse with the finger-tip mark vehicle-to-target vehicle Information can refer to that the current speed of target vehicle, step 105 can specifically include following steps:
(1) according to target lane and target range information, determine target vehicle in the target position information of predetermined time；
(2) target carriage is calculated based on initial position message, target position information, initial velocity information, target speed information Motion profile.
Target vehicle is calculated from currently to the method for the motion profile future time instance with a variety of.For example, specifically may be used To carry out trajectory calculation, etc. using cubic polynomial curve, quintic algebra curve curve, S curve, step curve etc..
For example, calculation can refer to Fig. 1 f, V1 is the present speed of target vehicle, and V2 is the target carriage that prediction obtains Future time instance (for example, after 3 seconds) target velocity；D1 is that target vehicle is currently currently located between lane center with it Distance, d2 be target vehicle at a distance from where future time instance with its future time instance between target lane center, d is target Vehicle is currently located the distance between lane and target lane.
Wherein, lane center can be obtained from high-precision map, between target vehicle and Future targets vehicle away from It is as follows from the calculation formula of D:
D=d1+d2+d
At this point, by distance D and target vehicle present speed V1 and target vehicle future time instance target velocity Target vehicle can be calculated from currently to the motion profile future time instance in V2.
Specifically, the method that the calculating of motion profile Q is carried out using quintic algebra curve curve is as follows:
Q(tstart,tend)=q (tstart)+q(tstart+1)+q(tstart+2)+...q(tend)=∑ q (ti)
Wherein, location point when q is certain moment where target vehicle；Q is the line being linked to be by multiple location points, that is, is moved Track；tstartFor initial time；tendFor future time instance；tiSampling instant between initial time and future time instance, the sampling Moment can be configured by technical staff, can also be set by the user.
Target vehicle is in the location of sampling instant point q (ti) can be found out by following formula:
q(ti)=q (tsatrt)+w1(ti-tstrat)+w2(ti-tstrat)2+w3(ti-tstrat)3+w4(ti-tstrat)4+w5(ti- tstrat)5
Wherein, w1、w2、w3、w4、w5For the coefficient of the quintic algebra curve.
Known physical equation is as follows, in tiMoment, speed v can ask first derivative to obtain by location point q, acceleration of motion A can ask second dervative to obtain by location point q:
q′(ti)=v (ti), q " (ti)=a (ti)
It can be concluded that coefficient w1、w2、w3、w4、w5Calculation formula it is as follows:
w1=vstart
Wherein, h=q (ti)-q(tstart)。
From the foregoing, it will be observed that the available cartographic information of the embodiment of the present invention, the driving information of target vehicle, lane prediction mould Type, motion prediction model, wherein lane prediction model, motion prediction model are formed by training sample training；Based on cartographic information And the driving information of target vehicle, determine the row of the association lane and target vehicle of target vehicle relative to association lane Vehicle feature；Using lane prediction model, according to the driving feature target carriage that prediction target vehicle will drive into association lane Road；Using motion prediction model, according to motion information of the driving feature prediction target vehicle on target lane；Believed based on traveling Breath, target lane and motion information, calculate the motion profile of target vehicle.
Thus this programme can predict vehicle by different prediction models target lane and vehicle are in target carriage Motion information on road, to carry out trajectory calculation.The program can promote the precision of prediction of trajectory predictions as a result,.
The method according to described in above-described embodiment, will now be described in further detail below.
Trajectory predictions scheme provided in an embodiment of the present invention can be applied in various traffic scenes, for each rank Automatic driving vehicle system, to realize that automatic driving vehicle predicts the motion profile of surrounding vehicles.
In the present embodiment, will be with server training lane prediction model and motion prediction model, automatic driving vehicle is logical It crosses for the obtained model of these servers training predicts the motion profile in surrounding vehicles 3 seconds, with reference to Fig. 2, to this The method of inventive embodiments is described in detail, and detailed process is as follows:
201, server obtains training sample and initial predicted model, and each training sample corresponds to multiple sample marks.
Wherein, the mode of server acquisition training sample and initial predicted model has a variety of, for example, server passes through Network obtains training sample, server directly reads initial predicted model in its local memory, server is obtained by network Training data, and training data is labeled, obtain available training sample, etc..
The training data can be by being equipped with the sensors such as laser radar, camera, High Accuracy Inertial and corresponding perception The acquisition vehicle of algorithm acquires.Wherein, acquisition vehicle can be the vehicle for being exclusively used in collecting training data, be also possible to be equipped with The vehicle of automated driving system.
For example, acquisition vehicle can acquire surrounding traffic condition, and traffic condition institute is recorded according to accurately seal The place of generation.For example, acquisition vehicle can acquire the picture of its peripheral obstacle, the lane at place, the position at place, speed Degree, the direction of motion, acceleration, angular speed and number of barrier, etc..Wherein, barrier may include vehicle, traffic Indicator light, pedestrian etc. can cause the people hindered and object to vehicle driving.
For example, acquisition vehicle can acquire travel speed, the driving direction in 15 meters in other vehicles 3 seconds around, and Initial position, final position in this 3 seconds, thus obtain around in 15 meters lane sums where other vehicle initial times 3 Target lane where after second.
In some embodiments, server can be labeled training data, obtain available training sample, specific to walk It is rapid as follows:
(1) the association lane and target lane of barrier are determined.
Wherein, target lane is barrier in the lane where future time instance (such as after 3 seconds), which can To be included in training data.
Wherein, the association lane of barrier, which refers to, is currently located the associated lane in lane with barrier, specific to determine barrier The mode in the association lane of object is hindered to have a variety of, such as shown in following steps:
