CN106548645A - Vehicle route optimization method and system based on deep learning - Google Patents
Vehicle route optimization method and system based on deep learning Download PDFInfo
- Publication number
- CN106548645A CN106548645A CN201610976649.6A CN201610976649A CN106548645A CN 106548645 A CN106548645 A CN 106548645A CN 201610976649 A CN201610976649 A CN 201610976649A CN 106548645 A CN106548645 A CN 106548645A
- Authority
- CN
- China
- Prior art keywords
- data
- network model
- belief network
- depth belief
- path
- 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
Links
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/09—Arrangements for giving variable traffic instructions
- G08G1/0962—Arrangements for giving variable traffic instructions having an indicator mounted inside the vehicle, e.g. giving voice messages
- G08G1/0968—Systems involving transmission of navigation instructions to the vehicle
- G08G1/096805—Systems involving transmission of navigation instructions to the vehicle where the transmitted instructions are used to compute a route
- G08G1/096811—Systems involving transmission of navigation instructions to the vehicle where the transmitted instructions are used to compute a route where the route is computed offboard
- G08G1/096816—Systems involving transmission of navigation instructions to the vehicle where the transmitted instructions are used to compute a route where the route is computed offboard where the complete route is transmitted to the vehicle at once
Landscapes
- Engineering & Computer Science (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Traffic Control Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of vehicle route optimization method and system based on deep learning, the method includes obtaining Real-time Road data and history road data, the data of acquisition is carried out the data set of tape label is formed after pretreatment；Construction depth belief network model, trains depth belief network model；The depth belief network model completed using training is predicted to the All Paths of vehicle to destination, exports the congestion coefficient in each path；According to congestion coefficient and apart from this two indexs come comprehensive assessment path, optimal path is exported；Optimal path is for congestion coefficient and apart from the path corresponding to this two index linear superpositions minimums.The present invention can obtain information needed from the highway traffic data of various dimensions by the powerful feature extraction functions of depth belief network, reduce interference, accurate reasonable prediction is made to jam situation, improve and seek footpath efficiency, reduce human error, be that mitigation salvaging has saved valuable time.
Description
Technical field
The present invention relates to a kind of vehicle route optimization method and system based on deep learning, in particular for special vehicle
Road distance and the equilibrium problem between the distance time are weighed under the complexity traffic environment of city.
Background technology
In past half a century, Vehicle Routing Problem-it is directly one of focus of field of traffic research, with traffic flow
Being significantly increased for amount, has no longer been focal point apart from.Need more to consider a problem from time angle.In real traffic
In, as the uncertainty (difference of the travel amount of the people of different time points) of demand and the uncertainty of supply are (as handed over
The decline of the traffic capacity in the section that interpreter's event, weather reason, road maintenance etc. are caused), the journey time in section is one
Vary in fixed scope.The uncertainty of travel time greatly have impact on the reliability of traffic system and cause mostly
Several congestions, also contributes to the Route choice behavior of people.
For vehicle driving route planning modeling and optimization, particularly with special vehicle task (such as：Put out a fire, rescue people, go out
Alert, emergency repair), the planing method of traditional circuit is carried out under the constant hypothesis of network structure mostly：Assume traffic shape
Condition is certain, sets up traffic network design, then by solving shortest path model, draws circuit.Method by emulating,
Image Via Gis (GIS), it is considered to the model based on Petri network is set up after factors.But these methods are not all right
The reliability of travel route is analysed in depth, and the model of foundation, method and theory are also limited to the side of traditional route optimization
Method and thinking.And in practical problem, reliability is often more important than effectiveness, so traditional method will be improved, with full
Sufficient current demand.
The content of the invention
In order to solve the shortcoming of prior art, the first object of the present invention is to provide a kind of vehicle road based on deep learning
Footpath optimization method.The vehicle route optimization method based on deep learning is somebody's turn to do, including：
Real-time Road data and history road data are obtained, the data of acquisition are carried out tape label is formed after pretreatment
Data set；
Construction depth belief network model, using the data set of tape label, Real-time Road data and history road data
To train depth belief network model；
The depth belief network model completed using training is predicted to the All Paths of vehicle to destination, and output is each
The congestion coefficient in individual path；
According to congestion coefficient and apart from this two indexs come comprehensive assessment path, optimal path is exported；The optimal path
For congestion coefficient and apart from the path corresponding to this two index linear superposition minimums.
The vehicle route optimization method based on deep learning of the present invention is by the powerful feature extraction of depth belief network
Function, can obtain information needed from the highway traffic data of various dimensions, reduce the interference of redundant data, jam situation is done
Go out accurate reasonable prediction, footpath efficiency is sought in raising；
The present invention's applies also for the special vehicle with special duty based on the vehicle route optimization method of deep learning,
Such as putting out a fire, rescuing people, responding, emergency repair, these vehicles are when accident occurs, each to need accident to be reached now with prestissimo
, the method for the present invention can make accurate reasonable prediction to jam situation, improve and seek footpath efficiency, reduce human error,
Valuable time is saved for mitigation salvaging, has selected an optimal travel route improve rescue efficiency, maximum journey
Safeguard people life property safety degree.
Preferably, the process that the data of acquisition carry out pretreatment is included：
The data of acquisition are normalized；
When the data for obtaining have missing values, using average interpolation method polishing, arrangement form road in a fixed order
Characteristic vector.
The method of the present invention is quantified by the road data to obtaining and normalization, can be according to reality by road
Data are finely divided label, further form the data set with label, are more accurately to train depth belief network model to establish
Basis is determined.
Depth belief network model is made up of the limited Boltzmann machine of multilamellar and one layer of reverse transmittance nerve network；Wherein,
Reverse transmittance nerve network is located at top layer, and the activation primitive of depth belief network model adopts PReLU functions.The depth of the present invention
Degree belief network model is using PReLU as activation primitive, it is to avoid the impact that gradient disappears, and accelerates convergence rate, improves
The efficiency of depth belief network model training and subsequent path optimizing.
Preferably, in training depth belief network model during the limited Boltzmann machine of any layer, by history road
Circuit-switched data trains the weights of any layer limited Boltzmann machine by non-supervisory greedy successively method as training sample, this
Sample can reduce computing cost, quickly obtain the model parameter of depth belief network model.
Preferably, input of the output of last layer of limited Boltzmann machine as depth belief network model top layer.
Preferably, depth belief network model top layer passes through the data set of tape label to the mould in depth belief network model
Shape parameter is finely adjusted, and obtains ID belief network model.
