CN104867329A - Vehicle state prediction method of Internet of vehicles - Google Patents

Abstract

The invention relates to a vehicle state prediction method of Internet of vehicles. The method comprises the following steps: 1), obtaining vehicle environment information of a target vehicle and information of traffic lights in front of the target vehicle; 2), according to the vehicle environment information and historical state information of the target vehicle, predicting vehicle behavior of the target vehicle; and 3), according to the vehicle environment information, the information of the traffic lights in front of the target vehicle and the vehicle behavior obtained through prediction, through combination with the historical state information of the target vehicle, predicting the state of the vehicle. Compared to the prior art, the method provided by the invention has the following advantages: the vehicle behavior of the vehicle is predicted based on the vehicle environment information and the historical state of the vehicle, then the future state of the vehicle is predicted through the vehicle behavior, the prediction based on the vehicle behavior is more accurate, and the method is good in expandability, is simple and highly-efficient and is suitable for large-scale network conditions.

Description

A kind of car online vehicles trend prediction method

Technical field

The present invention relates to a kind of vehicle-state forecasting techniques, especially relate to a kind of car online vehicles trend prediction method.

Background technology

Vehicle-state comprises the position of vehicle, speed and acceleration, the movement properties of these three vehicles is for the communication in car networking between vehicle and vehicle, communication between vehicle and infrastructure is most important, studies these three and runs the important foundation that attribute is network connectivty in the networking of research car.Whether position relationship is to two communication nodes in mutual infinite network coverage, and namely whether two communication nodes can communicate.Communication node can be vehicle or static roadside infrastructure.Speed and acceleration action are to the duration of two communication nodes in communication range, and namely communication keeps duration.Therefore the position of vehicle, speed and acceleration are the key factors affecting car intranet network connectedness.Traditional network routing protocol does not consider vehicle location in car networking, speed and acceleration attribute, be difficult to adapt to the feature that in car networking, height of car runs, comprise DSDV route, the classical Mobile Ad-hoc network network agreements such as DSR route and AODV route, the people such as Karp propose the stateless route GPSR on the greedy border of location-based Routing Protocol first, all vehicle node are all regular upgrades its positional information to location server, when source node needs to send data to destination node, to location server request destination node position, then with the next-hop node that greedy forward mode each chosen distance destination node is nearest.Due to its find down hop time only according to air line distance, do not consider the problem of crossing barrier, GPCR has done an improvement to this, then this node is directly transmitted to when namely finding that crossing exists node, but whether there is node to crossing to be difficult to judge, judge whether crossing exists node easily by map, so current main study hotspot concentrates on the Routing Protocol based on map.Geographical source routing GSR is one, and typically based on the Routing Protocol of map, it also needs the position obtaining destination node from location server, then utilizes electronic chart and positional information to calculate the shortest path of source node to the node of object by dijkstra's algorithm.The Routing Protocol of above position-based and map make use of vehicle location well and carries out data retransmission, but do not consider the characteristic of vehicle high-speed movement, due to the position frequent variations of vehicle, forward node and destination node shift in position are very large, and data retransmission was probably lost efficacy.Forwarding possibility failure because of the shift in position of adjacent node of single-hop node in GPSR agreement, the research of the people such as Alsaqour R shows that inaccurate position obviously will reduce the efficiency of location-based agreement, and propose a kind of NWLBP (Neignborwireless link break prediction for this problem, adjacent node failure prediction) predict the position of adjacent node, but the method only can solve wall scroll adjacent node shift in position problem, the problem of destination node variation can not be solved, and GPSR agreement with the position of destination node for route leads, therefore the position of destination node is most important.