A. according to the position location of high-precision map and barrier, the current lane of barrier is determined；
B. topological analysis is carried out according to current lane of the high-precision map to barrier, obtains the association lane of barrier.
Wherein, the position location of barrier can collect its relative position between barrier by acquisition vehicle Relationship is derived with position of the acquisition vehicle on high-precision map to combine.
Wherein, topological relation can be obtained from high-precision map, which may include abutting, being associated with, including With the relationships such as be connected to.
(2) according to training data, determine barrier for the driving feature in association lane.
Wherein, barrier for be associated with lane driving feature may include association lane in barrier occupied area, Relative velocity, relative distance in size, type, motion state, and association lane between barrier and target vehicle, etc. Deng.
For example, shown in reference table 1 driving feature list it is found that barrier for be associated with lane driving feature have it is more Kind:
Table 1
(3) training data is labeled according to target lane.
Wherein, each training sample can correspond to the sample mark of multiple types, for example, in some embodiments, each Training sample can correspond to three sample marks, respectively lane mark, distance mark, speed mark, etc..
In the present embodiment, the association lane in training data can be labeled according to target lane.For example, working as When target lane is identical as association lane, generates the lane that the association lane corresponds to training data and be labeled as positive sample mark, when When target lane is with lane difference is associated with, generates the lane that the association lane corresponds to training data and be labeled as negative sample mark.
For example, the association lane of barrier is lane A, lane B, lane C, then can generate characteristic set according to lane [A]、[B]、[C]。
For example, lane of the barrier where after 3 seconds is lane B, then the target lane of barrier is lane B, it is known that barrier Hinder target lane and the lane A of object different, identical as lane B, different with lane C, then characteristic set [A] is used as negative sample at this time This, is labeled as [A, 0]；Characteristic set [B] is used as positive sample, is labeled as [B, 1]；Characteristic set [C] is used as negative sample, mark Note is [C, 0].
In some embodiments, characteristic set can also be labeled according to travel speed of the barrier after 3 seconds, In, the data of negative sample are noted as without marking travel speed, for example, travel speed of the barrier after 3 seconds is v, then will Negative sample [A, 0] is labeled as [A, 0,0]；Positive sample [B, 1] is labeled as [B, 1, v]；Negative sample [C, 0] is labeled as [C, 0, 0]。
It in some embodiments, can also be according to distance of the barrier after 3 seconds relative to target lane to characteristic set Be labeled, wherein be noted as the data of negative sample without marking distance, for example, barrier after 3 seconds with target lane Distance is D, then negative sample [A, 0,0] is labeled as [A, 0,0,0], positive sample [B, 1, v] is labeled as [B, 1, v, D], will be born Sample [C, 0,0] is labeled as [C, 0,0,0].
202, server carries out rejecting processing to training sample corresponding sample mark, obtains lane training sample and fortune Dynamic training sample.
After different types of sample mark corresponding to training sample carries out rejecting processing, training sample can retain part The sample of type marks.For example, in some embodiments, sample mark includes lane mark, distance mark, speed mark, fortune Dynamic training sample includes carrying out at rejecting apart from training sample and speed training sample to training sample corresponding sample mark Reason, obtains lane training sample and training sample can specifically include following steps:
(a) the distance mark and speed mark for abandoning training sample are only remained the lane training sample of lane mark This；
(b) it abandons the lane mark of training sample and apart from mark, is only remained the speed training sample of speed mark This；
(c) the lane mark and speed mark for abandoning training sample are only remained the distance training sample of distance mark This.
For example, when training sample be negative sample [A, 0,0,0], positive sample [B, 1, v, D], negative sample [C, 0,0,0], In, the first item of training sample refers to candidate lane, and Section 2 refers to sample type (such as the positive sample type, negative sample of training sample Type), Section 3 refers to travel speed of the barrier on target lane, Section 4 refer between barrier and target lane away from From.
For example, executing step a, i.e., the speed mark and Section 4 of discardable training sample three to above-mentioned training sample Distance mark, then obtain lane training sample [A, 0], [B, 1], [C, 0].
For example, executing step b, i.e., the lane mark and Section 4 of discardable training sample first item to above-mentioned training sample Distance mark, then obtain speed training sample [0,0], [1, v], [0,0].
For example, executing step c, i.e., the lane mark and Section 3 of discardable training sample first item to above-mentioned training sample Speed mark, then obtain apart from training sample [0,0], [1, D], [0,0].
203, server is trained initial predicted model using lane training sample, obtains lane prediction model.
For example, server is trained initial predicted model using lane training sample [A, 0], [B, 1], [C, 0], directly To convergence, lane prediction model is obtained.
204, server is trained initial predicted model using training sample, obtains motion prediction model.
In some embodiments, training sample includes speed training sample, apart from training sample, speed training sample It include distance training subsample apart from training sample, motion prediction model includes prediction of speed mould including speed training positive sample Type, range prediction model are trained initial predicted model using training sample, obtain motion prediction model and specifically may be used With the following steps are included:
(a) initial predicted model is trained using speed training subsample, obtains speed prediction model；
(b) initial predicted model is trained using distance training subsample, obtains range prediction model.
For example, speed training sample [0,0], [1, v], first item is sample type in [0,0], wherein 0 represents negative sample This, 1 represents positive sample, then speed training sample [0,0], [1, v], include two negative samples [0,0] and one in [0,0] Positive sample [1, v].
