CN106990714A - Adaptive Control Method and device based on deep learning - Google Patents
Adaptive Control Method and device based on deep learning Download PDFInfo
- Publication number
- CN106990714A CN106990714A CN201710411253.1A CN201710411253A CN106990714A CN 106990714 A CN106990714 A CN 106990714A CN 201710411253 A CN201710411253 A CN 201710411253A CN 106990714 A CN106990714 A CN 106990714A
- Authority
- CN
- China
- Prior art keywords
- car
- vehicle
- driving
- deep learning
- control method
- 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
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/0265—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric the criterion being a learning criterion
- G05B13/027—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric the criterion being a learning criterion using neural networks only
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course or altitude of land, water, air, or space vehicles, e.g. automatic pilot
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
- G05D1/0221—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory involving a learning process
Abstract
The invention discloses a kind of Adaptive Control Method and device based on deep learning, using deep neural network learning experience driver under a variety of operating modes of driving, fed back by body-sensing, On-line Control of the people in loop is realized to vehicle, generate adaptation controller, applied to driving car certainly, it is adaptable to various vehicles.Beneficial effects of the present invention are, pass through the cognitive behavior of materialization experience driver, autonomous driving control is decoupled, using the Recognition with Recurrent Neural Network architecture design in deep learning and realize from drive car adaptation controller, adaptation controller receives to drive the cognitive arrow commander of brain decision making in the application, tackle the uncertain control vehicle traveling such as road, weather, load, it is ensured that from safe, steady, the energy-conservation for driving car.
Description
Technical field
The invention belongs to unmanned technical field, more particularly to a kind of Adaptive Control Method and dress based on deep learning
Put.
Background technology
Unmanned technology is related to computer science, communication science, cognitive science, Vehicle Engineering, electric and electronic engineering, control
The subjects such as scientific and engineering processed, systematic science and technology, human engineering science, artificial intelligence.It is a kind of wheel from car is driven
Formula mobile robot, is the unmanned product for developing into advanced stage, is to weigh national a research strength and industrial level
One of important symbol.
It is related to two aspects of soft and hardware from the exploitation for driving car.Software aspects need dozens or even hundreds of software mould
Block cooperates, and completes environment sensing, drives cognition, intelligent decision, path planning, the task such as automatically controls.Hardware aspect is needed
A variety of different types of sensors to be installed in vehicle diverse location, meet real-time, accurate, Overall Acquisition surrounding enviroment information
It is required that.Simultaneously, it is necessary to carry out line traffic control repacking and electric transformation to vehicle, and it is equipped with the sufficient master controller of computing resource.
However, different types of vehicle dynamics characteristics are different, in order to ensure implementing safe and reliable autonomous driving control, car to it
Chassis is completed after line traffic control repacking, and people are often through PID (Proportion Integration Differentiation)
Control ensures the stationarity and robustness that vehicle is manipulated automatically, or passes through Model Predictive Control (Model Predictive
Control, MPC) pid parameter to be adjusted, difficulty is big, and experience cost and time cost are high, and parameter is difficult to optimize, usually
Attend to one thing and lose sight of another, it is in a dilemma.
The content of the invention
It is an object of the invention to overcome above-mentioned deficiency of the prior art, and provide a kind of adaptation based on deep learning
Control method and device.
To realize the purpose of the present invention, the invention provides a kind of Adaptive Control Method based on deep learning, depth is utilized
Neural network learning experience driver is spent under a variety of operating modes of driving, and is fed back by body-sensing, realizes people in loop on vehicle
On-line Control, generate adaptation controller, applied to from drive car, it is adaptable to various vehicles.
Correspondingly, a kind of adaptation control device based on deep learning is additionally provided, including big-sample data collecting unit,
Deep neural network training unit and adaptation controller：
The big-sample data collecting unit, for gathering experience driver under a variety of operating modes of driving, operating and controlling vehicle
The parameter of data and reflection travel condition of vehicle；
The deep neural network training unit, for being obtained using a general Recognition with Recurrent Neural Network framework study is above-mentioned
The big-sample data obtained, obtains the weight of each state parameter in the network structure and network structure of stabilization, forms data-driven
Model；
The adaptation controller, is, into adaptation controller chip, car to be driven applied to oneself by modelling obtained above,
Suitable for various vehicles.