Preferably, the model parameter in depth belief network model is finely adjusted using error backpropagation algorithm.Can
So that model is preferably fitted training sample and has higher generalization ability,
Preferably, using the dual average algorithm RDA of regularization and ID belief network model, to Real-time Road data
On-line training is carried out, depth belief network model is continued to optimize.The present invention is trained to model online using RDA algorithms, is led to
Cross and continue to optimize parameter, make model prediction result more accurate.
The second object of the present invention there is provided a kind of vehicle route optimization system based on deep learning, should be based on depth
A kind of example structure of the vehicle route optimization system of study is：
A kind of vehicle route optimization system based on deep learning includes：Data acquisition and processing module, which is used to obtain
Real-time Road data and history road data, the data of acquisition are carried out the data set of tape label is formed after pretreatment；
Depth belief network model training module, which is used for construction depth belief network model, using the data of tape label
Collection, Real-time Road data and history road data are training depth belief network model；
Path congestion coefficients calculation block, which is used for using the depth belief network model that completes of training to vehicle to purpose
The All Paths on ground are predicted, and export the congestion coefficient in each path；
Optimal path computation module, which is used for according to congestion coefficient and apart from this two indexs come comprehensive assessment path, defeated
Go out optimal path；The optimal path is for congestion coefficient and apart from the path corresponding to this two index linear superpositions minimums.
The vehicle route optimization system based on deep learning of the present invention is by the powerful feature extraction of depth belief network
Function, can obtain information needed from the highway traffic data of various dimensions, reduce interference, jam situation be made accurately rationally
Footpath efficiency is sought in prediction, raising.
Further, data acquisition and processing module also include：
Normalization module, which is used to be normalized the data of acquisition；
Roadway characteristic vector forms module, and which is used for when the data for obtaining have missing values, is mended using average interpolation method
Together, arrangement form roadway characteristic is vectorial in a fixed order.
Present invention also offers based on another kind of example structure of the vehicle route optimization system of deep learning being：
Vehicle route optimization system of the another kind of the present invention based on deep learning, including：
Data acquisition unit, which is configured to gather Real-time Road data and history road data；
The data acquisition unit is connected with server, and the server is configured to：
Real-time Road data and history road data are obtained, the data of acquisition are carried out tape label is formed after pretreatment
Data set；
Construction depth belief network model, using the data set of tape label, Real-time Road data and history road data
To train depth belief network model；
The depth belief network model completed using training is predicted to the All Paths of vehicle to destination, and output is each
The congestion coefficient in individual path；
According to congestion coefficient and apart from this two indexs come comprehensive assessment path, optimal path is exported；The optimal path
For congestion coefficient and apart from the path corresponding to this two index linear superposition minimums.
The vehicle route optimization system being somebody's turn to do based on deep learning of the present invention is carried by the powerful feature of depth belief network
Function is taken, information needed can be obtained from the highway traffic data of various dimensions, reduce interference, accurately conjunction is made to jam situation
Footpath efficiency is sought in reason prediction, raising.
The system also includes display device, and which is connected with server, and is display configured to optimal path.
Beneficial effects of the present invention are：
(1) the method for the invention, can be from the road of various dimensions by the powerful feature extraction functions of depth belief network
Information needed is obtained in the traffic data of road, redundant data interference is reduced, accurate reasonable prediction is made to jam situation, footpath is sought in raising
Efficiency.
(2) the method for the invention applies also for the special vehicle with special duty, such as puts out a fire, rescues people, responding, answers
Suddenly speedily carry out rescue work, when accident occurs, each to need to reach the scene of the accident with prestissimo, the method for the present invention can be right for these vehicles
Jam situation makes accurate reasonable prediction, improves and seeks footpath efficiency, reduces human error, is that mitigation salvaging has saved preciousness
Time, select an optimal travel route can improve rescue efficiency, farthest safeguard people life property safety.
(3) the method for the invention in conventional depth belief network using PReLU as activation primitive, it is to avoid gradient
The impact of disappearance, accelerates convergence rate, improves the efficiency of algorithm；Online model is trained using RDA algorithms, is passed through
Parameter is continued to optimize, makes model prediction result more accurate.
(4) the of the invention vehicle route optimization system being somebody's turn to do based on deep learning is by the powerful feature of depth belief network
Abstraction function, can obtain information needed from the highway traffic data of various dimensions, reduce interference, jam situation is made accurately
Reasonable prediction, raising seek footpath efficiency.
Description of the drawings
Fig. 1 is a kind of schematic flow sheet of vehicle route optimization method based on deep learning of the present invention；
Fig. 2 is the depth belief network model schematic of the present invention；
Fig. 3 is the structural representation of the vehicle route optimization system embodiment one based on deep learning of the present invention；
Fig. 4 is the structural representation of the vehicle route optimization system embodiment two based on deep learning of the present invention.
Specific embodiment
Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, complete
Site preparation is described.Fig. 1 is a kind of schematic flow sheet of vehicle route optimization method based on deep learning of the present invention, as shown in the figure
Can be included based on the vehicle route optimization method of deep learning in the present embodiment：
S101, obtains Real-time Road data and history road data, forms band after the data of acquisition are carried out pretreatment
The data set of label.
In implementing, the road data of a large amount of real-time and history can provide safeguard for model prediction accuracy rate afterwards.
Real-time Road data include：
(1) by arranging the average speed and vehicle flowrate data that sensing equipment obtains response section in real time in each section.
(2) by the video data in monitoring device Real-time Collection track.
(3) by gas such as the visibility in each section of meteorological information acquisition equipment Real-time Collection, temperature, humidity, wind-force, sleet
Wait data, Comprehensive Assessment weather alert grade.
History road data includes：
(1) the history average speed and history vehicle flowrate data in response section are obtained.
(2) the history video data in track.
(3) the climatic data such as the history visibility in each section, temperature, humidity, wind-force, sleet, Comprehensive Assessment weather alert etc.
Level.
As a example by the special vehicle of special duty is taken the post as：
In view of the sudden and randomness of special assignment, responding period and war preparedness period can be divided within one day, for not
With the frequency difference of period gathered data, it is specifically described below.
In the war preparedness period, the interval of delta t of gathered data is set to 10 minutes；
In the responding period, the interval of delta t of gathered data is set to 1 minute；
Road data mainly includes vehicle number of the unit interval by crossing, Emergency Vehicle Lane occupancy situation (take as 0.9,
Without car for 0.1), weather alert quantify (without early warning be 0.1, blue early warning be 0.3, yellow early warning be that 0.5, orange early warning is
0.7th, red early warning is for 0.9), period factors quantization (working day morning peak be 0.9, evening peak be 0.7, festivals or holidays be 0.6), its
His factor (it is 0.9 that road maintenance is 0.7, vehicle accident 0.8, march-rally), listed numerical value is not unique, can be according to locality
Practical situation is adjusted flexibly numerical value.