205, automatic driving vehicle obtains cartographic information, lane prediction model, motion prediction model from server, and passes through The driving information of sensor acquisition target vehicle.
Wherein, automatic driving vehicle can obtain cartographic information, lane prediction model, movement in advance from server by network Model is surveyed, cartographic information, lane prediction model, motion prediction model, etc. can also be imported from server by storage equipment.
206, driving information of the automatic driving vehicle based on cartographic information, target vehicle, using lane prediction model and fortune The motion profile of dynamic prediction model prediction target vehicle.
Step (6) can refer to above-mentioned steps 102,103,104,105, and this will not be repeated here.
From the foregoing, it will be observed that in embodiments of the present invention, server, which passes through, obtains training sample and initial predicted model, and right The corresponding sample mark of training sample carries out rejecting processing, obtains lane training sample and training sample, server can To be trained using lane training sample to initial predicted model, lane prediction model is obtained, using training sample pair Initial predicted model is trained, and obtains motion prediction model.Automatic driving vehicle is pre- from server acquisition cartographic information, lane Model, motion prediction model are surveyed, and acquires the driving information of target vehicle by sensor；Automatic driving vehicle is obtained from server Cartographic information, lane prediction model, motion prediction model are taken, and acquires the driving information of target vehicle by sensor.
The target lane and vehicle for predicting vehicle by different prediction models in embodiments of the present invention as a result, exist Motion information on target lane, to carry out trajectory calculation.The efficiency of trajectory predictions can be improved in the program as a result, simultaneously Promote the precision of prediction of trajectory predictions.
In order to better implement above method, the embodiment of the present invention also provides a kind of trajectory predictions device, the trajectory predictions Device specifically can integrate in the electronic device, which can be the equipment such as terminal, server.Wherein, which sets It is standby to be specifically as follows automatic pilot, server, etc..
For example, in the present embodiment, it will be by taking trajectory predictions device be integrated in the server as an example, to the embodiment of the present invention Method is described in detail.
For example, as shown in Figure 3a, which may include acquiring unit 301, associative cell 302, lane list Member 303, moving cell 304 and trajectory unit 305 are as follows:
(1) acquiring unit 301:
Acquiring unit 301, for obtaining driving information, the lane prediction model, motion prediction of cartographic information, target vehicle Model, wherein lane prediction model, motion prediction model are formed by training sample training.
With reference to Fig. 3 b, in some embodiments, acquiring unit 301 can also include obtaining subelement 3011, rejecting son list Member 3012, lane model subelement 3013 and motion model subelement 3014, as follows:
(1) subelement 3011 is obtained:
Subelement is obtained, for obtaining training sample and initial predicted model, each training sample corresponds to multiple samples Mark；
(2) subelement 3012 is rejected:
Subelement is rejected, for carrying out rejecting processing to training sample corresponding sample mark, obtains lane training sample And training sample；
(3) lane model subelement 3013:
It is pre- to obtain lane for being trained using lane training sample to initial predicted model for lane model subelement Survey model；
(4) motion model subelement 3014:
It is pre- to obtain movement for being trained using training sample to initial predicted model for motion model subelement Survey model.
In some embodiments, sample mark includes lane mark, distance mark, speed mark, training sample packet It includes apart from training sample and speed training sample, rejecting subelement 3012 specifically can be used for:
The distance mark and speed mark for abandoning training sample, are only remained the lane training sample of lane mark；
It abandons the lane mark of training sample and apart from mark, is only remained the speed training sample of speed mark；
The lane mark and speed mark for abandoning training sample, are only remained apart from mark apart from training sample.
In some embodiments, training sample includes speed training sample, apart from training sample, speed training sample It include distance training subsample apart from training sample, motion prediction model includes prediction of speed mould including speed training positive sample Type, range prediction model, motion model subelement 3014 specifically can be used for:
Initial predicted model is trained using speed training subsample, obtains speed prediction model；
Initial predicted model is trained using distance training subsample, obtains range prediction model.
(2) associative cell 302:
Associative cell 302 determines the association of target vehicle for the driving information based on cartographic information and target vehicle The driving feature of lane and target vehicle relative to association lane.
In some embodiments, associative cell 302 specifically can be used for:
The driving information of information and target vehicle according to the map, determines the current lane of target vehicle；
Information carries out topological analysis to the current lane of target vehicle according to the map, obtains pass associated with current lane Join lane；
Driving information based on association lane and target vehicle, the driving for calculating target vehicle relative to association lane are special Sign.
(3) lane unit 303:
Lane unit 303 predicts target vehicle in association lane according to driving feature for using lane prediction model The target lane that will be driven into.
In some embodiments, with reference to Fig. 3 c, lane unit 303 may include probability subelement 3031 and lane subelement 3032, as follows:
(1) probability subelement 3031:
Probability subelement is sailed according to driving feature calculation target vehicle in predetermined time for using lane prediction model Enter to be associated with lane drives into probability；
(2) lane subelement 3032:
Lane subelement, for determining target lane from association lane according to probability is driven into.
In some embodiments, lane prediction model includes multiple lane prediction submodels, and probability subelement 3031 is specific It can be used for:
Submodel is predicted using lane, sailing for association lane is driven into predetermined time according to driving feature calculation target vehicle Enter sub- probability；