Compared with prior art, beneficial effects of the present invention are, by the cognitive behavior of materialization experience driver, to incite somebody to action autonomous
Driving control is decoupled, using the Recognition with Recurrent Neural Network architecture design in deep learning and realize from drive car adaptation control
Device processed, in the application adaptation controller receiving drives the cognitive arrow commander of brain decision making, reply road, weather, load
Deng uncertainty control vehicle traveling, it is ensured that from safe, steady, the energy-conservation for driving car.
Brief description of the drawings
Fig. 1 autonomous drivings control decoupling figure；
Fig. 2 is whole vehicle state adaptation control schematic diagram；
Fig. 3 is a long short-term memory temporally deployed (Long-Short Term Memory, LSTM) network structure
Schematic diagram；
Fig. 4 is helix road schematic diagram.The radial dimension in place, should meet vehicle it is a certain determination speed under turn to when
Tire does not break away.For example, when speed is 36km/h, 260m × 230m is at least wanted in place；When speed is 72km/h, place is at least
Want 500m × 360m.
Fig. 5 is adaptation controller design framework figure；
Fig. 6 is deep neural network structural representation in preferred embodiment.
Embodiment
The present invention is described in further detail below in conjunction with the drawings and specific embodiments.It should be appreciated that described herein
Specific embodiment only to explain the present invention, be not intended to limit the present invention.
It should be noted that " connection " described herein and the word for expressing " connection ", such as " being connected ",
" connected " etc., was both directly connected to including a certain part with another part, also passed through miscellaneous part and another portion including a certain part
Part is connected.
It should be noted that in the case where not conflicting, the feature in embodiment and embodiment in the application can phase
Mutually combination.
The success of century-old car industry is the success of ergonomics, is the decoupling of automobile longitudinal motion and transverse movement, is
Driver to steering wheel, throttle and braking can with feel qualitatively manipulate, it is controllable need not to survey, be cognitive aspect (rather than
Data and signal aspect) Autonomic prediction and control of the people in loop are realized, tackle the uncertainty in vehicle traveling.Driver is led to
Cross steering wheel, throttle and the feedback control of braking critically important, different vehicle, different load, different weather, different road surfaces, difference
When road curvature, friction speed, Real-time Feedback of the people that the on-line operation of driver is formed in loop：By holding steering wheel
Tire corner, rotation and sliding that feel feedback comes；By the feel of pedal, by vision, the sense of hearing, smell and body-sensing,
The acceleration, deceleration, car body for knowing from experience vehicle are turned round, vibrate, shake, jolted, and are formed people in the real-time online feedback in loop and are controlled
System, is accumulated as driving efficiency.
Experience driver drives, and not only to meet driving norms, and safety, civilization traveling, its experience are also embodied in energy-conservation skill
Ingeniously, ride comfort, to the vehicle compatibility of different dynamic in terms of.The experience of driver and behavior are such as the walking appearance of people
State is different, can also referred to as drive fingerprint with identification of the driving behavior as driver.Mark post driver is that experience is driven
Outstanding representative in the person of sailing.
Autonomous driving hardly possible will study driving behavior and psychology, the perception of materialization driver, decision-making, note in personification
Recall, control and behavior skill, rather than improve the dynamic performance of vehicle simply, realize automatic Pilot.Researched and developed from car is driven
Difficulty, the not exclusively property of automobile dynamics and various sensor requirements, it is often more important that to research and develop and drive
Online " machine driving brain " as member, Autonomic prediction and control of the simulated implementation people in loop.Autonomous driving is controlled to decouple
For the wagon control of different levels, such as Fig. 1 shows：Large scale is seen, is the tracing control to driving trace；Mesoscale is seen, is to whole
The control of car state；Small yardstick is seen, is to the control inside each actuator.The present invention focuses on the control of the whole vehicle state of mesoscale
Preparative layer face, such as Fig. 2 show, using long short-term memory (Long-Short Term Memory, LSTM) model in deep learning, adopt
With the sensing data for reflecting vehicle-state on automobile CAN-bus, such as speed, vibration, the tire pressure of tire, corner and rotating speed are learned
Habit experience driver strong wind, fully loaded, bend, inclination, ice and snow, paddle, it is muddy and cheat recessed road surface etc. and realize that people exists by body-sensing
The On-line Control in loop, generation obtains " driving for experience driver from the adaptation Controlling model (being commonly called as " driving cerebellum ") for driving car
Sail fingerprint ", applied to driving car certainly, it is ensured that steady, comfortable, the energy-conservation of vehicle traveling.LSTM is a kind of Recognition with Recurrent Neural Network
(Recurrent Neural Networks, RNN), avoids perplexing traditional RNN networks by a kind of structure for being referred to as " door "
Long-term Dependence Problem.Fig. 3 is a LSTM schematic network structure temporally deployed.In LSTM networks, the transmission of state
Long Memory control in short-term is realized by " forgetting door ", " input gate ", " out gate " etc..