The concrete grammar of above-mentioned data prediction is：
The road data for obtaining is quantified and normalization, leading indicator include unit interval crossing vehicle pass-through number,
Average speed, Emergency Vehicle Lane occupancy situation, weather alert grade, time factor, other factors etc..
Gained index composition characteristic vector, missing values are adopted into average interpolation method polishing, in a fixed order arrangement form
Roadway characteristic vector.
Data acquisition intervals are Δ t, can flexible as the case may be.
Below as a example by using road video monitoring information：
Artificially congestion in road situation is graded, as label (unimpeded is 0, and crowded is 1, is blocked as 2, is blocked as 3),
Merely just give an example, can this makes it possible to form the data set of tape label according to actual subdivision label.
S102, construction depth belief network model, using the data set of tape label, Real-time Road data and history road
Data are training depth belief network model.
In implementing, depth belief network is a generative probabilistic model, sets up one between training data and label
Joint Distribution.Depth belief network model is by the limited Boltzmann machine RBM of multilamellar and one layer of reverse transmittance nerve network BP group
Into；Wherein, reverse transmittance nerve network BP is located at top layer, and the activation primitive of depth belief network model adopts PReLU functions.
In training depth belief network model during the limited Boltzmann machine of any layer, history road data is made
For training sample, the weights of the limited Boltzmann machine of any layer are trained by non-supervisory greedy successively method, depth letter is obtained
Read the model parameter of network model.
By the weights of non-supervisory greedy successively method pre-training model, mainly include the following steps that：
(1) first RBM is trained up；
(2) the weight and side-play amount of first RBM are fixed, then using the state of its recessive neuron, as second
The input vector of RBM；
(3), after training up second RBM, second RBM is stacked on into the top of first RBM；
(4) it is multiple that three above step is repeated；
(5) top layer BP neural network is used to predict classification as layer is returned.
Input of the output of last layer of limited Boltzmann machine as depth belief network model top layer.
Depth belief network model top layer is by the data set of tape label to the model parameter in depth belief network model
It is finely adjusted, obtains ID belief network model.
The model parameter in depth belief network model is finely adjusted using error backpropagation algorithm.
Model using tape label data set using error backpropagation algorithm to estimated performance tuning, concrete grammar include with
Lower step：
(1), when the BP neural network of top layer is trained, dominant neurologic unit will be carried out together with the neuron for representing tag along sort
Training, to each training data, corresponding label neuron is opened and is set to 1, and others are then closed and are set to 0；
(2) BP neural network propagates to each layer of RBM by error message top-down, finely tunes whole DBN networks；
(3) Wake stages：Cognitive process, weight (cognitive weight) by extraneous feature and upwards produce each layer
Abstract representation (node state), and decline the descending weight (generation weight) of modification interlayer using gradient.
(4) Sleep stages：Generating process, is represented and downward weight by top layer, generates the state of bottom, while modifying layer
Between weight upwards.
Specifically, as shown in Fig. 2 the process of training depth belief network model DBN is carried out layer by layer.In each layer
In, infer hidden layer with data vector, then this hidden layer as next layer of data vector.
First RBM being trained up first, showing layer corresponding to input, hidden layer corresponds to feature detection, their joint group
State energy equation is：
V in formulaiAnd hjI-th input and j-th feature are represented respectively；biAnd αjIt is their side-play amount respectively；ωijFor
Their weight matrix, n represent the number of input, and n is the positive integer more than or equal to 1；V and h is respectively input into set and spy
Collection is closed.Each input is input into set i.e. the combination in all paths namely per paths.
When v or h is given, their conditional probability distribution can be calculated：
Give one group of data { Vc| c ∈ { 1,2 ..., C } }, maximize the log-likelihood function of this model：
Parameter ω is approximately tried to achieve by gibbs sampler methodij, bi, aj, its update rule be：
Δωij=εω(Edata[vihj]-Emod[vihj])
Δbi=εb(Edata[vi]-Emod[hj])
Δaj=εa(Edata[hj]-Emod[hj])
In formula：εω、εb、εaThe learning rate constant being in the range of [0,1], EdataBe initial model hidden layer expect it is defeated
Go out, EmodIt is the desired output estimated by non-supervisory greedy method.
Using PReLU as activation primitive, its form is each neural unit：
Wherein yiRepresent input, { αi|0<αi≤ 1 } represent learning rate constant.
The weight and side-play amount of first RBM are fixed afterwards, using the state of its recessive neuron, as second RBM
Input vector.
After training up second RBM, the top that second RBM is stacked on first RBM by that analogy, is successively instructed
Practice, obtain a depth Boltzmann machine, add afterwards and return layer composition depth belief network.
Input of the output of last RBM as top layer, top layer enter parameter adjustment by the training set for having label again.It is main
Comprise the following steps：
(1) label data is chosen, depth belief network is trained with back-propagation algorithm, calculate each layer output；
(2) reconstructed error amendment weights and the establishment according to each layer；
(3) determine whether model meets requirement according to predictablity rate, if can not if repeat step (1) and (2), Zhi Daomo
Type output result meets to be expected to require.
Using the dual average algorithm RDA of regularization and ID belief network model, Real-time Road data are carried out
Line training, continues to optimize depth belief network model.
On-line training is carried out to Real-time Road data using RDA algorithms, its weight more new formula is：
Wherein<G(t), W>Represent G(t)Integral mean to W；ψ (W) is regular terms；For extra regular terms,
It is a strictly convex function.
S103, the depth belief network model completed using training are predicted to the All Paths of vehicle to destination,
Export the congestion coefficient in each path.
The road data on vehicle to the All Paths of destination is obtained, the depth belief network mould for completing is input into training
Type, exports the congestion coefficient in each path.
All Paths of the exhaustive special vehicle to accident spot, predict gathering around for all roads using the model ergod for training
Stifled coefficient.
The special circumstances such as Emergency Vehicle Lane can be driven in the wrong direction and are taken for special vehicle, and model can increase the weight of Emergency Vehicle Lane
And predict the congestion coefficient of twocouese on route.
S104, according to congestion coefficient and apart from this two indexs come comprehensive assessment path, exports optimal path；The optimum
Path is for congestion coefficient and apart from the path corresponding to this two index linear superpositions minimums.