It is weighted summation to sub- probability is driven into, target vehicle is obtained in predetermined time and drives into driving into generally for association lane Rate.
(4) moving cell 304:
Moving cell 304, for using motion prediction model, according to driving feature prediction target vehicle on target lane Motion information.
In some embodiments, motion prediction model includes the first motion prediction model, the second motion prediction model, movement Information includes target speed information and target range information, and with reference to Fig. 3 d, moving cell 304 may include the first movement subelement 3041 and second move subelement 3042, as follows:
(1) first moving cell 3041:
First moving cell, for using the first motion prediction model, according to driving feature prediction target vehicle default Target speed information of the moment relative to target lane；
(2) second moving cells 3042:
Second moving cell, for using the second motion prediction model, according to driving feature prediction target vehicle default Target range information of the moment relative to target lane.
(5) trajectory unit 305:
Trajectory unit 305 calculates the movement rail of target vehicle for being based on driving information, target lane and motion information Mark.
In some embodiments, trajectory unit 305 specifically can be used for based on driving information, target lane, target velocity Information and target range information, calculate the motion profile of target vehicle.
In some embodiments, driving information includes initial position message, initial velocity information；With reference to Fig. 3 e, track is single Member 305 may include location subunit 3051 and track subelement 3052, as follows:
(1) location subunit 3051:
Location subunit, for determining target vehicle in the mesh of predetermined time according to target lane and target range information Cursor position information；
(2) track subelement 3052:
Track subelement, for based on initial position message, target position information, initial velocity information, target velocity letter Breath calculates the motion profile of target vehicle.
When it is implemented, above each unit can be used as independent entity to realize, any combination can also be carried out, is made It is realized for same or several entities, the specific implementation of above each unit can be found in the embodiment of the method for front, herein not It repeats again.
From the foregoing, it will be observed that the trajectory predictions device of the present embodiment is obtained the traveling of cartographic information, target vehicle by acquiring unit Information, lane prediction model, motion prediction model；Driving information by associative cell based on cartographic information and target vehicle, Determine the driving feature of the association lane and target vehicle of target vehicle relative to association lane；Vehicle is used by lane unit Road prediction model, according to the driving feature target lane that prediction target vehicle will drive into association lane；By moving cell Using motion prediction model, according to motion information of the driving feature prediction target vehicle on target lane；By trajectory unit base In driving information, target lane and motion information, the motion profile of target vehicle is calculated.
The target lane that vehicle can be predicted by different prediction models due to the program and vehicle are in target carriage Motion information on road, to carry out trajectory calculation.The program can promote the precision of prediction of trajectory predictions as a result,.
The embodiment of the present invention also provides a kind of electronic equipment, which can be smart phone, smartwatch, plate Computer, microcomputer, automatic pilot, server etc..As shown in figure 4, it illustrates involved in the embodiment of the present invention The structural schematic diagram of electronic equipment, specifically:
The electronic equipment may include one or more than one processing core processor 401, one or more Memory 402, power supply 403, the input unit 404 of computer readable storage medium, in addition to this it is possible to include sensor system The components such as system 405, positioning system 406.It will be understood by those skilled in the art that the not structure of electronic devices structure shown in Fig. 4 The restriction of paired electrons equipment may include perhaps combining certain components or different than illustrating more or fewer components Component layout.Wherein:
Processor 401 is the control centre of the electronic equipment, utilizes various interfaces and the entire electronic equipment of connection Various pieces by travelling or execute the software program and/or module that are stored in memory 402, and are called and are stored in Data in reservoir 402 execute the various functions and processing data of electronic equipment, to carry out integral monitoring to electronic equipment. In some embodiments, processor 401 may include one or more processing cores；In some embodiments, processor 401 can collect At application processor and modem processor, wherein the main processing operation system of application processor, user interface and apply journey Sequence etc., modem processor mainly handle wireless communication.It is understood that above-mentioned modem processor can not also collect At in processor 401.
Memory 402 can be used for storing software program and module, and processor 401 is stored in memory 402 by traveling Software program and module, thereby executing various function application and data processing.Memory 402 can mainly include storage journey Sequence area and storage data area, wherein storing program area can application program needed for storage program area, at least one function etc.； Storage data area, which can be stored, uses created data etc. according to electronic equipment.In addition, memory 402 may include high speed with Machine access memory, can also include nonvolatile memory, a for example, at least disk memory, flush memory device or its His volatile solid-state part.Correspondingly, memory 402 can also include Memory Controller, right to provide processor 401 The access of memory 402.
Electronic equipment further includes the power supply 403 powered to all parts, and in some embodiments, power supply 403 can pass through Power-supply management system and processor 401 are logically contiguous, to realize management charging, electric discharge, Yi Jigong by power-supply management system The functions such as consumption management.Power supply 403 can also include one or more direct current or AC power source, recharging system, power supply The random components such as fault detection circuit, power adapter or inverter, power supply status indicator.
