CN108333959A - A kind of energy saving method of operating of locomotive based on convolutional neural networks model - Google Patents
A kind of energy saving method of operating of locomotive based on convolutional neural networks model Download PDFInfo
- Publication number
- CN108333959A CN108333959A CN201810192684.8A CN201810192684A CN108333959A CN 108333959 A CN108333959 A CN 108333959A CN 201810192684 A CN201810192684 A CN 201810192684A CN 108333959 A CN108333959 A CN 108333959A
- Authority
- CN
- China
- Prior art keywords
- locomotive
- data
- convolutional neural
- neural networks
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B17/00—Systems involving the use of models or simulators of said systems
- G05B17/02—Systems involving the use of models or simulators of said systems electric
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Automation & Control Theory (AREA)
- Traffic Control Systems (AREA)
Abstract
The locomotive smart steering method based on CNN that the present invention provides a kind of.Its step is:Acquisition has the history of the driver of abundant driving experience to drive data and locomotive operation monitor daily record；Feature extraction is carried out to collected data, obtains training dataset and test data set；Then CNN model parameters are adjusted, model training is carried out using training data, this step of iteration is until model is restrained；Trained CNN models are used for the prediction of locomotive operation gear；Emulation testing is carried out to test data set based on prediction gear, speed and gear curve that driver's practical the case where driving and CNN model predictions go out are compared.Data preprocessing method, modelling and training method proposed by the present invention can make full use of data information, reach preferable locomotive gear prediction effect.
Description
Technical field
The present invention relates to the energy saving manipulation field of railway locomotive more particularly to a kind of base convolutional neural networks (CNN) models
The energy saving method of operating of locomotive.
Background technology
Locomotive operation control system is typical multiple target, multiple constraint, a nonlinear complex control system, needs to protect
Reliability, safety, punctuality and fuel-economizing are demonstrate,proved, so the energy saving manipulation problem of locomotive is a non-linear constrained more mesh
Mark optimization problems.And constraints of such problem due to needing consideration numerous complicated during solving, it is entire to optimize
Search space it is very big, be that more difficult the problem of searching optimal solution, the realistic meaning studied it are great in a short time.
The optimization method of the existing energy saving operation and control of locomotive can be divided into numerical search method, Analytical Solution method and open
Hairdo strategy design method.Wherein numerical search refers to carries out optimizing search to obtain by numerical search algorithm to manipulating sequence
The manipulation sequence of optimization, common algorithm has genetic algorithm, group hunting algorithm, Dynamic Programming etc., but time-consuming for this mode,
And it is difficult to converge to optimal result；Analytical Solution method refers to based on domain knowledge to the pass under different situations during manipulation and control
Key transfer point obtains final optimized handling sequence according to analytic formula solution, but this mode major defect is transfer point
Analytic formula derivation is complicated, it is more difficult to handle multi-constraint condition；Heuristic strategies design method refer to consider it is many it is complicated because
Element manually passes through the tactful design of the didactic progress such as some working specifications in existing field, the excessive introducing people of this kind of mode
The analysis and design of work greatly reduce the efficiency of strategy design, simultaneously because people's thinking is limited in scope, can not cover all
Possible situation, this will certainly cause part optimization solution to be omitted.So there is all kinds of drawbacks for current solution.
At present due to the significantly promotion of computer computation ability, depth learning technology is widely used in multiple fields
In challenge solution, such as the control of image recognition, machine translation, robot, automatic Pilot.Depth learning technology relies on
Its stronger complex state characterization ability and complex characteristic abstracting power, achieve in the above field than traditional solution party
The more significant effect of method.Convolutional neural networks (Convolutional Neural Network, CNN) are in depth learning technology
One of the network structure of great representative achieves prodigious success in image processing field, and convolutional neural networks are a kind of special
Deep layer neural network model, its particularity is embodied in two aspects, and on the one hand its interneuronal connection is non-complete
Connection, the weight of the connection in another aspect same layer between certain neurons is shared (i.e. identical).It non-complete
Connection and weights share network structure be allowed to be more closely similar to biological neural network, reduce network model complexity (for
It is difficult for the deep structure of study, this is very important.So the problem of manipulation that CNN is energy saving applied to locomotive, solves
On, have great importance in terms of breaking through the complicated engineer of existing method, the bottleneck that complicated formulas derives.