The linear superposition of path congestion coefficient and distance is calculated, concrete formula is as follows：
In formula, α and β are weight coefficients, add up to 1；EiRepresent the congestion coefficient of i-th input, that is, the i-th paths
Congestion coefficient, DiFor the distance of the i-th paths.
Optimal path Lbest=min (L).
The present invention, can be from the highway traffic data of various dimensions by the powerful feature extraction functions of depth belief network
Information needed is obtained, redundant data interference is reduced, accurate reasonable prediction is made to jam situation, footpath efficiency is sought in raising.
The present disclosure additionally applies for the special vehicle with special duty, such as puts out a fire, rescues people, responding, emergency repair, these
Vehicle when accident occurs, each to need to reach the scene of the accident with prestissimo, can do to jam situation by the method for the present invention
Go out accurate reasonable prediction, improve and seek footpath efficiency, reduce human error, be that mitigation salvaging has saved valuable time, choosing
Select an optimal travel route and can improve rescue efficiency, farthest safeguard people life property safety.
Fig. 3 be the present invention the vehicle route optimization system embodiment one based on deep learning structural representation, such as Fig. 3
The shown present invention's is included based on the vehicle route optimization system of deep learning：
Data acquisition and processing module, which is used to obtain Real-time Road data and history road data, the number that will be obtained
According to the data set for carrying out formation tape label after pretreatment；
Depth belief network model training module, which is used for construction depth belief network model, using the data of tape label
Collection, Real-time Road data and history road data are training depth belief network model；
Path congestion coefficients calculation block, which is used for using the depth belief network model that completes of training to vehicle to purpose
The All Paths on ground are predicted, and export the congestion coefficient in each path；
Optimal path computation module, which is used for according to congestion coefficient and apart from this two indexs come comprehensive assessment path, defeated
Go out optimal path；The optimal path is for congestion coefficient and apart from the path corresponding to this two index linear superpositions minimums.
Wherein, data acquisition and processing module also include：
Normalization module, which is used to be normalized the data of acquisition；
Roadway characteristic vector forms module, and which is used for when the data for obtaining have missing values, is mended using average interpolation method
Together, arrangement form roadway characteristic is vectorial in a fixed order.
Fig. 4 be the present invention the vehicle route optimization system embodiment two based on deep learning structural representation, such as Fig. 4
The shown present invention's is included based on the vehicle route optimization system of deep learning：Data acquisition unit, the data acquisition dress
Put and be connected with server.
Wherein, data acquisition unit, which is configured to gather Real-time Road data and history road data.
Wherein, data acquisition unit can be achieved using data collecting card, and its structure is prior art.
Server is configured to：
Real-time Road data and history road data are obtained, the data of acquisition are carried out tape label is formed after pretreatment
Data set；
Construction depth belief network model, using the data set of tape label, Real-time Road data and history road data
To train depth belief network model；
The depth belief network model completed using training is predicted to the All Paths of vehicle to destination, and output is each
The congestion coefficient in individual path；
According to congestion coefficient and apart from this two indexs come comprehensive assessment path, optimal path is exported；The optimal path
For congestion coefficient and apart from the path corresponding to this two index linear superposition minimums.
In implementing, the road data of a large amount of real-time and history can provide safeguard for model prediction accuracy rate afterwards.
Real-time Road data include：
(1) by arranging the average speed and vehicle flowrate data that sensing equipment obtains response section in real time in each section.
(2) by the video data in monitoring device Real-time Collection track.
(3) by gas such as the visibility in each section of meteorological information acquisition equipment Real-time Collection, temperature, humidity, wind-force, sleet
Wait data, Comprehensive Assessment weather alert grade.
History road data includes：
(1) the history average speed and history vehicle flowrate data in response section are obtained.
(2) the history video data in track.
(3) the climatic data such as the history visibility in each section, temperature, humidity, wind-force, sleet, Comprehensive Assessment weather alert etc.
Level.
As a example by the special vehicle of special duty is taken the post as：
In view of the sudden and randomness of special assignment, responding period and war preparedness period can be divided within one day, for not
With the frequency difference of period gathered data, it is specifically described below.
In the war preparedness period, the interval of delta t of gathered data is set to 10 minutes；
In the responding period, the interval of delta t of gathered data is set to 1 minute；
Road data mainly includes vehicle number of the unit interval by crossing, Emergency Vehicle Lane occupancy situation (take as 0.9,
Without car for 0.1), weather alert quantify (without early warning be 0.1, blue early warning be 0.3, yellow early warning be that 0.5, orange early warning is
0.7th, red early warning is for 0.9), period factors quantization (working day morning peak be 0.9, evening peak be 0.7, festivals or holidays be 0.6), its
His factor (it is 0.9 that road maintenance is 0.7, vehicle accident 0.8, march-rally), listed numerical value is not unique, can be according to locality
Practical situation is adjusted flexibly numerical value.
The system also includes display device, and which is connected with server, and is display configured to optimal path.
One of ordinary skill in the art will appreciate that all or part of flow process in realizing above-described embodiment method, can be
Instruct related hardware to complete by computer program, described program can be stored in a computer read/write memory medium
In, the program is upon execution, it may include such as the flow process of the embodiment of above-mentioned each method.Wherein, described storage medium can be magnetic
Dish, CD, read-only memory (Read-Only Memory, ROM) or random access memory (Random
AccessMemory, RAM) etc..
Although the above-mentioned accompanying drawing that combines is described to the specific embodiment of the present invention, not to present invention protection model
The restriction enclosed, one of ordinary skill in the art should be understood that on the basis of technical scheme those skilled in the art are not
The various modifications made by needing to pay creative work or deformation are still within protection scope of the present invention.
Claims (10)
1. a kind of vehicle route optimization method based on deep learning, it is characterised in that include：
Step 1：Real-time Road data and history road data are obtained, after the data of acquisition are carried out pretreatment, tape label is formed
Data set；
Step 2：Construction depth belief network model, using the data set of tape label, Real-time Road data and history road way
According to training depth belief network model；
Step 3：The depth belief network model completed using training is predicted to the All Paths of vehicle to destination, is exported
The congestion coefficient in each path；
Step 4：According to congestion coefficient and apart from this two indexs come comprehensive assessment path, optimal path is exported；The optimum road
Footpath is for congestion coefficient and apart from the path corresponding to this two index linear superpositions minimums.
2. the vehicle route optimization method based on deep learning as claimed in claim 1, it is characterised in that the data that will be obtained
The process for carrying out pretreatment includes：
The data of acquisition are normalized；
When the data for obtaining have missing values, using average interpolation method polishing, arrangement form roadway characteristic in a fixed order
Vector.