Specifically, the invention discloses a kind of Adaptive Control Method based on deep learning, deep neural network is utilized
Habit experience driver is fed back under a variety of operating modes of driving by body-sensing, and On-line Control of the people in loop is realized to vehicle, raw
Into adaptation controller, applied to driving car certainly, it is adaptable to various vehicles.
Wherein, a variety of operating modes of driving can reflect the uncertainty for operating mode of being driven in practical application, including：Not on the same day
Gas and/or different road surfaces, different vehicle platform, different loads, different road curvatures, friction speed.
Operating mode of typically driving has strong wind, is fully loaded with, bend, rolls, and ice and snow is paddled, and muddy and hole is recessed.
Wherein, applicable vehicle includes the car category provided in standard GB/T/T 3730.1-2001, including：Commonly
Passenger car, top passenger car living, high-end vehicles, small passenger car, open car, storehouse back of the body passenger car, station wagon, multipurpose are riding
Car, brachycephaly passenger car, cross-country passenger car, special passenger car, lodging vehicle, bullet-proof car, ambulance, hearse, car, small-sized visitor
Car, city bus, coach, sightseeing bus, articulated coach, trolleybus, cross-country car, special coach, tractor truck,
Lorry, general wagon, multiple purpose truck, full ailer-towing vehicle, off-the-road truck, special operation car, unitrain, truck combination.
Wherein, the Adaptive Control Method based on deep learning specifically includes following steps：
(1) the big-sample data collection of experience driver operating and controlling vehicle；
(2) deep neural network is trained, using a general Recognition with Recurrent Neural Network framework to the large sample in step (1)
Data carry out deep learning, obtain stable network structure, form the model of data-driven；
(3) the modelling adaptation controller chip obtained using step (2).
Wherein, above-mentioned steps (1) are specifically included：
Experience driver drives vehicle in the case where linear road, helix road and difference park operating mode, and road will embody not
In addition pavement friction, different pitching with attachment, operating mode of parking will also embody the manipulation to vehicle in small space；
(1.1) according to the primary system plan track on driving map, experience driver is driven by pre-set velocity straight line, or presses spiral shell
Spin line driving path, such as Fig. 4 show that the lateral error for deviateing desired guiding trajectory is maintained in prescribed limit；
According to the primary system plan track on driving map, experience driver is completed to limit in spatial dimension by pre-set velocity
Park, the lateral error for deviateing desired guiding trajectory is maintained in prescribed limit；
(1.2) change preset vehicle speed to continue repeatedly to test, and record data.
Wherein, in step (1) experience driver drive operating mode setting, it is necessary to reference to specific adaptation controller applied field
Scape.
Wherein, the parameter of the reflection travel condition of vehicle of record is in step (1)：Body speed of vehicle, subtotal mileage, engine
Power, car body course angle, the angle of pitch, side drift angle, 4 wheel or 2 driving wheels rotating speed, 4 wheel or 2 driving wheels tire pressure, 2 turn
To the corner of wheel, body oscillating frequency, body oscillating amplitude.
Wherein, the general Recognition with Recurrent Neural Network framework used in step (2) uses LSTM models.
Wherein, applied to from when driving car, the input of adaptation controller is the accelerator pedal displacement at current time, brakes and step on
Plate displacement, steering wheel angle and travel condition of vehicle parameter, are output as accelerator pedal displacement, the brake pedal position of subsequent time
Shifting, steering wheel angle, so as to realize to the control from driving car.