Existing vehicle-state forecasting techniques is mainly used in vehicle and exempts from the system of hitting, and according to track and its position of prediction of speed of surrounding vehicles, the people such as Hermes C utilize the historical position information of vehicle to use position after trajectory predictions technological prediction vehicle 1-2s.And use track sorting technique, extraction training is carried out to a large amount of vehicle operating track, different tracks is divided into the multiple behavioural characteristics such as left-hand rotation, right-hand rotation, then pattern recognition classifier is carried out to existing track and predict.Dagdelen M, Reymond G, Kemeny A, et al.Model-based predictive motioncueing strategy for vehicle driving simulators [J] .Control Engineering Practice, 2009, 17 (9): 995-1003 and Sorstedt J, Svensson L, Sandblom F, et al.A new vehicle motionmodel for improved predictions and situation assessment [J] .Intelligent TransportationSystems, IEEE Transactions on, 2011, driver turns left by 12 (4): 1209-1219, turn right and become the subjective purposes of driving such as track and add forecast model.Andita R, Caveney D S.Preceding vehicle stateprediction:U.S.Patent Application 13/535,513 [P] .2012-6-28 proposes the vehicle-state forecasting techniques of the fleet of some car compositions, employs Intelligent Driver Model to simulate with garage to be.Houenou A, Bonnifait P, Cherfaoui V, et al.Vehicle trajectory prediction based on motion model andmaneuver recognition [C] //Intelligent Robots and Systems (IROS), 2013IEEE/RSJInternational Conference on.IEEE, 2013:4363-4369 realizes track of vehicle forecasting techniques in conjunction with trajectory predictions technology and vehicle Activity recognition technology, vehicle Activity recognition technology comprises and keeps track, changing Lane and turning to, finite automaton state automata model can be used or drive context identification technology.This type of forecasting techniques only considers the vehicle of surrounding, and quantity is little, requires higher degree of accuracy.

Summary of the invention

Object of the present invention be exactly in order to overcome above-mentioned prior art exist defect and a kind of car online vehicles trend prediction method is provided.

Object of the present invention can be achieved through the following technical solutions:

A kind of car online vehicles trend prediction method, comprising:

1) the vehicle environmental information of acquisition residing for target vehicle and the traffic lights information in target vehicle front;

2) according to vehicle environmental information, and the vehicle behavior of the historic state information prediction target vehicle of target vehicle;

3) according to the vehicle behavior that traffic lights information and the prediction in vehicle environmental information, target vehicle front obtain, the state of the historic state information prediction vehicle of combining target vehicle.

Vehicle environmental information residing for described vehicle is specially:

Road＝(long _c,inserction,vehicles)

Wherein: Road is vehicle environmental information, long _cfor the distance between target vehicle and junction ahead, inserction is the junction ahead information of target vehicle, and vehicles is the front vehicles information of target vehicle;

Described junction ahead information is specially:

inserction＝(locations,lanes)

Wherein: locations is the positional information of junction ahead, lanes is the track data of each merging road in junction ahead place,

The positional information of described junction ahead is specially:

locations＝[(x ₁,y ₁,θ ₁),(x ₂,y ₂.θ ₂),(x ₃,y ₃,θ ₃),(x ₄,y ₄,θ ₄)]

Wherein: x _ifor the longitude on junction ahead i-th summit, y _ifor the latitude on junction ahead i-th summit, θ _ifor the angle of two merging roads in junction ahead i-th summit place, i herein represents the sequence number on summit, crossing,

The track data of each merging road in described junction ahead place are specially:

lanes＝[(forw ₁,rev ₁),(forw ₂,rev ₂),(forw ₃,rev ₃),(forw ₄,rev ₄)]

Wherein: forw _ifor the forward lane information of junction ahead i-th road, rev _ifor the reverse lane information of junction ahead i-th road, i herein represents the sequence number of road;

Described front vehicles information is specially:

vehicles＝[(S ₁,lane _S1),(S ₂,lane _S2),...(S _i,lane _Si),...(S _n,lane _Sn)]

Wherein: S _ifor the current state of i-th vehicle in front vehicles, lane _sifor the lane information in front vehicles residing for i-th vehicle;

Described traffic lights information is specially:

Light＝(sequence,long _L)

Wherein: Light is traffic lights information, long _lfor the distance between traffic lights and target vehicle, sequence be after traffic lights sequence in setting-up time Δ t, be specially:

sequence＝[(color ₁,time ₁),...(color _i，time _i),...(color _m，time _m)]

Wherein: time _ifor color _iduration, color _ifor the color state of traffic lights,

color _i∈{red,green,yellow}。

Described step 2) specifically comprise step:

201) judge whether the time that target vehicle walks to junction ahead is greater than setting-up time Δ t, if yes, then perform step 202), if NO, then perform step 205);