Invention content
The object of the present invention is to provide a kind of energy saving method of operating of the locomotive based on CNN models.CNN models are in image procossing
Aspect has a wide range of applications, can be with the original spatial positional information of retention data, and has light weight, reduces model training
The characteristics of complexity.
The present invention is achieved through the following technical solutions：
The energy saving method of operating of a kind of locomotive based on convolutional neural networks model, which is characterized in that the method includes：
Step S101, acquisition driver's history driving data and locomotive operation monitor daily record, as initial training data；
Step S102 pre-processes initial training data, obtains training dataset and test data set；
Step S103, repetition training convolutional neural networks model simultaneously preserve trained model；
Step S104, the good model of application training is predicted, and carries out analog simulation test.
Further, in step 1, driver's history driving data of acquisition includes locomotive attribute, line properties and locomotive
Travel daily record.
Further, in step S102, preprocessing process is divided into two stages：First stage, from original training data
The data set of driver driving behavior is chosen, second stage continues to pre-process the driving data selected.
Further, for a locomotive driving state, then the feature extracted can be divided into two parts, be in the locomotive respectively
The driving information of locomotive and road information and the road information after the locomotive driving state before transport condition.
Further, forward direction feature include car weight, vehicle commander, weight vehicle number, light vehicle number, when scarp slope section mean inclination, when
Scarp slope section total length, at scarp slope section average speed, the speed limit of current location, the residue length of current speed limit, current point
In two stations between average speed, current location reach apart from next post distance, current location the residue at next station
Time, current gear, present speed, the gradient of current location point, the difference of present speed and speed limit.
Further, backward feature includes being grown when scarp slope section mean inclination, when scarp slope section total length, when scarp slope section residue
Degree, when scarp slope section average speed, the speed limit of current location, current location distance apart from next station, current location point
The difference of the gradient, the speed of current locomotive and extraction feature location point speed limit.
The beneficial effects of the invention are as follows：
(1) present invention devises 17 forward direction features and 8 backward features, input feature vector number of these features as model
According to taking full advantage of data information.The design of forward direction feature considers timing and physical location feature simultaneously, and backward feature is examined
Position characteristics are considered, have further improved state representation ability of the model to locomotive, to finally influence the accurate of prediction gear
Property；
(2) present invention is improved on the basis of existing CNN models, using two convolutional neural networks, in input layer
It is middle to input preceding come to feature and backward character separation respectively, then by the conscientious unified integration of the result handled respectively.Originally it asks
Forward direction feature and backward feature in topic are had any different in feature quantity and characteristic meaning, so this is solved according to the present invention
The certainly particularity of problem and the structure that designs, the design of the structure is more suitable for solving the problems, such as the energy saving manipulation of locomotive；
(3) present invention proposes the step of repetitive exercise CNN and method, and after repetitive exercise, the gear of model is predicted
The promotion that ability will be obtained further, or even obtain the more optimized manipulation gear of driver more outstanding than the mankind；
(4) present invention devises the locomotive gear prediction model based on CNN, which utilizes with abundant driving experience
The history driving data of human operators and the road section information data of locomotive driving etc. carry out the manipulation gear during locomotive driving
Prediction, the prediction of the gear compromise between security (no hypervelocity risk etc.), punctuality and energy saving simultaneously；To make full use of data
Information, the present invention proposes the extracting method of temporal aspect (to feature and backward feature before being divided into), respectively with different CNN nets
Network handles forward and backward feature, is then being used for the defeated of output layer altogether by treated abstract data is unified
Enter.
Description of the drawings
Fig. 1 is the training process that locomotive of the present invention predicts gear model；
Fig. 2 is the network structure details of the CNN of the present invention；
Fig. 3 is the model structure frame diagram of the present invention；
Specific implementation mode
To keep the present invention relatively sharp, the present invention is described in detail below in conjunction with the accompanying drawings.
The present embodiment provides a kind of locomotive smart steering methods being based on convolutional neural networks (CNN) model, specifically include：
Step S101, acquisition driver's history driving data and locomotive operation monitor daily record, as initial training data.
The history driving data of locomotive driver and locomotive operation monitor daily record can be from the LKJ (row in railway locomotive
Vehicle operation control recording device) it obtains.