3. the vehicle route optimization method based on deep learning as claimed in claim 1, it is characterised in that depth belief network
Model is made up of the limited Boltzmann machine of multilamellar and one layer of reverse transmittance nerve network；Wherein, reverse transmittance nerve network is located at
Top layer, the activation primitive of depth belief network model adopt PReLU functions.
4. the vehicle route optimization method based on deep learning as claimed in claim 3, it is characterised in that in training depth letter
During reading the limited Boltzmann machine of any layer in network model, using history road data as training sample, by non-prison
Superintend and direct greedy successively method to train the weights of the limited Boltzmann machine of any layer, obtain the model ginseng of depth belief network model
Number.
5. the vehicle route optimization method based on deep learning as claimed in claim 3, it is characterised in that last layer is limited
Input of the output of Boltzmann machine as depth belief network model top layer.
6. the vehicle route optimization method based on deep learning as claimed in claim 4, it is characterised in that depth belief network
Model top layer is finely adjusted to the model parameter in depth belief network model by the data set of tape label, obtains ID
Belief network model.
7. the vehicle route optimization method based on deep learning as claimed in claim 6, it is characterised in that reverse using error
Propagation algorithm is finely adjusted to the model parameter in depth belief network model；Or
Using the dual average algorithm RDA of regularization and ID belief network model, Real-time Road data are instructed online
Practice, continue to optimize depth belief network model.
8. a kind of vehicle route optimization system based on deep learning, it is characterised in that include：
Data acquisition and processing module, which is used to obtain Real-time Road data and history road data, the data of acquisition is entered
The data set of tape label is formed after row pretreatment；
Depth belief network model training module, which is used for construction depth belief network model, using the data set of tape label, reality
When road data and history road data training depth belief network model；
Path congestion coefficients calculation block, its depth belief network model for being used to complete using training arrive destination to vehicle
All Paths are predicted, and export the congestion coefficient in each path；
Optimal path computation module, which is used for according to congestion coefficient and apart from this two indexs come comprehensive assessment path, and output is most
Shortest path；The optimal path is for congestion coefficient and apart from the path corresponding to this two index linear superpositions minimums.
9. the vehicle route optimization system based on deep learning as claimed in claim 8, it is characterised in that the data acquisition
And processing module also includes：
Normalization module, which is used to be normalized the data of acquisition；
Roadway characteristic vector forms module, and which is used for when the data for obtaining have missing values, using average interpolation method polishing, presses
According to permanent order arrangement form roadway characteristic vector.
10. a kind of vehicle route optimization system based on deep learning, it is characterised in that include：
Data acquisition unit, which is configured to gather Real-time Road data and history road data；
The data acquisition unit is connected with server, and the server is configured to：
Real-time Road data and history road data are obtained, the data of acquisition are carried out after pretreatment, to form the data of tape label
Collection；
Construction depth belief network model, is instructed using the data set of tape label, Real-time Road data and history road data
Practice depth belief network model；
The depth belief network model completed using training is predicted to the All Paths of vehicle to destination, exports each road
The congestion coefficient in footpath；
According to congestion coefficient and apart from this two indexs come comprehensive assessment path, optimal path is exported；The optimal path is to gather around
Stifled coefficient and apart from the path corresponding to this two index linear superpositions minimum.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610976649.6A CN106548645B (en) | 2016-11-03 | 2016-11-03 | Vehicle route optimization method and system based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610976649.6A CN106548645B (en) | 2016-11-03 | 2016-11-03 | Vehicle route optimization method and system based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106548645A true CN106548645A (en) | 2017-03-29 |
| CN106548645B CN106548645B (en) | 2019-07-12 |
Family
ID=58394818
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610976649.6A Active CN106548645B (en) | 2016-11-03 | 2016-11-03 | Vehicle route optimization method and system based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106548645B (en) |
Cited By (39)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107270925A (en) * | 2017-07-27 | 2017-10-20 | 三星电子（中国）研发中心 | A kind of user's Vehicular navigation system, device and method |
| CN107368076A (en) * | 2017-07-31 | 2017-11-21 | 中南大学 | Robot motion's pathdepth learns controlling planning method under a kind of intelligent environment |
| CN107424426A (en) * | 2017-07-10 | 2017-12-01 | 南京邮电大学 | A kind of logistics route planing method based on deep learning |
| CN107491072A (en) * | 2017-09-05 | 2017-12-19 | 百度在线网络技术（北京）有限公司 | Vehicle obstacle-avoidance method and apparatus |
| CN107609477A (en) * | 2017-08-09 | 2018-01-19 | 五邑大学 | It is a kind of that detection method is fallen down with what Intelligent bracelet was combined based on deep learning |
| CN107886049A (en) * | 2017-10-16 | 2018-04-06 | 江苏省气象服务中心 | A kind of visibility identification method for early warning based on camera probe |
| CN108108808A (en) * | 2018-01-08 | 2018-06-01 | 北京邮电大学 | A kind of position predicting method and device based on depth belief network |
| CN108108716A (en) * | 2017-12-29 | 2018-06-01 | 中国电子科技集团公司信息科学研究院 | A kind of winding detection method based on depth belief network |