202) judge whether the density of target vehicle place lane preceding vehicle is greater than setting density threshold ρ *, if yes, then the predicting the outcome as lane change of vehicle behavior of target vehicle, and perform step 3), if NO, then perform step 203);

203) judge whether the velocity contrast of nearest vehicle in target vehicle and its front vehicles is greater than setting speed difference limen value δ v, if yes, then the predicting the outcome as lane change of vehicle behavior of target vehicle, and perform step 3), if NO, then step 204 is performed);

204) according to the acceleration and deceleration behavior of the range prediction vehicle of target vehicle and fore-aft vehicle, and step 3 is performed);

205) turning behavior of vehicle at crossing place is judged according to the historic state information of target vehicle.

Described step 204) specifically comprise step:

241) judge whether the distance in target vehicle and front vehicles between nearest vehicle is greater than safe distance, if yes, then perform step 243), if NO, then perform step 242);

242) judge whether the speed of target vehicle is greater than the speed of nearest vehicle in front vehicles, if yes, then target vehicle vehicle behavior predict the outcome for slow down, and perform step 3), if NO, then perform step 243);

243) judge whether the distance in target vehicle and front vehicle between nearest vehicle is greater than safe distance, if yes, then target vehicle vehicle behavior predict the outcome as at the uniform velocity, and perform step 3), if NO, then perform step 244);

244) judge whether the speed of target vehicle is greater than the speed of nearest vehicle in front vehicle, if yes, then target vehicle vehicle behavior predict the outcome as at the uniform velocity, and perform step 3), if NO, then target vehicle vehicle behavior predict the outcome for accelerate, and perform step 3);

Described step 242) in vehicle behavior predict the outcome for slow down time acceleration magnitude be:

$a_1=\mathrm{\σ}\frac{1}{d_{\mathrm{ij}}(v_i-v_j)}$

Wherein: a ₁acceleration magnitude during for predicting the outcome as slowing down, d _ijfor the distance between vehicle nearest in target vehicle and front vehicles, σ is constant, v _ifor the present speed size of target vehicle, v _jfor the present speed size of vehicle nearest in front vehicles.

Described step 244) in vehicle behavior predict the outcome for accelerate time acceleration magnitude be:

$a_2=-a\frac{1}{d_{\mathrm{ij}}(v_i-v_j)}+\mathrm{\β}\frac{(v_t-v_i)}{d_{\mathrm{it}}}$

Wherein: a ₂acceleration magnitude during for predicting the outcome as accelerating, d _ijfor the distance between vehicle nearest in target vehicle and front vehicles, α, β are constant, v _ifor the present speed size of target vehicle, v _jfor the present speed size of vehicle nearest in front vehicles, v _tfor the present speed size of vehicle nearest in front vehicle.

Described step 205) specifically comprise step:

251) travel through the historic state information of target vehicle from the current state of target vehicle forward, and determine start offset state TurnPoint, be specially:

TurnPoint＝min({g|Offset _k＞T&Offset _k-1＞T&,...,Offset _g＞T})

Wherein:

Offset _i＝(OffsetP _i,OffsetV _i,OffsetA _i)

T＝(TP,TV,TA)

Wherein:

OffsetP _i, OffsetV _i, OffsetA _ibe respectively the side-play amount of position in the i-th-1 state to the i-th state, speed and acceleration, TP, TV, TA are constant;

252) judge whether continue skew number of times exceedes set point number: and whether acceleration direction migration included angle meets circular motion characteristic, is specially judgment formula:

k-TurnPoint＞TK

w(OffsetA _k)＜TW&w(OffsetA _k-1)＜TW,...,w(OffsetA _TurnPoint+1)＜TW

Wherein: k-TurnPoint is for continuing skew number of times, and TK is set point number, and TW is deviation angle tolerable error,

If the determination result is YES, then target of prediction vehicle is turned, and turn direction is the offset direction of target vehicle current state position, if judged result is no, then target of prediction vehicle is not turned.

Compared with prior art, the present invention has the following advantages:

1) the present invention is based on vehicle environmental information and the vehicle behavior of vehicle historic state to vehicle is predicted, predicted by the to-be of vehicle behavior to vehicle again, prediction based on vehicle behavior is more accurate, and there is extendability well, simple efficient, be suitable for large scale network situation.