For locomotive driving data of the specific driver on specific route, driver's history driving data of acquisition
Including：Locomotive attribute, line properties and locomotive driving daily record.Wherein, locomotive attribute include car weight, vehicle commander, weight vehicle number and
Light vehicle number；Line properties include the run time information between the gradient of circuit, speed-limiting messages, station information, two stations；Locomotive
Traveling daily record includes the information such as timestamp, travel speed, locomotive driving gear, fuel consumption record.The data being collected into are constituted just
Beginning training data.
Step S102 pre-processes initial training data, obtains training dataset and test data set.
Include more redundancy in original training data, and data format disunity, cannot directly as training data,
It needs to pre-process original training data.
Preprocessing process is divided into two stages：
First stage, from original training data choose driver driving behavior data set, these data generally have with
Lower feature：Meet safety, on schedule property, flat vehicle heavy oil consumption it is small, gear change is steady.Wherein safety refers to nothing in driving process
Safety issue occurs, and property refers in driving process without overdue phenomenon on schedule.Two above index can be directly from locomotive driving day
It is obtained in will.
Second stage continues to pre-process the driving data selected.
To consider that locomotive current operating conditions and next distance situation, the present invention carry when in view of locomotive driving simultaneously
The temporal aspect extracting method for locomotive smart steering problem is gone out.
For a specific locomotive driving state, if locomotive is at the kp kilometer posts of road, then the feature extracted can
Be divided into two parts, be respectively before the state driving information of (kilometer post be less than kp) locomotive and road information and the state it
The road information of (kilometer post is more than kp) afterwards, to feature and backward feature before being referred to as in the present invention.At this stage, according to
Temporal aspect extracting method extracts corresponding data from initial data, finally constitutes training dataset and test data
Collection.
Specifically, the forward direction feature that the present invention designs includes 17 features, respectively car weight, vehicle commander, weight vehicle number, light car
Number, when scarp slope section mean inclination, when scarp slope section total length, when scarp slope section average speed, the speed limit of current location, current limit
Average speed, current location between two stations that residue length, the current point of fast value are in apart from next post distance, currently
Position reach the remaining time at next station, current gear, present speed, current location point the gradient, present speed and speed limit
The difference of value；
Backward feature includes 8 features, respectively when scarp slope section mean inclination, when scarp slope section total length, when scarp slope section it is surplus
Remaining length, when distance apart from next station of scarp slope section average speed, the speed limit of current location, current location, current location
The gradient, the difference of the speed of current locomotive and extraction feature location point speed limit of point.
This step carries out data time sequence feature point extraction, the spy extracted by two benches preprocess method described above
Sign point data is divided into training dataset and test data set.
Step S103, model is simultaneously for repetition training convolutional neural networks (Convolutional Neural Network, CNN)
Preserve trained model.
In this step, the training of CNN models is carried out according to the training data that step S102 is extracted.The present invention is directed to machine
The CNN network structure models that car stop position prediction problem proposes are as follows：
Convolutional neural networks (Convolutional Neural Network, CNN) model is pole in depth learning technology
Have one of the network structure represented, achieves prodigious success in image processing field, convolutional neural networks are a kind of special
The neural network model of deep layer, its particularity are embodied in two aspects, and on the one hand its interneuronal connection is non-to connect entirely
It connects, the weight of the connection in another aspect same layer between certain neurons is shared (i.e. identical).The non-of it connects entirely
It connects the network structure shared with weights to be allowed to be more closely similar to biological neural network, reduces the complexity of network model, reduce
The quantity of weights.Gear prediction model proposed by the present invention based on CNN is as shown in Figures 2 and 3.
CNN locomotive gear prediction models as shown in Figure 3 are divided into 4 input layer, convolutional layer, pond layer and output layer portions
Point.Its detail is respectively intended to before processing to spy as shown in Fig. 2, wherein input layer and convolutional layer has been divided into two parts
Levy data and backward characteristic.
To characteristic after being tieed up to feature and 8 before 17 dimensions mentioned in S102 steps.Due to forward and backward characteristic
According to dimension difference, so the network structure for handling this two-part input layer and convolutional layer respectively is also different.Convolutional layer packet
The length of convolution kernel is set as being variable by the convolution kernel containing many specifications in the present invention, and width is to fix to be equal to input
Slice width degree, this is because in this application scene, it is the operation characteristic attribute of some point per data line, is indivisible
, therefore the width of convolution kernel is set as the width of input layer.Pond layer, can also using max-pooling methods
It is changed to other pond methods.It is finally output layer, the full Connection Neural Network of output layer adds softmax regression functions to realize, defeated
The target for going out layer is output prediction gear, predicts to be -8--8 grades in this application scene.