The electronic equipment may also include input unit 404, the input unit 404 can be used for receiving the number of input, character, Image, location information etc., and generate key related with user setting and function control, dummy keyboard, steering wheel, behaviour Make the signals such as bar, sensor input, for example, input unit can receive sensing system and positioning system input image, Manage position and driving information etc..
The electronic equipment may also include sensing system 405, which may include multiple sensors, than Such as, radar, camera, infrared sensor, etc..The structure of sensing system 405 can for centralization, distribution, stagewise, Hybrid and multi-stag etc., multiple sensors therein may include sensing element, conversion original part, accessory power supply and transformation The components such as circuit, sensing element can be with direct feeling, measurement, and exports and be measured the physical quantity signal for having determining relationship；Turn It changes element and the physical quantity signal that sensing element exports is converted into electric signal；Translation circuit is responsible for the telecommunications exported to conversion element Number amplify modulation；Conversion element and translation circuit generally also need accessory power supply to power.
The electronic equipment may also include positioning system 406, which can receive, tracks, converts and measure position Confidence number provides the position and speed of carrier in real time.Positioning system 406 can be by antenna element, receiver main computer unit and electricity Three, source composition, the location navigation signal that can be will acquire of antenna element is converted into electric current, and carries out to this signal code Amplification and frequency-conversion processing；Receiver unit can be tracked, handled and be surveyed to the signal power source by amplification and frequency-conversion processing Amount.
Although being not shown, electronic equipment can also include display unit, communication unit etc., and details are not described herein.Specifically exist In the present embodiment, the processor 401 in electronic equipment can be according to following instruction, by one or more application program The corresponding executable file of process is loaded into memory 402, and is travelled and be stored in memory 402 by processor 401 Application program, thus realize various functions, it is as follows:
Obtain cartographic information, the driving information of target vehicle, lane prediction model, motion prediction model, wherein lane is pre- Survey model, motion prediction model is formed by training sample training；
Driving information based on cartographic information and target vehicle determines association lane and the target carriage of target vehicle Relative to association lane driving feature；
Using lane prediction model, according to the driving feature target carriage that prediction target vehicle will drive into association lane Road；
Using motion prediction model, according to motion information of the driving feature prediction target vehicle on target lane；
Based on driving information, target lane and motion information, the motion profile of target vehicle is calculated.
The specific implementation of above each operation can be found in the embodiment of front, and details are not described herein.
From the foregoing, it will be observed that in embodiments of the present invention, the traveling letter of the available cartographic information of electronic equipment, target vehicle Breath, lane prediction model, motion prediction model；Driving information based on cartographic information and target vehicle, determines target vehicle Association lane and target vehicle relative to association lane driving feature；Using lane prediction model, according to driving feature The target lane that prediction target vehicle will drive into association lane；Using motion prediction model, according to driving feature prediction Motion information of the target vehicle on target lane；Based on driving information, target lane and motion information, target vehicle is calculated Motion profile.Thus the program can predict vehicle by different prediction models target lane and vehicle are in target carriage Motion information on road, to carry out trajectory calculation.The program can promote the precision of prediction of trajectory predictions as a result,.
It will appreciated by the skilled person that all or part of the steps in the various methods of above-described embodiment can be with It is completed by instructing, or relevant hardware is controlled by instruction to complete, which can store computer-readable deposits in one In storage media, and is loaded and executed by processor.
For this purpose, the embodiment of the present invention provides a kind of storage medium, wherein being stored with a plurality of instruction, which can be processed Device is loaded, to execute the step in any trajectory predictions method provided by the embodiment of the present invention.For example, the instruction can To execute following steps:
Obtain cartographic information, the driving information of target vehicle, lane prediction model, motion prediction model, wherein lane is pre- Survey model, motion prediction model is formed by training sample training；
Driving information based on cartographic information and target vehicle determines association lane and the target carriage of target vehicle Relative to association lane driving feature；
Using lane prediction model, according to the driving feature target carriage that prediction target vehicle will drive into association lane Road；
Using motion prediction model, according to motion information of the driving feature prediction target vehicle on target lane；
Based on driving information, target lane and motion information, the motion profile of target vehicle is calculated.
Wherein, which may include: read-only memory (ROM, Read Only Memory), random access memory Body (RAM, Random Access Memory), disk or CD etc..
By the instruction stored in the storage medium, it is pre- that any track provided by the embodiment of the present invention can be executed Step in survey method, it is thereby achieved that achieved by any trajectory predictions method provided by the embodiment of the present invention Beneficial effect is detailed in the embodiment of front, and details are not described herein.
It is provided for the embodiments of the invention a kind of trajectory predictions method and apparatus above to be described in detail, herein Apply that a specific example illustrates the principle and implementation of the invention, the explanation of above example is only intended to help Understand method and its core concept of the invention；Meanwhile for those skilled in the art, according to the thought of the present invention, having There will be changes in body embodiment and application range, in conclusion the content of the present specification should not be construed as to the present invention Limitation.