2) vehicle behaviour decision making, is judged the acceleration of vehicle, deceleration by conditional decision, at the uniform velocity, is changed trains and the behavior such as turning, reaches the object of the prediction of vehicle-state under the complicated road in predicted city in conjunction with current vehicle condition information and road information.

3) vehicle behaviour decision making distinguishes straight way and crossing process, straight way considers the impact of surrounding vehicles, use for reference the method to fish school behavior simulation in artificial fish-swarm algorithm, vehicle behavior is simulated, modeling is carried out to vehicle's current condition and surrounding vehicles environment, derives effective conditional decision and express formula.Whether crossing situation, by carrying out modeling to the skew of vehicle historical position, speed and acceleration, is derived vehicle and is turned and the conditional decision of turn direction.

Accompanying drawing explanation

Fig. 1 is key step schematic flow sheet of the present invention;

Fig. 2 is crossing place schematic diagram;

Fig. 3 is that crossing turning time calculates schematic diagram.

Embodiment

Below in conjunction with the drawings and specific embodiments, the present invention is described in detail.The present embodiment is implemented premised on technical solution of the present invention, give detailed embodiment and concrete operating process, but protection scope of the present invention is not limited to following embodiment.

A kind of car online vehicles trend prediction method, as shown in Figure 1, comprising:

1) the vehicle environmental information of acquisition residing for target vehicle and the traffic lights information in target vehicle front;

Vehicle environmental information residing for vehicle is specially:

Road＝(long _c,inserction,vehicles)

Wherein: Road is vehicle environmental information, long _cfor the distance between target vehicle and junction ahead, inserction is the junction ahead information of target vehicle, and vehicles is the front vehicles information of target vehicle;

As shown in Figure 2, in the present embodiment, junction ahead is modeled as parallelogram, and junction ahead information is specially:

inserction＝(locations,lanes)

Wherein: locations is the positional information of junction ahead, lanes is the track data of each merging road in junction ahead place,

The positional information of junction ahead is specially:

locations＝[(x ₁,y ₁,θ ₁),(x ₂,y ₂.θ ₂),(x ₃,y ₃,θ ₃),(x ₄,y ₄,θ ₄)]

Wherein: x _ifor the longitude on junction ahead i-th summit, y _ifor the latitude on junction ahead i-th summit, θ _ifor the angle of two merging roads in junction ahead i-th summit place, i ∈ { 1,2,3,4}, wherein θ ₁as shown in Figure 2,

The track data of each merging road in junction ahead place are specially:

lanes＝[(forw ₁,rev ₁),(forw ₂,rev ₂),(forw ₃,rev ₃),(forw ₄,rev ₄)]

Wherein: forw _ifor the forward lane information of junction ahead i-th road, rev _ifor the reverse lane information of junction ahead i-th road, i ∈ { 1,2,3,4}, if the junction of three roads, can be 0 by the forward track of a road in setting four roads and reverse number of lanes, the forward in track wherein and can be oppositely set according to the travel direction of target vehicle, also can move towards setting according to road;

Front vehicles information is specially:

vehicles＝[(S ₁,lane _S1),(S ₂,lane _S2),...(S _i,lane _Si),...(S _n,lane _Sn)]

Wherein: S _ifor the current state of i-th vehicle in front vehicles, lane _sifor the lane information in front vehicles residing for i-th vehicle, the integer of i to be span be 1 ~ n;

Traffic lights information is specially:

Light＝(sequence,long _L)

Wherein: Light is traffic lights information, long _lfor the distance between traffic lights and target vehicle, sequence be after traffic lights sequence in setting-up time Δ t, be specially:

sequence＝[(color ₁,time ₁),...(color _i，time _i),...(color _m，time _m)]

Wherein: time _ifor color _iduration, color _ifor the color state of traffic lights,

Color _i∈ { the integer of red, green, yellow}, i to be span be 1 ~ m.