As shown in figure 5, giving adaptation controller design framework figure.As seen from the figure, the controlled quentity controlled variable that the t+ τ moment exports be by
What the controlled quentity controlled variable and travel condition of vehicle parameter of t were together decided on, namely：
Wherein, x1It is steering wheel angle, x2It is driving pedal displacement, x3It is brake pedal displacement；α1,...,αnIt is n car
Running state parameter, is followed successively by, is shown in Table 2：
The travel condition of vehicle parameter list of table 2
Due to being difficult to find or even f determination analytic expression can not be solved at all, therefore utilize a general circulation god
Deep learning is carried out to the big-sample data in step 1 through the network architecture, the weight of each state parameter in network structure is obtained,
Form the model of data-driven.The general Recognition with Recurrent Neural Network framework uses long short-term memory (LSTM) model.
Describe the embodiment of the present invention in detail with reference to preferred embodiment.
Drive to be kept straight at test site by experience driver, turn left, turning right, going up a slope, the traveling, data acquisition frequency such as descending
Rate is 100Hz, and the partial data example such as table 3 of acquisition and recording shows.
The gathered data example of table 3
Note：The meaning that secondary series " throttle shift voltage " data are stated in upper table is equal to driving pedal displacement, the 3rd row
The meaning of " brake pressure " data statement is equal to brake pedal displacement, and the meaning of the 9th row " jolting " data statement is equal to car
Body Oscillation Amplitude.
The model of data-driven by training generation is as follows, including seven-layer structure：
First layer is input layer, for initial data to be accessed into Recognition with Recurrent Neural Network；
The second layer is classification presentation layer, and input layer is connected to classification presentation layer with full connected mode；Presentation layer of classifying is set
64 neurons, classification presentation layer has two effects：One is that initial data is reclassified and combed, and two be isolation input
Layer with LSTM layers, it is ensured that even if parameter spread or change, do not interfere with LSTM layers of network structure yet；
Third layer is LSTM memory layers 1, for integrating current input information and previous experiences information, carries out preliminary function and reflects
Penetrate；
4th layer is LSTM memory layers 2, and the function representation ability for strengthening network enables the network to support more complicated
The mapping of memory problems；It is 32 that LSTM, which remembers layer 1 and LSTM memory layer 2 node numbers,；
Layer 5 is middle decision-making level, for the output state after two layers of LSTM memory layer processing linearly to be reflected
Penetrate, form preliminary decision-making；
Layer 6 is output decision-making level, is finally arranged for the output to middle decision-making level；
Layer 7 is output layer, and the variable of output is directly used in steering wheel, throttle and the control of brake.
(3) the modelling adaptation controller of above-mentioned acquisition is utilized.
In the chip that the model of above-mentioned acquisition is write to adaptation controller using computer program, when in use, receive to drive
The cognitive arrow commander for sailing brain decision making, outbound course disk corner, accelerator pedal displacement, brake pedal are moved to execution
Device, controls vehicle smooth-ride.With above-mentioned Adaptive Control Method accordingly, present invention also offers one kind be based on deep learning
Adaptation control device, including big-sample data collecting unit, deep neural network training unit and adaptation controller：
Specifically, the big-sample data collecting unit, for gathering experience driver under a variety of operating modes of driving, manipulation
The data of vehicle and the parameter of reflection travel condition of vehicle；
The deep neural network training unit, for being obtained using a general Recognition with Recurrent Neural Network framework study is above-mentioned
The big-sample data obtained, obtains the weight of each state parameter in the network structure and network structure of stabilization, forms data-driven
Model；
The adaptation controller, is, into adaptation controller chip, car to be driven applied to oneself by modelling obtained above,
Suitable for various vehicles.Specifically, the general Recognition with Recurrent Neural Network framework uses LSTM models.
Specifically, applied to from when driving car, the input of adaptation controller is the accelerator pedal displacement at current time, braking
Pedal displacement, steering wheel angle and travel condition of vehicle parameter, are output as accelerator pedal displacement, the brake pedal position of subsequent time
Shifting, steering wheel angle, so as to realize to the control from driving car.
Described above is only the preferred embodiment of the present invention, it is noted that for the common skill of the art
For art personnel, under the premise without departing from the principles of the invention, some improvements and modifications can also be made, these improvements and modifications
Also it should be regarded as protection scope of the present invention.