So-called model training refers to the matrix-vector for constantly updating model, the variation in front and back not structure, only
Matrix-vector in model is constantly updated, and keeps its prediction result more and more accurate.
Repetitive exercise several times after, trained model can be preserved, be can be applied in test data set
Gear prediction work, the initial model that can also be used as next training process advanced optimize training.
Step S104, application model are predicted, and carry out analog simulation test.
According to trained CNN models in this step, gear is carried out to test data set, prediction gear is obtained and is imitated
True simulation test.Locomotive driving effect under emulation simulation true environment completely, it is possible to intuitively compare this model prediction and go out
Driving scheme and pilot steering scheme.
The data source of the model of the convolutional neural networks to be trained of the embodiment comes from the outstanding driver at scene
Driving record data, therefore model can learn out the driving habit of these outstanding drivers, therefore also can when train automatic Pilot
It is steady, punctual, fuel-efficient when open as outstanding driver.And fuel consumption is can directly to be calculated by gear
It out, can be to reduce oil consumption as one of training objective in training pattern.
For the performance of comprehensive and accurate assessment method proposed by the invention, real section (the two of Shenyang-Dandong line are chosen
A website) and outstanding driver driving data carry out model training result contrast test.The locomotive car weight of this experiment is 3440.00
Ton, vehicle commander are 66.9 meters.The curve of comparing result display prediction is basic consistent with the driving gear change curve of outstanding driver, says
Bright model proposed by the present invention has the level of the outstanding driver driving of the mankind；Corresponding model cootrol locomotive running speed is bent
Line and outstanding driver driving curve are almost the same, and it is especially noted that the lower bound in 412000 kilometer post attachmentes is fast
Under the conditions of, model has also effectively predicted suitable manipulation gear, and locomotive running speed also completely in safe range, does not produce
Raw hypervelocity equivalent risk situation.And it coincide for the study of gear change point is also substantially all.It is proposed by the present invention shown in sum up
A kind of locomotive gear prediction technique based on CNN models can make full use of locomotive historical data and road information, and combine pre- place
Obtained temporal aspect sequence is managed, the manipulation gear for ensureing the multiple targets such as safety, punctuality can be predicted.Based on true road
The experiment test of line and locomotive, it was demonstrated that the present invention has reached saving manpower object on the basis of ensureing reliability and safety
The automatic Pilot target of power.
Although the principle of the present invention is described in detail above in conjunction with the preferred embodiment of the present invention, this field skill
Art personnel are it should be understood that above-described embodiment is only the explanation to the exemplary implementation of the present invention, not to present invention packet
Restriction containing range.Details in embodiment is simultaneously not meant to limit the scope of the invention, without departing substantially from the present invention spirit and
In the case of range, any equivalent transformation, simple replacement based on technical solution of the present invention etc. obviously changes, and all falls within
Within the scope of the present invention.
Claims (6)
1. a kind of energy saving method of operating of locomotive based on convolutional neural networks model, which is characterized in that the method includes：
Step S101, acquisition driver's history driving data and locomotive operation monitor daily record, as initial training data；
Step S102 pre-processes initial training data, obtains training dataset and test data set；
Step S103, repetition training convolutional neural networks model simultaneously preserve trained model；
Step S104, the good model of application training is predicted, and carries out analog simulation test.
2. a kind of energy saving method of operating of locomotive based on convolutional neural networks model, which is characterized in that in step 1, the department of acquisition
Machine history driving data includes locomotive attribute, line properties and locomotive driving daily record.
3. the energy saving method of operating of the locomotive according to claim 1 based on convolutional neural networks model, is characterized in that, step
In S102, preprocessing process is divided into two stages：First stage chooses the data of driver driving behavior from original training data
Collection, second stage continue to pre-process the driving data selected.
4. the energy saving method of operating of the locomotive according to claim 3 based on convolutional neural networks model, is characterized in that, for
One locomotive driving state, the then feature extracted can be divided into two parts, be the traveling of the locomotive before the locomotive driving state respectively
Information and road information and the road information after the locomotive driving state.