2) according to vehicle environmental information, and the vehicle behavior of the historic state information prediction target vehicle of target vehicle, specifically comprise step:

201) judge whether the time that target vehicle walks to junction ahead is greater than setting-up time Δ t, if yes, then judges that vehicle travels on and knows, and perform step 202), if NO, then judge that vehicle travels on crossing, and perform step 205);

202) (ρ * is specially and sets as required to judge the density of the front vehicles in the visual range of track, target vehicle place whether to be greater than setting density threshold ρ *, can be empirical value), if yes, then target vehicle vehicle behavior predict the outcome as lane change, and perform step 3), if NO, then step 203 is performed);

203) judge whether the velocity contrast of nearest vehicle in target vehicle and its front vehicles is greater than setting speed difference limen value δ v, if yes, then the predicting the outcome as lane change of vehicle behavior of target vehicle, and perform step 3), if NO, then step 204 is performed);

204) according to the acceleration and deceleration behavior of the range prediction vehicle of target vehicle and fore-aft vehicle, in the present embodiment, in order to simplify forecasting process, when without the need to carrying out acceleration and deceleration, the vehicle behavior of target vehicle predict the outcome as at the uniform velocity, and perform step 3), step 204) specifically comprise step:

241) judge whether the distance in target vehicle and front vehicles between nearest vehicle is greater than safe distance, if yes, then perform step 243), if NO, then perform step 242);

242) judge whether the speed of target vehicle is greater than the speed of nearest vehicle in front vehicles, if yes, then target vehicle vehicle behavior predict the outcome for slow down, and perform step 3), if NO, then perform step 243);

Wherein vehicle behavior predict the outcome for slow down time acceleration magnitude be:

$a_1=\mathrm{\σ}\frac{1}{d_{\mathrm{ij}}(v_i-v_j)}$

Wherein: a ₁acceleration magnitude during for predicting the outcome as slowing down, d _ijfor the distance between vehicle nearest in target vehicle and front vehicles, σ is constant, v _ifor the present speed size of target vehicle, v _jfor the present speed size of vehicle nearest in front vehicles.

243) judge whether the distance in target vehicle and front vehicle between nearest vehicle is greater than safe distance, if yes, then target vehicle vehicle behavior predict the outcome as at the uniform velocity, and perform step 3), if NO, then perform step 244);

244) judge whether the speed of target vehicle is greater than the speed of nearest vehicle in front vehicle, if yes, then target vehicle vehicle behavior predict the outcome as at the uniform velocity, and perform step 3), if NO, then target vehicle vehicle behavior predict the outcome for accelerate, and perform step 3);

Wherein vehicle behavior predict the outcome for accelerate time acceleration magnitude be:

$a_2=-a\frac{1}{d_{\mathrm{ij}}(v_i-v_j)}+\mathrm{\β}\frac{(v_t-v_i)}{d_{\mathrm{it}}}$

Wherein: a ₂acceleration magnitude during for predicting the outcome as accelerating, d _ijfor the distance between vehicle nearest in target vehicle and front vehicles, α, β are constant, v _ifor the present speed size of target vehicle, v _jfor the present speed size of vehicle nearest in front vehicles, v _tfor the present speed size of vehicle nearest in front vehicle.

205) judge the turning behavior of vehicle at crossing place according to the historic state information of target vehicle, specifically comprise step, specifically comprise step:

251) travel through the historic state information of target vehicle from the current state of target vehicle forward, and determine start offset state TurnPoint, be specially:

TurnPoint＝min({g|Offset _k＞T&Offset _k-1＞T&,...,Offset _g＞T})

Wherein:

Offset _i＝(OffsetP _i,OffsetV _i,OffsetA _i)

T＝(TP,TV,TA)

Wherein:

OffsetP _i, OffsetV _i, OffsetA _ibe respectively the side-play amount of position in the i-th-1 state to the i-th state, speed and acceleration, TP, TV, TA are constant;

252) judge whether continue skew number of times exceedes set point number: and whether acceleration direction migration included angle meets circular motion characteristic, is specially judgment formula:

k-TurnPoint＞TK

w(OffsetA _k)＜TW&w(OffsetA _k-1)＜TW,...,w(OffsetA _TurnPoint+1)＜TW

Wherein: k-TurnPoint is for continuing skew number of times, and TK is set point number, and TW is deviation angle tolerable error, and k represents the sequence number of target vehicle current state,

If the determination result is YES, then target of prediction vehicle is turned, and turn direction is the offset direction of target vehicle current state position, if judged result is no, then target of prediction vehicle is not turned.