Claims (10)

1. a kind of trajectory predictions method characterized by comprising
Obtain cartographic information, the driving information of target vehicle, lane prediction model, motion prediction model, wherein the lane is pre- Survey model, motion prediction model is formed by training sample training；
Driving information based on the cartographic information and target vehicle determines association lane and the target carriage of target vehicle Relative to it is described association lane driving feature；
Using lane prediction model, according to the driving feature mesh that prediction target vehicle will drive into the association lane Mark lane；
Using motion prediction model, according to motion information of the driving feature prediction target vehicle on the target lane；
Based on the target lane, motion information and driving information, the motion profile of target vehicle is calculated.
2. trajectory predictions method as described in claim 1, which is characterized in that the motion prediction model includes that the first movement is pre- Model, the second motion prediction model are surveyed, the motion information on the target lane is included in predetermined time relative to institute The target speed information and target range information in target lane are stated, the target lane includes target lane center；
It is described to use motion prediction model, believed according to movement of the driving feature prediction target vehicle on the target lane Breath, comprising:
Using the first motion prediction model, predict target vehicle in predetermined time relative to the target according to the driving feature The target speed information in lane；
Using the second motion prediction model, predict target vehicle in predetermined time relative to the target according to the driving feature The target range information of lane center；
It is described to be based on the driving information, target lane and motion information, calculate the motion profile of target vehicle, comprising:
Based on the driving information, target lane, target speed information and target range information, the movement rail of target vehicle is calculated Mark.
3. trajectory predictions method as claimed in claim 2, which is characterized in that the driving information include initial position message, Initial velocity information；
Based on the driving information, target lane, target speed information and target range information, the movement rail of target vehicle is calculated Mark, comprising:
According to the target lane and target range information, determine target vehicle in the target position information of predetermined time；
Target vehicle is calculated based on the initial position message, target position information, initial velocity information, target speed information Motion profile.
4. trajectory predictions method as described in claim 1, which is characterized in that it is described to use lane prediction model, according to described The driving feature target lane that prediction target vehicle will drive into the association lane, comprising:
Using lane prediction model, the association lane is driven into predetermined time according to the driving feature calculation target vehicle Drive into probability；
Probability is driven into according to described, target lane is determined from the association lane.
5. trajectory predictions method as claimed in claim 4, which is characterized in that the lane prediction model includes that multiple lanes are pre- It surveys submodel and the association is driven into predetermined time according to the driving feature calculation target vehicle using lane prediction model Probability is driven into lane, comprising:
Submodel is predicted using lane, the association lane is driven into predetermined time according to the driving feature calculation target vehicle Drive into sub- probability；
It drives into sub- probability to described and is weighted summation, obtain target vehicle in predetermined time and drive into driving into for the association lane Probability.
6. trajectory predictions method as described in claim 1, which is characterized in that obtain the traveling letter of cartographic information, target vehicle Breath, lane prediction model, motion prediction model, before, further includes:
Obtain training sample and initial predicted model, the corresponding multiple sample marks of each of described training sample；
Rejecting processing is carried out to the training sample corresponding sample mark, obtains lane training sample and training sample This；
Initial predicted model is trained using the lane training sample, obtains lane prediction model；
Initial predicted model is trained using the training sample, obtains motion prediction model.
7. trajectory predictions method as claimed in claim 6, which is characterized in that the sample mark includes lane mark, distance Mark, speed mark, the training sample includes apart from training sample and speed training sample；
Rejecting processing is carried out to the training sample corresponding sample mark, obtains lane training sample and training sample This, comprising:
The distance mark and speed mark for abandoning the training sample, are only remained the lane training sample of lane mark；
It abandons the lane mark of the training sample and apart from mark, is only remained the speed training sample of speed mark；
The lane mark and speed mark for abandoning the training sample, are only remained apart from mark apart from training sample.
8. trajectory predictions method as claimed in claim 6, which is characterized in that the training sample includes speed training sample Originally, apart from training sample, the speed training sample includes speed training subsample, described to include distance instruction apart from training sample Practice subsample, the motion prediction model includes speed prediction model, range prediction model；
Initial predicted model is trained using the training sample, obtains motion prediction model, comprising:
Initial predicted model is trained using the speed training subsample, obtains speed prediction model；
Initial predicted model is trained using distance training subsample, obtains range prediction model.
9. trajectory predictions method as described in claim 1, which is characterized in that based on the cartographic information and target vehicle Driving information determines the driving feature of the association lane and target vehicle of target vehicle relative to the association lane, packet It includes:
According to the cartographic information and the driving information of target vehicle, the current lane of target vehicle is determined；
Topological analysis is carried out according to current lane of the cartographic information to target vehicle, is obtained associated with the current lane Association lane；
Driving information based on the association lane and target vehicle calculates row of the target vehicle relative to the association lane Vehicle feature.
10. a kind of trajectory predictions device characterized by comprising
Acquiring unit, for obtaining the driving information, lane prediction model, motion prediction model of cartographic information, target vehicle, In, the lane prediction model, motion prediction model are formed by training sample training；
Associative cell determines the association vehicle of target vehicle for the driving information based on the cartographic information and target vehicle The driving feature of road and target vehicle relative to the association lane；
Lane unit predicts target carriage in the association lane according to the driving feature for using lane prediction model The target lane that will be driven into；
Moving cell predicts target vehicle in the target lane for using motion prediction model according to the driving feature On motion information；
Trajectory unit calculates the motion profile of target vehicle for being based on the driving information, target lane and motion information.
CN201910729763.2A 2019-08-08 2019-08-08 Trajectory prediction method and apparatus Active CN110400490B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910729763.2A CN110400490B (en) 2019-08-08 2019-08-08 Trajectory prediction method and apparatus