| CN108122048A (en) * | 2017-11-23 | 2018-06-05 | 浙江专线宝网阔物联科技有限公司 | A kind of transportation route dispatching method and its system |
| CN108198413A (en) * | 2017-12-20 | 2018-06-22 | 河南中裕广恒科技股份有限公司 | Blocking method is delayed in the intelligent transportation of a kind of big data and autonomous deep learning |
| CN108268038A (en) * | 2018-01-19 | 2018-07-10 | 广东美的智能机器人有限公司 | The dispatching method and system of multiple mobile robot |
| CN108665091A (en) * | 2018-04-07 | 2018-10-16 | 深圳供电局有限公司 | Material intelligent scheduling method based on machine learning |
| WO2018209576A1 (en) * | 2017-05-16 | 2018-11-22 | Beijing Didi Infinity Technology And Development Co., Ltd. | Systems and methods for digital route planning |
| CN108876045A (en) * | 2018-06-25 | 2018-11-23 | 上海应用技术大学 | Emergency tender optimal route recommended method based on LSTM model prediction |
| WO2018227374A1 (en) * | 2017-06-13 | 2018-12-20 | Beijing Didi Infinity Technology And Development Co., Ltd. | Systems and methods for route planning based on deep convolutional neural network |
| CN109166311A (en) * | 2018-08-15 | 2019-01-08 | 广州市康软信息科技有限公司 | A kind of region first aid monitoring method, system and device based on vehicle scheduling |
| CN109556609A (en) * | 2018-11-15 | 2019-04-02 | 武汉南华工业设备工程股份有限公司 | A kind of collision prevention method and device based on artificial intelligence |
| CN109697875A (en) * | 2017-10-23 | 2019-04-30 | 华为技术有限公司 | Plan the method and device of driving trace |
| CN109948835A (en) * | 2019-02-28 | 2019-06-28 | 江苏理工学院 | A method of optimal path and calculating path time based on artificial neural network |
| CN110197593A (en) * | 2019-05-20 | 2019-09-03 | 河海大学 | A kind of road conditions broadcasting method and system based on cloud path prediction in short-term |
| CN110309957A (en) * | 2019-06-17 | 2019-10-08 | 北京航空航天大学 | It is a kind of towards in the car-haul time pre-judging method with the unmanned transportation system of mine car |
| CN110837253A (en) * | 2019-10-31 | 2020-02-25 | 华中科技大学 | Intelligent addressing system based on memristor synapse |
| CN111538059A (en) * | 2020-05-11 | 2020-08-14 | 东华大学 | Self-adaptive rapid dynamic positioning system and method based on improved Boltzmann machine |
| CN111653114A (en) * | 2020-06-12 | 2020-09-11 | 北京百度网讯科技有限公司 | Information reminding method and device, electronic equipment and storage medium |
| CN111768629A (en) * | 2019-11-04 | 2020-10-13 | 北京京东乾石科技有限公司 | Vehicle scheduling method, device and system |
| CN112216130A (en) * | 2020-09-30 | 2021-01-12 | 长沙理工大学 | Emergency vehicle guidance method under cooperative vehicle and road environment |
| CN112735129A (en) * | 2020-12-25 | 2021-04-30 | 北京中交兴路信息科技有限公司 | Method and device for truck parking scheduling |
| CN113207127A (en) * | 2021-04-27 | 2021-08-03 | 重庆邮电大学 | Dynamic spectrum access method based on hierarchical deep reinforcement learning in NOMA system |
| CN113762575A (en) * | 2020-12-02 | 2021-12-07 | 北京京东乾石科技有限公司 | Conveying line path optimization method and device |
| WO2022006873A1 (en) * | 2020-07-10 | 2022-01-13 | Beijing Didi Infinity Technology And Development Co., Ltd. | Vehicle repositioning on mobility-on-demand platforms |
| WO2022016901A1 (en) * | 2020-07-20 | 2022-01-27 | 华为技术有限公司 | Method for planning driving route of vehicle, and intelligent vehicle |
| CN114518750A (en) * | 2021-12-15 | 2022-05-20 | 电子科技大学 | Path planning method based on amoebic movement mechanism |
| CN115294791A (en) * | 2022-07-29 | 2022-11-04 | 广州市粤迅特数码技术有限公司 | Intelligent traffic guidance system for smart city |
| CN115527372A (en) * | 2022-11-29 | 2022-12-27 | 湖南工商大学 | Intelligent traffic real-time induction variable information system based on global optimization |
| CN116128224A (en) * | 2022-12-23 | 2023-05-16 | 山东省人工智能研究院 | Intelligent driving decision method integrating road network equalization and accident risk assessment |
| CN116403426A (en) * | 2023-04-10 | 2023-07-07 | 广东海洋大学 | Road network leading method and system based on road condition big data |
| CN117133131A (en) * | 2023-10-26 | 2023-11-28 | 深圳市地铁集团有限公司 | Intelligent traffic control system based on ARM technology system |
| CN117970892A (en) * | 2024-03-29 | 2024-05-03 | 合肥焕智科技有限公司 | Control method and device of conveying system |
| CN118155419A (en) * | 2024-04-02 | 2024-06-07 | 广科通立(广东)数据服务有限公司 | Emergency event supervision system based on traffic network data analysis |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104077595A (en) * | 2014-06-15 | 2014-10-01 | 北京工业大学 | Deep belief network image recognition method based on Bayesian regularization |
| CN104217593A (en) * | 2014-08-27 | 2014-12-17 | 北京航空航天大学 | Real-time road condition information acquisition method orienting to cellphone traveling speed |
| CN105046365A (en) * | 2015-07-29 | 2015-11-11 | 余意 | Method and device for route optimization of logistics delivery vehicle |
| CN105096614A (en) * | 2015-09-23 | 2015-11-25 | 南京遒涯信息技术有限公司 | Newly established crossing traffic flow prediction method based on generating type deep belief network |
| CN105160866A (en) * | 2015-08-07 | 2015-12-16 | 浙江高速信息工程技术有限公司 | Traffic flow prediction method based on deep learning nerve network structure |
| CN105390018A (en) * | 2015-10-16 | 2016-03-09 | 上海物联网有限公司 | Underground parking lot intelligent guiding system based on machine learning |
| CN105444766A (en) * | 2015-12-16 | 2016-03-30 | 清华大学 | Indoor navigation method based on deep learning |
| EP3038323A1 (en) * | 2014-12-26 | 2016-06-29 | Mattias Bergstorm | Method and system for adaptive virtual broadcasting of digital content |
| WO2016154440A1 (en) * | 2015-03-24 | 2016-09-29 | Hrl Laboratories, Llc | Sparse inference modules for deep learning |
-
2016
- 2016-11-03 CN CN201610976649.6A patent/CN106548645B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104077595A (en) * | 2014-06-15 | 2014-10-01 | 北京工业大学 | Deep belief network image recognition method based on Bayesian regularization |
| CN104217593A (en) * | 2014-08-27 | 2014-12-17 | 北京航空航天大学 | Real-time road condition information acquisition method orienting to cellphone traveling speed |
| EP3038323A1 (en) * | 2014-12-26 | 2016-06-29 | Mattias Bergstorm | Method and system for adaptive virtual broadcasting of digital content |