253) turning time of origin t is predicted _turn.By acceleration angular deflection, calculate turning rate ω, by crossing angle calcu-lation deviation angle, thus estimation is turned the time occurred.As shown in Figure 3, the angular deflection occurred when turning from crossing is 180 degree and deducts crossing angle, namely

ω＝(w(OffsetA _k)+w(OffsetA _k-1)+,....,w(OffsetA _TurnPoint+1))/(k-TurnPoint)Δt

t _turn＝(180°-θ)/ω

254) at turning time of origin t _turninterior acceleration, speed and change in location, calculate according to circular motion rule, acceleration magnitude is constant, and direction at the uniform velocity offsets, and speed is obtained by acceleration and Time Calculation, and position is obtained by speed and Time Calculation.

3) according to the vehicle behavior that traffic lights information and the prediction in vehicle environmental information, target vehicle front obtain, the state of the historic state information prediction vehicle of combining target vehicle, is specially:

s _k+1＝g(s _k,Road,Light,Behavior)

Wherein: s _k+1for predicting the vehicle-state obtained, s _kfor the vehicle-state that target vehicle is current, Behavior predicts the vehicle behavior obtained.

Claims (8)

1. a car online vehicles trend prediction method, is characterized in that, comprising:

1) the vehicle environmental information of acquisition residing for target vehicle and the traffic lights information in target vehicle front;

2) according to vehicle environmental information, and the vehicle behavior of the historic state information prediction target vehicle of target vehicle;

3) according to the vehicle behavior that traffic lights information and the prediction in vehicle environmental information, target vehicle front obtain, the state of the historic state information prediction vehicle of combining target vehicle.

2. a kind of car online vehicles trend prediction method according to claim 1, it is characterized in that, the vehicle environmental information residing for described vehicle is specially:

Road＝(long _c,inserction,vehicles)

Wherein: Road is vehicle environmental information, long _cfor the distance between target vehicle and junction ahead, inserction is the junction ahead information of target vehicle, and vehicles is the front vehicles information of target vehicle;

Described junction ahead information is specially:

inserction＝(locations,lanes)

Wherein: locations is the positional information of junction ahead, lanes is the track data of each merging road in junction ahead place,

The positional information of described junction ahead is specially:

locations＝[(x ₁,y ₁,θ ₁),(x ₂,y ₂.θ ₂),(x ₃,y ₃,θ ₃),(x ₄,y ₄,θ ₄)]

Wherein: x _ifor the longitude on junction ahead i-th summit, y _ifor the latitude on junction ahead i-th summit, θ _ifor the angle of two merging roads in junction ahead i-th summit place,

The track data of each merging road in described junction ahead place are specially:

lanes＝[(forw ₁,rev ₁),(forw ₂,rev ₂),(forw ₃,rev ₃),(forw ₄,rev ₄)]

Wherein: forw _ifor the forward lane information of junction ahead i-th road, rev _ifor the reverse lane information of junction ahead i-th road;

Described front vehicles information is specially:

vehicles＝[(S ₁,lane _S1),(S ₂,lane _S2),…(S _i,lane _Si),…(S _n,lane _Sn)]

Wherein: S _ifor the current state of i-th vehicle in front vehicles, lane _sifor the lane information in front vehicles residing for i-th vehicle.

3. a kind of car online vehicles trend prediction method according to claim 2, it is characterized in that, described traffic lights information is specially:

Light＝(sequence,long _L)

Wherein: Light is traffic lights information, long _lfor the distance between traffic lights and target vehicle, sequence be after traffic lights sequence in setting-up time Δ t, be specially:

sequence＝[(color ₁,time ₁),…(color _i，time _i),…(color _m，time _m)]

Wherein: time _ifor color _iduration, color _ifor the color state of traffic lights,

color _i∈{red,green,yellow}。

4. a kind of car online vehicles trend prediction method according to claim 3, is characterized in that, described step 2) specifically comprise step:

201) judge whether the time that target vehicle walks to junction ahead is greater than setting-up time Δ t, if yes, then perform step 202), if NO, then perform step 205);