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910729763.2A CN110400490B (en) 2019-08-08 2019-08-08 Trajectory prediction method and apparatus

Publications (2)

Publication Number Publication Date
CN110400490A true CN110400490A (en) 2019-11-01
CN110400490B CN110400490B (en) 2022-02-25

Family

ID=68327772

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910729763.2A Active CN110400490B (en) 2019-08-08 2019-08-08 Trajectory prediction method and apparatus

Country Status (1)

Country Link
CN (1) CN110400490B (en)

Cited By (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111114554A (en) * 2019-12-16 2020-05-08 苏州智加科技有限公司 Method, device, terminal and storage medium for predicting travel track
CN111340880A (en) * 2020-02-17 2020-06-26 北京百度网讯科技有限公司 Method and apparatus for generating a predictive model
CN111428943A (en) * 2020-04-23 2020-07-17 福瑞泰克智能系统有限公司 Method, device and computer device for predicting obstacle vehicle track
CN111583715A (en) * 2020-04-29 2020-08-25 宁波吉利汽车研究开发有限公司 Vehicle track prediction method, vehicle collision early warning method, device and storage medium
CN111752272A (en) * 2020-04-16 2020-10-09 北京京东乾石科技有限公司 Trajectory prediction method, apparatus, device and storage medium
CN111824157A (en) * 2020-07-14 2020-10-27 广州小鹏车联网科技有限公司 Automatic driving method and device
CN112562328A (en) * 2020-11-27 2021-03-26 腾讯科技（深圳）有限公司 Vehicle behavior prediction method and device
CN112833903A (en) * 2020-12-31 2021-05-25 广州文远知行科技有限公司 Trajectory prediction method, apparatus, device and computer readable storage medium
WO2021174445A1 (en) * 2020-03-04 2021-09-10 华为技术有限公司 Method and device for predicting exit for vehicle
CN113997954A (en) * 2021-11-29 2022-02-01 广州文远知行科技有限公司 Vehicle driving intention prediction method, device and equipment and readable storage medium
CN114005280A (en) * 2021-11-17 2022-02-01 同济大学 Vehicle track prediction method based on uncertainty estimation
WO2022022384A1 (en) * 2020-07-31 2022-02-03 华为技术有限公司 Method and device for recognizing vehicle motion state
CN114283576A (en) * 2020-09-28 2022-04-05 华为技术有限公司 Vehicle intention prediction method and related device
WO2022178858A1 (en) * 2021-02-26 2022-09-01 华为技术有限公司 Vehicle driving intention prediction method and apparatus, terminal and storage medium
CN115123252A (en) * 2022-07-05 2022-09-30 小米汽车科技有限公司 Vehicle control method, vehicle control device, vehicle and storage medium
CN116153084A (en) * 2023-04-20 2023-05-23 智慧互通科技股份有限公司 Vehicle flow direction prediction method, prediction system and urban traffic signal control method
WO2023123456A1 (en) * 2021-12-31 2023-07-06 深圳市大疆创新科技有限公司 Vehicle location prediction method and apparatus, and vehicle and storage medium
WO2023159915A1 (en) * 2022-02-23 2023-08-31 中国第一汽车股份有限公司 Vehicle trajectory prediction method and device

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140032108A1 (en) * 2012-07-30 2014-01-30 GM Global Technology Operations LLC Anchor lane selection method using navigation input in road change scenarios
CN103809593A (en) * 2012-11-06 2014-05-21 现代摩比斯株式会社 Control apparatus of vehicle for changing lane and control method of the same
EP2950294A1 (en) * 2014-05-30 2015-12-02 Honda Research Institute Europe GmbH Method and vehicle with an advanced driver assistance system for risk-based traffic scene analysis
CN107919027A (en) * 2017-10-24 2018-04-17 北京汽车集团有限公司 Predict the methods, devices and systems of vehicle lane change
CN108981729A (en) * 2017-06-02 2018-12-11 腾讯科技（深圳）有限公司 Vehicle positioning method and device
CN109583151A (en) * 2019-02-20 2019-04-05 百度在线网络技术（北京）有限公司 The driving trace prediction technique and device of vehicle
CN109885066A (en) * 2019-03-26 2019-06-14 北京经纬恒润科技有限公司 A kind of motion profile prediction technique and device
CN110020748A (en) * 2019-03-18 2019-07-16 杭州飞步科技有限公司 Trajectory predictions method, apparatus, equipment and storage medium

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140032108A1 (en) * 2012-07-30 2014-01-30 GM Global Technology Operations LLC Anchor lane selection method using navigation input in road change scenarios
CN103809593A (en) * 2012-11-06 2014-05-21 现代摩比斯株式会社 Control apparatus of vehicle for changing lane and control method of the same
EP2950294A1 (en) * 2014-05-30 2015-12-02 Honda Research Institute Europe GmbH Method and vehicle with an advanced driver assistance system for risk-based traffic scene analysis
CN108981729A (en) * 2017-06-02 2018-12-11 腾讯科技（深圳）有限公司 Vehicle positioning method and device
CN107919027A (en) * 2017-10-24 2018-04-17 北京汽车集团有限公司 Predict the methods, devices and systems of vehicle lane change
CN109583151A (en) * 2019-02-20 2019-04-05 百度在线网络技术（北京）有限公司 The driving trace prediction technique and device of vehicle
CN110020748A (en) * 2019-03-18 2019-07-16 杭州飞步科技有限公司 Trajectory predictions method, apparatus, equipment and storage medium
CN109885066A (en) * 2019-03-26 2019-06-14 北京经纬恒润科技有限公司 A kind of motion profile prediction technique and device