Claims (10)
1. a kind of Adaptive Control Method based on deep learning, it is characterised in that driven using deep neural network learning experience
Member is fed back under a variety of operating modes of driving by body-sensing, and On-line Control of the people in loop, generation adaptation control are realized to vehicle
Device, applied to driving car certainly, it is adaptable to various vehicles.
2. the Adaptive Control Method according to claim 1 based on deep learning, it is characterised in that a variety of works of driving
Condition can reflect the uncertainty for operating mode of being driven in practical application, including：Different weather and/or different road surfaces, different vehicle are put down
Platform, different loads, different road curvatures, friction speed.
3. the Adaptive Control Method according to claim 1 based on deep learning, it is characterised in that the various vehicles refer to
The car category provided in national standard, including：It is ordinary passenger car, top passenger car living, high-end vehicles, small passenger car, spacious
Van, storehouse the back of the body passenger car, station wagon, multipurpose passenger car, brachycephaly passenger car, cross-country passenger car, special passenger car, lodging vehicle,
Bullet-proof car, ambulance, hearse, car, station wagon, city bus, coach, sightseeing bus, articulated coach, trackless electricity
Car, cross-country car, special coach, tractor truck, lorry, general wagon, multiple purpose truck, full ailer-towing vehicle, off-the-road truck,
Special operation car, unitrain, truck combination.
4. the Adaptive Control Method based on deep learning according to claim 1 or 2 or 3, it is characterised in that the adaptation
Control method specifically includes following steps：
(1) the big-sample data collection of experience driver operating and controlling vehicle；
(2) deep neural network is trained, using a general Recognition with Recurrent Neural Network framework to the big-sample data in step (1)
Deep learning is carried out, stable network structure is obtained, the model of data-driven is formed；
(3) the modelling adaptation controller chip obtained using step (2).
5. the Adaptive Control Method according to claim 4 based on deep learning, it is characterised in that step (1) is specifically wrapped
Include：
Experience driver drives vehicle in the case where linear road, helix road and difference park operating mode, and road will embody different attached
In addition the pavement friction, different pitching, operating mode of parking will also embody the manipulation to vehicle in small space；
(1.1) according to the primary system plan track on driving map, experience driver is driven by pre-set velocity straight line, or by helix
Driving path, the lateral error for deviateing desired guiding trajectory is maintained in prescribed limit；
According to the primary system plan track on driving map, experience driver is completed to limit the pool in spatial dimension by pre-set velocity
Car, the lateral error for deviateing desired guiding trajectory is maintained in prescribed limit；
(1.2) change preset vehicle speed to continue repeatedly to test, and record data.
6. the Adaptive Control Method according to claim 4 based on deep learning, it is characterised in that experience in step (1)
Driver drive operating mode setting, it is necessary to reference to specific adaptation controller application scenarios.
7. the Adaptive Control Method according to claim 4 based on deep learning, it is characterised in that recorded in step (1)
The parameter of reflection travel condition of vehicle be：Body speed of vehicle, subtotal mileage, engine power, car body course angle, the angle of pitch, side
Drift angle, the rotating speed of 4 wheels or 2 driving wheels, the tire pressure of 4 wheels or 2 driving wheels, the corner of 2 deflecting rollers, body oscillating frequency,
Oscillation Amplitude.
8. the Adaptive Control Method according to claim 4 based on deep learning, it is characterised in that described in step (2)
General Recognition with Recurrent Neural Network framework uses long short-term memory (Long-Short Term Memory, LSTM) model.
9. the Adaptive Control Method according to claim 1 based on deep learning, it is characterised in that applied to driving car certainly
When, the input of adaptation controller runs for the accelerator pedal displacement at current time, brake pedal displacement, steering wheel angle and vehicle
State parameter, is output as the accelerator pedal displacement of subsequent time, brake pedal displacement, steering wheel angle, so as to realize to self-driving
Sail the control of car.