202) judge whether the density of target vehicle place lane preceding vehicle is greater than setting density threshold ρ *, if yes, then the predicting the outcome as lane change of vehicle behavior of target vehicle, and perform step 3), if NO, then perform step 203);

203) judge whether the velocity contrast of nearest vehicle in target vehicle and its front vehicles is greater than setting speed difference limen value δ v, if yes, then the predicting the outcome as lane change of vehicle behavior of target vehicle, and perform step 3), if NO, then step 204 is performed);

204) according to the acceleration and deceleration behavior of the range prediction vehicle of target vehicle and fore-aft vehicle, and step 3 is performed);

205) turning behavior of vehicle at crossing place is judged according to the historic state information of target vehicle.

5. a kind of car online vehicles trend prediction method according to claim 4, is characterized in that, described step 204) specifically comprise step:

241) judge whether the distance in target vehicle and front vehicles between nearest vehicle is greater than safe distance, if yes, then perform step 243), if NO, then perform step 242);

242) judge whether the speed of target vehicle is greater than the speed of nearest vehicle in front vehicles, if yes, then target vehicle vehicle behavior predict the outcome for slow down, and perform step 3), if NO, then perform step 243);

243) judge whether the distance in target vehicle and front vehicle between nearest vehicle is greater than safe distance, if yes, then target vehicle vehicle behavior predict the outcome as at the uniform velocity, and perform step 3), if NO, then perform step 244);

244) judge whether the speed of target vehicle is greater than the speed of nearest vehicle in front vehicle, if yes, then target vehicle vehicle behavior predict the outcome as at the uniform velocity, and perform step 3), if NO, then target vehicle vehicle behavior predict the outcome for accelerate, and perform step 3).

6. a kind of car online vehicles trend prediction method according to claim 5, is characterized in that, described step 242) in vehicle behavior predict the outcome for slow down time acceleration magnitude be:

$a_1=\mathrm{\σ}\frac{1}{d_{\mathrm{ij}}(v_i-v_j)}$

Wherein: a ₁acceleration magnitude during for predicting the outcome as slowing down, d _ijfor the distance between vehicle nearest in target vehicle and front vehicles, σ is constant, v _ifor the present speed size of target vehicle, v _jfor the present speed size of vehicle nearest in front vehicles.

7. a kind of car online vehicles trend prediction method according to claim 5, is characterized in that, described step 244) in vehicle behavior predict the outcome for accelerate time acceleration magnitude be:

$a_2=-\mathrm{\α}\frac{1}{d_{\mathrm{ij}}(v_i-v_j)}+\mathrm{\β}\frac{(v_t-v_i)}{d_{\mathrm{it}}}$

Wherein: a ₂acceleration magnitude during for predicting the outcome as accelerating, d _ijfor the distance between vehicle nearest in target vehicle and front vehicles, α, β are constant, v _ifor the present speed size of target vehicle, v _jfor the present speed size of vehicle nearest in front vehicles, v _tfor the present speed size of vehicle nearest in front vehicle.

8. a kind of car online vehicles trend prediction method according to claim 4, is characterized in that, described step 205) specifically comprise step:

251) travel through the historic state information of target vehicle from the current state of target vehicle forward, and determine start offset state TurnPoint, be specially:

TurnPoint＝min({g|Offset _k＞T&Offset _k-1＞T&,…,Offset _g＞T})

Wherein:

Offset _i＝(OffsetP _i,OffsetV _i,OffsetA _i)

T＝(TP,TV,TA)

Wherein:

OffsetP _i, OffsetV _i, OffsetA _ibe respectively the side-play amount of position in the i-th-1 state to the i-th state, speed and acceleration, TP, TV, TA are constant;

252) judge whether continue skew number of times exceedes set point number: and whether acceleration direction migration included angle meets circular motion characteristic, is specially judgment formula:

k-TurnPoint＞TK

w(OffsetA _k)＜TW&w(OffsetA _k-1)＜TW,…,w(OffsetA _TurnPoint+1)＜TW

Wherein: k-TurnPoint is for continuing skew number of times, and TK is set point number, and TW is deviation angle tolerable error,

If the determination result is YES, then target of prediction vehicle is turned, and turn direction is the offset direction of target vehicle current state position, if judged result is no, then target of prediction vehicle is not turned.