Cited By (27)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111114554A (en) * 2019-12-16 2020-05-08 苏州智加科技有限公司 Method, device, terminal and storage medium for predicting travel track
CN111114554B (en) * 2019-12-16 2021-06-11 苏州智加科技有限公司 Method, device, terminal and storage medium for predicting travel track
CN111340880A (en) * 2020-02-17 2020-06-26 北京百度网讯科技有限公司 Method and apparatus for generating a predictive model
CN111340880B (en) * 2020-02-17 2023-08-04 北京百度网讯科技有限公司 Method and apparatus for generating predictive model
WO2021174445A1 (en) * 2020-03-04 2021-09-10 华为技术有限公司 Method and device for predicting exit for vehicle
CN111752272A (en) * 2020-04-16 2020-10-09 北京京东乾石科技有限公司 Trajectory prediction method, apparatus, device and storage medium
CN111428943A (en) * 2020-04-23 2020-07-17 福瑞泰克智能系统有限公司 Method, device and computer device for predicting obstacle vehicle track
CN111428943B (en) * 2020-04-23 2021-08-03 福瑞泰克智能系统有限公司 Method, device and computer device for predicting obstacle vehicle track
CN111583715A (en) * 2020-04-29 2020-08-25 宁波吉利汽车研究开发有限公司 Vehicle track prediction method, vehicle collision early warning method, device and storage medium
CN111824157A (en) * 2020-07-14 2020-10-27 广州小鹏车联网科技有限公司 Automatic driving method and device
WO2022022384A1 (en) * 2020-07-31 2022-02-03 华为技术有限公司 Method and device for recognizing vehicle motion state
CN114056347A (en) * 2020-07-31 2022-02-18 华为技术有限公司 Vehicle motion state identification method and device
CN114283576A (en) * 2020-09-28 2022-04-05 华为技术有限公司 Vehicle intention prediction method and related device
CN112562328A (en) * 2020-11-27 2021-03-26 腾讯科技（深圳）有限公司 Vehicle behavior prediction method and device
CN112833903B (en) * 2020-12-31 2024-04-23 广州文远知行科技有限公司 Track prediction method, device, equipment and computer readable storage medium
CN112833903A (en) * 2020-12-31 2021-05-25 广州文远知行科技有限公司 Trajectory prediction method, apparatus, device and computer readable storage medium
EP4286972A4 (en) * 2021-02-26 2024-03-27 Huawei Technologies Co., Ltd. Vehicle driving intention prediction method and apparatus, terminal and storage medium
WO2022178858A1 (en) * 2021-02-26 2022-09-01 华为技术有限公司 Vehicle driving intention prediction method and apparatus, terminal and storage medium
CN114005280A (en) * 2021-11-17 2022-02-01 同济大学 Vehicle track prediction method based on uncertainty estimation
CN113997954B (en) * 2021-11-29 2023-11-21 广州文远知行科技有限公司 Method, device and equipment for predicting vehicle driving intention and readable storage medium
CN113997954A (en) * 2021-11-29 2022-02-01 广州文远知行科技有限公司 Vehicle driving intention prediction method, device and equipment and readable storage medium
WO2023123456A1 (en) * 2021-12-31 2023-07-06 深圳市大疆创新科技有限公司 Vehicle location prediction method and apparatus, and vehicle and storage medium
WO2023159915A1 (en) * 2022-02-23 2023-08-31 中国第一汽车股份有限公司 Vehicle trajectory prediction method and device
CN115123252B (en) * 2022-07-05 2023-03-31 小米汽车科技有限公司 Vehicle control method, vehicle control device, vehicle and storage medium
CN115123252A (en) * 2022-07-05 2022-09-30 小米汽车科技有限公司 Vehicle control method, vehicle control device, vehicle and storage medium
CN116153084A (en) * 2023-04-20 2023-05-23 智慧互通科技股份有限公司 Vehicle flow direction prediction method, prediction system and urban traffic signal control method
CN116153084B (en) * 2023-04-20 2023-09-08 智慧互通科技股份有限公司 Vehicle flow direction prediction method, prediction system and urban traffic signal control method

Also Published As

Publication number Publication date
CN110400490B (en) 2022-02-25

Similar Documents

Publication Publication Date Title
CN110400490A (en) Trajectory predictions method and apparatus
CN110111566A (en) Trajectory predictions method, apparatus and storage medium
CN109059944B (en) Motion planning method based on driving habit learning
CN110388931A (en) The two-dimentional bounding box of object is converted into the method for the three-dimensional position of automatic driving vehicle
CN110345955A (en) Perception and planning cooperation frame for automatic Pilot
CN110378483A (en) The system and method for training machine learning model being deployed on analog platform
CN109937343A (en) Appraisal framework for the prediction locus in automatic driving vehicle traffic forecast
CN108732589A (en) The training data of Object identifying is used for using 3D LIDAR and positioning automatic collection
CN110239562A (en) The real-time perception adjustment based on surrounding vehicles behavior of automatic driving vehicle is adjusted with driving
CN110389584A (en) Method for assessing the track candidate item of automatic driving vehicle
CN110377024A (en) Automaticdata for automatic driving vehicle marks
CN110389583A (en) The method for generating the track of automatic driving vehicle
CN109429518A (en) Automatic Pilot traffic forecast based on map image
CN110325935A (en) The lane guide line based on Driving Scene of path planning for automatic driving vehicle
CN110377025A (en) Sensor aggregation framework for automatic driving vehicle
CN110389581A (en) Method for the prediction locus for automatic driving vehicle dyspoiesis object
CN108089572A (en) For the algorithm and infrastructure of steady and effective vehicle location
CN108981730A (en) For generating the method and system of reference path for operation automatic driving vehicle
CN110389585A (en) The speed planning device based on study for automatic driving vehicle
CN110386142A (en) Pitch angle calibration method for automatic driving vehicle
CN110096053A (en) Driving locus generation method, system and machine readable media for automatic driving vehicle
CN108139884A (en) The method simulated the physical model of automatic driving vehicle movement and combine machine learning
CN108995657A (en) Operate the method and data processing system of automatic driving vehicle
CN111771207A (en) Enhanced vehicle tracking
CN107310550A (en) Road vehicles travel control method and device

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