| WO2016154440A1 (en) * | 2015-03-24 | 2016-09-29 | Hrl Laboratories, Llc | Sparse inference modules for deep learning |
| CN105046365A (en) * | 2015-07-29 | 2015-11-11 | 余意 | Method and device for route optimization of logistics delivery vehicle |
| CN105160866A (en) * | 2015-08-07 | 2015-12-16 | 浙江高速信息工程技术有限公司 | Traffic flow prediction method based on deep learning nerve network structure |
| CN105096614A (en) * | 2015-09-23 | 2015-11-25 | 南京遒涯信息技术有限公司 | Newly established crossing traffic flow prediction method based on generating type deep belief network |
| CN105390018A (en) * | 2015-10-16 | 2016-03-09 | 上海物联网有限公司 | Underground parking lot intelligent guiding system based on machine learning |
| CN105444766A (en) * | 2015-12-16 | 2016-03-30 | 清华大学 | Indoor navigation method based on deep learning |
Non-Patent Citations (3)
| Title |
|---|
| 于乃功等: "基于深度自动编码器与Q学习的移动机器人", 《北京工业大学学报》 * |
| 江德浩: "基于深度信念网络的短时交通流预测", 《万方数据》 * |
| 谭娟，王胜春: "深度学习在城市交通流预测中的实践研究", 《计算机应用研究》 * |
Cited By (60)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10209084B2 (en) | 2017-05-16 | 2019-02-19 | Beijing Didi Infinity Technology And Development Co., Ltd. | Systems and methods for digital route planning |
| WO2018209576A1 (en) * | 2017-05-16 | 2018-11-22 | Beijing Didi Infinity Technology And Development Co., Ltd. | Systems and methods for digital route planning |
| US10782140B2 (en) | 2017-05-16 | 2020-09-22 | Beijing Didi Infinity Technology And Development Co., Ltd. | Systems and methods for digital route planning |
| US11644323B2 (en) | 2017-05-16 | 2023-05-09 | Beijing Didi Infinity Technology And Development Co., Ltd. | Systems and methods for digital route planning |
| WO2018227374A1 (en) * | 2017-06-13 | 2018-12-20 | Beijing Didi Infinity Technology And Development Co., Ltd. | Systems and methods for route planning based on deep convolutional neural network |
| US10816349B2 (en) | 2017-06-13 | 2020-10-27 | Beijing Didi Infinity Technology And Development Co., Ltd. | Systems and methods for route planning based on deep convolutional neural network |
| CN110691957A (en) * | 2017-06-13 | 2020-01-14 | 北京嘀嘀无限科技发展有限公司 | Path planning system and method based on deep convolutional neural network |
| CN110691957B (en) * | 2017-06-13 | 2023-05-12 | 北京嘀嘀无限科技发展有限公司 | Path planning system and method based on deep convolutional neural network |
| CN107424426B (en) * | 2017-07-10 | 2020-07-07 | 南京邮电大学 | Logistics path planning method based on deep learning |
| CN107424426A (en) * | 2017-07-10 | 2017-12-01 | 南京邮电大学 | A kind of logistics route planing method based on deep learning |
| CN107270925A (en) * | 2017-07-27 | 2017-10-20 | 三星电子（中国）研发中心 | A kind of user's Vehicular navigation system, device and method |
| CN107270925B (en) * | 2017-07-27 | 2021-02-05 | 三星电子（中国）研发中心 | User vehicle navigation system, device and method |
| CN107368076B (en) * | 2017-07-31 | 2018-03-27 | 中南大学 | Robot motion's pathdepth learns controlling planning method under a kind of intelligent environment |
| CN107368076A (en) * | 2017-07-31 | 2017-11-21 | 中南大学 | Robot motion's pathdepth learns controlling planning method under a kind of intelligent environment |
| CN107609477A (en) * | 2017-08-09 | 2018-01-19 | 五邑大学 | It is a kind of that detection method is fallen down with what Intelligent bracelet was combined based on deep learning |
| CN107491072A (en) * | 2017-09-05 | 2017-12-19 | 百度在线网络技术（北京）有限公司 | Vehicle obstacle-avoidance method and apparatus |
| CN107491072B (en) * | 2017-09-05 | 2021-03-30 | 百度在线网络技术（北京）有限公司 | Vehicle obstacle avoidance method and device |
| CN107886049A (en) * | 2017-10-16 | 2018-04-06 | 江苏省气象服务中心 | A kind of visibility identification method for early warning based on camera probe |
| CN109697875A (en) * | 2017-10-23 | 2019-04-30 | 华为技术有限公司 | Plan the method and device of driving trace |
| CN108122048A (en) * | 2017-11-23 | 2018-06-05 | 浙江专线宝网阔物联科技有限公司 | A kind of transportation route dispatching method and its system |
| CN108198413A (en) * | 2017-12-20 | 2018-06-22 | 河南中裕广恒科技股份有限公司 | Blocking method is delayed in the intelligent transportation of a kind of big data and autonomous deep learning |
| CN108108716A (en) * | 2017-12-29 | 2018-06-01 | 中国电子科技集团公司信息科学研究院 | A kind of winding detection method based on depth belief network |
| CN108108808B (en) * | 2018-01-08 | 2021-11-19 | 北京邮电大学 | Position prediction method and device based on deep belief network |
| CN108108808A (en) * | 2018-01-08 | 2018-06-01 | 北京邮电大学 | A kind of position predicting method and device based on depth belief network |
| CN108268038A (en) * | 2018-01-19 | 2018-07-10 | 广东美的智能机器人有限公司 | The dispatching method and system of multiple mobile robot |
| CN108665091A (en) * | 2018-04-07 | 2018-10-16 | 深圳供电局有限公司 | Material intelligent scheduling method based on machine learning |
| CN108876045A (en) * | 2018-06-25 | 2018-11-23 | 上海应用技术大学 | Emergency tender optimal route recommended method based on LSTM model prediction |
| CN109166311A (en) * | 2018-08-15 | 2019-01-08 | 广州市康软信息科技有限公司 | A kind of region first aid monitoring method, system and device based on vehicle scheduling |
| CN109556609A (en) * | 2018-11-15 | 2019-04-02 | 武汉南华工业设备工程股份有限公司 | A kind of collision prevention method and device based on artificial intelligence |
| CN109948835A (en) * | 2019-02-28 | 2019-06-28 | 江苏理工学院 | A method of optimal path and calculating path time based on artificial neural network |
| CN110197593A (en) * | 2019-05-20 | 2019-09-03 | 河海大学 | A kind of road conditions broadcasting method and system based on cloud path prediction in short-term |
| CN110309957A (en) * | 2019-06-17 | 2019-10-08 | 北京航空航天大学 | It is a kind of towards in the car-haul time pre-judging method with the unmanned transportation system of mine car |
| CN110837253A (en) * | 2019-10-31 | 2020-02-25 | 华中科技大学 | Intelligent addressing system based on memristor synapse |
| CN111768629A (en) * | 2019-11-04 | 2020-10-13 | 北京京东乾石科技有限公司 | Vehicle scheduling method, device and system |
| CN111768629B (en) * | 2019-11-04 | 2022-04-12 | 北京京东乾石科技有限公司 | Vehicle scheduling method, device and system |
| CN111538059A (en) * | 2020-05-11 | 2020-08-14 | 东华大学 | Self-adaptive rapid dynamic positioning system and method based on improved Boltzmann machine |