5. the energy saving method of operating of the locomotive according to claim 4 based on convolutional neural networks model, is characterized in that, forward direction
Feature include car weight, vehicle commander, weight vehicle number, light vehicle number, when scarp slope section mean inclination, when scarp slope section total length, when scarp slope section
Average speed between two stations that average speed, the speed limit of current location, the residue length of current speed limit, current point are in
Degree, current location apart from next post distance, current location reach remaining time at next station, current gear, present speed,
The gradient of current location point, the difference of present speed and speed limit.
6. the energy saving method of operating of the locomotive according to claim 4 based on convolutional neural networks model, is characterized in that, backward
Feature include when scarp slope section mean inclination, when scarp slope section total length, when scarp slope section residue length, when scarp slope section average speed, when
Distance, the gradient of current location point, the speed and pumping of current locomotive of the speed limit, current location of front position apart from next station
Take the difference of feature locations point speed limit.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810192684.8A CN108333959A (en) | 2018-03-09 | 2018-03-09 | A kind of energy saving method of operating of locomotive based on convolutional neural networks model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810192684.8A CN108333959A (en) | 2018-03-09 | 2018-03-09 | A kind of energy saving method of operating of locomotive based on convolutional neural networks model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108333959A true CN108333959A (en) | 2018-07-27 |
Family
ID=62930620
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810192684.8A Pending CN108333959A (en) | 2018-03-09 | 2018-03-09 | A kind of energy saving method of operating of locomotive based on convolutional neural networks model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108333959A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108984275A (en) * | 2018-08-27 | 2018-12-11 | 洛阳中科龙网创新科技有限公司 | The agricultural driver training method of Intelligent unattended based on Unity3D and depth enhancing study |
| EP3620880A1 (en) * | 2018-09-04 | 2020-03-11 | Autonomous Intelligent Driving GmbH | A method and a device for deriving a driving strategy for a self-driving vehicle and an electronic control unit for performing the driving strategy and a self-driving vehicle comprising the electronic control unit |
| CN112331034A (en) * | 2020-11-13 | 2021-02-05 | 重庆交通大学 | High-speed train driver monitoring and assisting system |
| CN112698578A (en) * | 2019-10-22 | 2021-04-23 | 北京车和家信息技术有限公司 | Automatic driving model training method and related equipment |
| WO2021146910A1 (en) * | 2020-01-21 | 2021-07-29 | 深圳元戎启行科技有限公司 | Graph computation based vehicle driving data processing method and apparatus, and computer device |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016077027A1 (en) * | 2014-11-13 | 2016-05-19 | Nec Laboratories America, Inc. | Hyper-class augmented and regularized deep learning for fine-grained image classification |
| CN106156807A (en) * | 2015-04-02 | 2016-11-23 | 华中科技大学 | The training method of convolutional neural networks model and device |
| CN106842925A (en) * | 2017-01-20 | 2017-06-13 | 清华大学 | A kind of locomotive smart steering method and system based on deeply study |
| CN106844949A (en) * | 2017-01-18 | 2017-06-13 | 清华大学 | A kind of training method for realizing the controllable two-way LSTM models of locomotive section |
| CN107423761A (en) * | 2017-07-24 | 2017-12-01 | 清华大学 | Feature based selects and the rail locomotive energy saving optimizing method of operating of machine learning |
| CN107516110A (en) * | 2017-08-22 | 2017-12-26 | 华南理工大学 | A kind of medical question and answer Semantic Clustering method based on integrated convolutional encoding |
-
2018
- 2018-03-09 CN CN201810192684.8A patent/CN108333959A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016077027A1 (en) * | 2014-11-13 | 2016-05-19 | Nec Laboratories America, Inc. | Hyper-class augmented and regularized deep learning for fine-grained image classification |
| CN106156807A (en) * | 2015-04-02 | 2016-11-23 | 华中科技大学 | The training method of convolutional neural networks model and device |
| CN106844949A (en) * | 2017-01-18 | 2017-06-13 | 清华大学 | A kind of training method for realizing the controllable two-way LSTM models of locomotive section |