10. a kind of adaptation control device based on deep learning, it is characterised in that including big-sample data collecting unit, depth
Neural metwork training unit and adaptation controller：
The big-sample data collecting unit, for gathering experience driver under a variety of operating modes of driving, the data of operating and controlling vehicle
With the parameter of reflection travel condition of vehicle；
The deep neural network training unit, for learning above-mentioned acquisition using a general Recognition with Recurrent Neural Network framework
Big-sample data, obtains the weight of each state parameter in the network structure and network structure of stabilization, forms the mould of data-driven
Type；
The adaptation controller, is, into adaptation controller chip, applied to from car is driven, to be applicable modelling obtained above
In various vehicles.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710411253.1A CN106990714A (en) | 2017-06-05 | 2017-06-05 | Adaptive Control Method and device based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710411253.1A CN106990714A (en) | 2017-06-05 | 2017-06-05 | Adaptive Control Method and device based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106990714A true CN106990714A (en) | 2017-07-28 |
Family
ID=59421302
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710411253.1A Pending CN106990714A (en) | 2017-06-05 | 2017-06-05 | Adaptive Control Method and device based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106990714A (en) |
Cited By (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107450593A (en) * | 2017-08-30 | 2017-12-08 | 清华大学 | A kind of unmanned plane autonomous navigation method and system |
| CN108216233A (en) * | 2017-12-28 | 2018-06-29 | 北京经纬恒润科技有限公司 | A kind of scaling method and device of self-adaption cruise system control parameter |
| CN108469815A (en) * | 2018-02-27 | 2018-08-31 | 重庆嵩岳贸易服务有限公司 | A kind of self-navigation of computer deep learning and control loop and its method based on intention |
| CN108470460A (en) * | 2018-04-11 | 2018-08-31 | 江苏大学 | A kind of nearby vehicle Activity recognition method based on smart mobile phone and RNN |
| CN108897313A (en) * | 2018-05-23 | 2018-11-27 | 清华大学 | A kind of end-to-end Vehicular automatic driving system construction method of layer-stepping |
| CN109035873A (en) * | 2018-08-14 | 2018-12-18 | 黄芷姗 | A kind of method of adjustment and adjustment device of unmanned aerial vehicle (UAV) control permission |
| CN109753047A (en) * | 2017-11-08 | 2019-05-14 | 通用汽车环球科技运作有限责任公司 | System and method for autonomous vehicle behaviour control |
| CN110221611A (en) * | 2019-06-11 | 2019-09-10 | 北京三快在线科技有限公司 | A kind of Trajectory Tracking Control method, apparatus and automatic driving vehicle |
| CN110422171A (en) * | 2018-04-27 | 2019-11-08 | 通用汽车环球科技运作有限责任公司 | The autonomous driving learnt using driver neural network based |
| CN110633596A (en) * | 2018-06-21 | 2019-12-31 | 北京京东尚科信息技术有限公司 | Method and device for predicting vehicle direction angle |
| CN110823603A (en) * | 2019-11-19 | 2020-02-21 | 上汽通用汽车有限公司 | Full-time monitoring system and method for automobile chassis, vehicle and storage medium |
| CN111231588A (en) * | 2020-01-17 | 2020-06-05 | 新石器慧通（北京）科技有限公司 | Tire pressure monitoring method and device, vehicle-mounted equipment and storage medium |
| CN111258217A (en) * | 2018-11-30 | 2020-06-09 | 百度（美国）有限责任公司 | Real-time object behavior prediction |
| CN111433102A (en) * | 2017-12-18 | 2020-07-17 | 智加科技公司 | Method and system for aggregate vehicle control prediction in autonomous vehicles |
| US10733510B2 (en) | 2018-08-24 | 2020-08-04 | Ford Global Technologies, Llc | Vehicle adaptive learning |
| CN112034829A (en) * | 2019-05-15 | 2020-12-04 | 广州汽车集团股份有限公司 | End-to-end automatic driving method and system and vehicle |
| CN112092825A (en) * | 2020-08-31 | 2020-12-18 | 英博超算(南京)科技有限公司 | Lane keeping method based on machine learning |
| CN112269314A (en) * | 2020-09-30 | 2021-01-26 | 南京航空航天大学 | Automobile chassis decoupling control method based on LSTM neural network |