| CN111538059B (en) * | 2020-05-11 | 2022-11-11 | 东华大学 | Self-adaptive rapid dynamic positioning system and method based on improved Boltzmann machine |
| CN111653114A (en) * | 2020-06-12 | 2020-09-11 | 北京百度网讯科技有限公司 | Information reminding method and device, electronic equipment and storage medium |
| US12061090B2 (en) | 2020-07-10 | 2024-08-13 | Beijing Didi Infinity Technology And Development Co., Ltd. | Vehicle repositioning on mobility-on-demand platforms |
| WO2022006873A1 (en) * | 2020-07-10 | 2022-01-13 | Beijing Didi Infinity Technology And Development Co., Ltd. | Vehicle repositioning on mobility-on-demand platforms |
| WO2022016901A1 (en) * | 2020-07-20 | 2022-01-27 | 华为技术有限公司 | Method for planning driving route of vehicle, and intelligent vehicle |
| CN112216130A (en) * | 2020-09-30 | 2021-01-12 | 长沙理工大学 | Emergency vehicle guidance method under cooperative vehicle and road environment |
| CN113762575A (en) * | 2020-12-02 | 2021-12-07 | 北京京东乾石科技有限公司 | Conveying line path optimization method and device |
| CN113762575B (en) * | 2020-12-02 | 2024-05-21 | 北京京东乾石科技有限公司 | Conveying line path optimization method and device |
| CN112735129A (en) * | 2020-12-25 | 2021-04-30 | 北京中交兴路信息科技有限公司 | Method and device for truck parking scheduling |
| CN113207127B (en) * | 2021-04-27 | 2022-04-12 | 重庆邮电大学 | Dynamic spectrum access method based on hierarchical deep reinforcement learning in NOMA system |
| CN113207127A (en) * | 2021-04-27 | 2021-08-03 | 重庆邮电大学 | Dynamic spectrum access method based on hierarchical deep reinforcement learning in NOMA system |
| CN114518750B (en) * | 2021-12-15 | 2023-08-15 | 电子科技大学 | Path planning method based on amoeba motion mechanism |
| CN114518750A (en) * | 2021-12-15 | 2022-05-20 | 电子科技大学 | Path planning method based on amoebic movement mechanism |
| CN115294791A (en) * | 2022-07-29 | 2022-11-04 | 广州市粤迅特数码技术有限公司 | Intelligent traffic guidance system for smart city |
| CN115527372A (en) * | 2022-11-29 | 2022-12-27 | 湖南工商大学 | Intelligent traffic real-time induction variable information system based on global optimization |
| CN116128224B (en) * | 2022-12-23 | 2024-01-26 | 山东省人工智能研究院 | Intelligent driving decision method integrating road network equalization and accident risk assessment |
| CN116128224A (en) * | 2022-12-23 | 2023-05-16 | 山东省人工智能研究院 | Intelligent driving decision method integrating road network equalization and accident risk assessment |
| CN116403426B (en) * | 2023-04-10 | 2024-01-05 | 广东海洋大学 | Road network leading method and system based on road condition big data |
| CN116403426A (en) * | 2023-04-10 | 2023-07-07 | 广东海洋大学 | Road network leading method and system based on road condition big data |
| CN117133131B (en) * | 2023-10-26 | 2024-02-20 | 深圳市地铁集团有限公司 | Intelligent traffic control system based on ARM technology system |
| CN117133131A (en) * | 2023-10-26 | 2023-11-28 | 深圳市地铁集团有限公司 | Intelligent traffic control system based on ARM technology system |
| CN117970892A (en) * | 2024-03-29 | 2024-05-03 | 合肥焕智科技有限公司 | Control method and device of conveying system |
| CN117970892B (en) * | 2024-03-29 | 2024-07-09 | 合肥焕智科技有限公司 | Control method and device of conveying system |
| CN118155419A (en) * | 2024-04-02 | 2024-06-07 | 广科通立(广东)数据服务有限公司 | Emergency event supervision system based on traffic network data analysis |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106548645B (en) | 2019-07-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106548645B (en) | Vehicle route optimization method and system based on deep learning | |
| CN109697852B (en) | Urban road congestion degree prediction method based on time sequence traffic events | |
| Yang et al. | A deep learning approach to real-time parking occupancy prediction in transportation networks incorporating multiple spatio-temporal data sources | |
| CN109993970B (en) | Urban area traffic accident risk prediction method | |
| Koesdwiady et al. | Improving traffic flow prediction with weather information in connected cars: A deep learning approach | |
| Yang et al. | Understanding and predicting travel time with spatio-temporal features of network traffic flow, weather and incidents | |
| Kumar et al. | Short term traffic flow prediction for a non urban highway using artificial neural network | |
| CN104376716B (en) | Method for dynamically generating bus timetables on basis of Bayesian network models | |
| Yao et al. | A support vector machine with the tabu search algorithm for freeway incident detection | |
| Wang et al. | Fine-grained traffic flow prediction of various vehicle types via fusion of multisource data and deep learning approaches | |
| CN114692984B (en) | Traffic prediction method based on multi-step coupling graph convolution network | |
| CN105493109A (en) | Air quality inference using multiple data sources | |
| CN110299011A (en) | A kind of traffic flow forecasting method of the highway arbitrary cross-section based on charge data | |
| US20240054321A1 (en) | Traffic prediction | |
| Akin et al. | A neural network (NN) model to predict intersection crashes based upon driver, vehicle and roadway surface characteristics | |
| CN111275971A (en) | Expressway traffic flow prediction method | |
| CN107909206A (en) | A kind of PM2.5 Forecasting Methodologies based on deep structure Recognition with Recurrent Neural Network | |
| Zhu et al. | TA-STAN: A deep spatial-temporal attention learning framework for regional traffic accident risk prediction | |
| Said et al. | An intelligent traffic control system using neutrosophic sets, rough sets, graph theory, fuzzy sets and its extended approach: a literature review | |
| CN117238126A (en) | Traffic accident risk assessment method under continuous flow road scene | |
| CN116052435B (en) | Urban road congestion influence range definition and road influence calculation method | |
| Quek et al. | A novel self-organizing fuzzy rule-based system for modelling traffic flow behaviour | |
| CN106384507A (en) | Travel time real-time estimation method based on sparse detector | |
| CN115759329A (en) | Scenic spot tourist number prediction method based on convolutional neural network | |
| D’Apuzzo et al. | An introductory step to develop Distance Decay Functions in the Italian context to assess the modal split to e-bike and e-scooter |
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 |