| CN106842925A (en) * | 2017-01-20 | 2017-06-13 | 清华大学 | A kind of locomotive smart steering method and system based on deeply study |
| CN107423761A (en) * | 2017-07-24 | 2017-12-01 | 清华大学 | Feature based selects and the rail locomotive energy saving optimizing method of operating of machine learning |
| CN107516110A (en) * | 2017-08-22 | 2017-12-26 | 华南理工大学 | A kind of medical question and answer Semantic Clustering method based on integrated convolutional encoding |
Non-Patent Citations (1)
| Title |
|---|
| 谈东奎: "人机共享的驾驶员横向辅助系统关键技术研究", 《中国博士学位论文全文数据库（电子期刊） 工程科技Ⅱ辑》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108984275A (en) * | 2018-08-27 | 2018-12-11 | 洛阳中科龙网创新科技有限公司 | The agricultural driver training method of Intelligent unattended based on Unity3D and depth enhancing study |
| EP3620880A1 (en) * | 2018-09-04 | 2020-03-11 | Autonomous Intelligent Driving GmbH | A method and a device for deriving a driving strategy for a self-driving vehicle and an electronic control unit for performing the driving strategy and a self-driving vehicle comprising the electronic control unit |
| CN112698578A (en) * | 2019-10-22 | 2021-04-23 | 北京车和家信息技术有限公司 | Automatic driving model training method and related equipment |
| CN112698578B (en) * | 2019-10-22 | 2023-11-14 | 北京车和家信息技术有限公司 | Training method of automatic driving model and related equipment |
| WO2021146910A1 (en) * | 2020-01-21 | 2021-07-29 | 深圳元戎启行科技有限公司 | Graph computation based vehicle driving data processing method and apparatus, and computer device |
| CN112331034A (en) * | 2020-11-13 | 2021-02-05 | 重庆交通大学 | High-speed train driver monitoring and assisting system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108333959A (en) | A kind of energy saving method of operating of locomotive based on convolutional neural networks model | |
| CN106844949A (en) | A kind of training method for realizing the controllable two-way LSTM models of locomotive section | |
| CN103679611B (en) | Operation method of city comprehensive emergency intelligent simulation system based on case-based reasoning | |
| CN112465395A (en) | Multi-dimensional comprehensive evaluation method and device for automatically-driven automobile | |
| CN112100865A (en) | Method for predicting remaining life of aircraft engine based on parallel CNN model | |
| CN107697070A (en) | Driving behavior Forecasting Methodology and device, unmanned vehicle | |
| CN107909206A (en) | A kind of PM2.5 Forecasting Methodologies based on deep structure Recognition with Recurrent Neural Network | |
| CN110481536B (en) | Control method and device applied to hybrid electric vehicle | |
| CN106781503A (en) | Method and apparatus for monitoring driving behavior | |
| CN112508164B (en) | End-to-end automatic driving model pre-training method based on asynchronous supervised learning | |
| CN103345165A (en) | Intelligent transportation hardware online simulation system of miniature intelligent vehicle group | |
| CN110555568A (en) | Road traffic running state real-time perception method based on social network information | |
| CN114358128A (en) | Method for training end-to-end automatic driving strategy | |
| CN104200422A (en) | Expert system for remote sensing image processing | |
| CN114282433A (en) | Automatic driving training method and system based on combination of simulation learning and reinforcement learning | |
| CN102262728A (en) | Road traffic sign identification method | |
| CN106647279B (en) | A kind of locomotive smart steering optimized calculation method based on fuzzy rule | |
| CN116050304B (en) | Intelligent fish flow field simulation control method, system, equipment and storage medium | |
| Ma et al. | Effects of driver acceleration behavior on fuel consumption of city buses | |
| Wang et al. | Pre-training with asynchronous supervised learning for reinforcement learning based autonomous driving | |
| Li et al. | Short term passenger flow forecast of metro based on inbound passenger plow and deep learning | |
| CN113487889A (en) | Traffic state anti-disturbance generation method based on single intersection signal control of rapid gradient descent | |
| CN116620327A (en) | Lane changing decision method for realizing automatic driving high-speed scene based on PPO and Lattice | |
| Wu et al. | A deep reinforcement learning based car following model for electric vehicle | |
| CN115096305A (en) | Intelligent driving automobile path planning system and method based on generation of countermeasure network and simulation learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180727 |
|
| RJ01 | Rejection of invention patent application after publication |