| CN113104010A (en) * | 2021-05-07 | 2021-07-13 | 的卢技术有限公司 | Vehicle brake control method, system, computer device and storage medium |
| CN113753063A (en) * | 2020-11-23 | 2021-12-07 | 北京京东乾石科技有限公司 | Vehicle driving instruction determination method, device, equipment and storage medium |
| CN114237049A (en) * | 2021-12-14 | 2022-03-25 | 西安建筑科技大学 | Intelligent building system prediction control parameter setting method based on LSTM |
| US11650586B2 (en) | 2017-12-18 | 2023-05-16 | Plusai, Inc. | Method and system for adaptive motion planning based on passenger reaction to vehicle motion in autonomous driving vehicles |
| US11884287B2 (en) | 2021-08-19 | 2024-01-30 | Ford Global Technologies, Llc | Enhanced vehicle operation |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5162997A (en) * | 1990-01-23 | 1992-11-10 | Nissan Motor Company, Limited | Control system for automotive vehicle for controlling vehicle driving behavior with feature of harmonization of vehicular driving condition dependent control and driver's driving tendency adapted control |
| CN101078935A (en) * | 2007-06-28 | 2007-11-28 | 华南农业大学 | Agricultural machine path tracking control method based on nerve network |
| CN105046196A (en) * | 2015-06-11 | 2015-11-11 | 西安电子科技大学 | Front vehicle information structured output method base on concatenated convolutional neural networks |
| CN105136469A (en) * | 2015-07-23 | 2015-12-09 | 江苏大学 | Unmanned vehicle speed control method based on PSO and RBF neutral network |
| CN105575119A (en) * | 2015-12-29 | 2016-05-11 | 大连楼兰科技股份有限公司 | Road condition climate deep learning and recognition method and apparatus |
| CN105892471A (en) * | 2016-07-01 | 2016-08-24 | 北京智行者科技有限公司 | Automatic automobile driving method and device |
| CN106096767A (en) * | 2016-06-07 | 2016-11-09 | 中国科学院自动化研究所 | A kind of link travel time prediction method based on LSTM |
| CN106203626A (en) * | 2016-06-30 | 2016-12-07 | 北京奇虎科技有限公司 | Car steering behavioral value method and device, automobile |
| CN106503393A (en) * | 2016-11-15 | 2017-03-15 | 浙江大学 | A kind of method for realizing that using emulation generation sample unmanned vehicle is independently advanced |
-
2017
- 2017-06-05 CN CN201710411253.1A patent/CN106990714A/en active Pending
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5162997A (en) * | 1990-01-23 | 1992-11-10 | Nissan Motor Company, Limited | Control system for automotive vehicle for controlling vehicle driving behavior with feature of harmonization of vehicular driving condition dependent control and driver's driving tendency adapted control |
| CN101078935A (en) * | 2007-06-28 | 2007-11-28 | 华南农业大学 | Agricultural machine path tracking control method based on nerve network |
| CN105046196A (en) * | 2015-06-11 | 2015-11-11 | 西安电子科技大学 | Front vehicle information structured output method base on concatenated convolutional neural networks |
| CN105136469A (en) * | 2015-07-23 | 2015-12-09 | 江苏大学 | Unmanned vehicle speed control method based on PSO and RBF neutral network |
| CN105575119A (en) * | 2015-12-29 | 2016-05-11 | 大连楼兰科技股份有限公司 | Road condition climate deep learning and recognition method and apparatus |
| CN106096767A (en) * | 2016-06-07 | 2016-11-09 | 中国科学院自动化研究所 | A kind of link travel time prediction method based on LSTM |
| CN106203626A (en) * | 2016-06-30 | 2016-12-07 | 北京奇虎科技有限公司 | Car steering behavioral value method and device, automobile |
| CN105892471A (en) * | 2016-07-01 | 2016-08-24 | 北京智行者科技有限公司 | Automatic automobile driving method and device |
| CN106503393A (en) * | 2016-11-15 | 2017-03-15 | 浙江大学 | A kind of method for realizing that using emulation generation sample unmanned vehicle is independently advanced |
Non-Patent Citations (5)
| Title |
|---|
| BOJARSKI M等: "End to end learning for self-driving cars", 《ARXIV PREPRINT ARXIV》 * |
| HUVAL, B.等: "An empirical evaluation of deep learning on highway driving", 《ARXIV PREPRINT ARXIV》 * |
| 柳金梅，等: "BP神经网络在智能车辆方向控制中的应用", 《自动化与仪器仪表》 * |
| 梁军，等: "基于动态神经网络集成的驾驶员行为学习算法", 《计算机应用研究》 * |
| 舒飞，等: "基于神经网络的智能车转向控制系统设计与实现", 《第六届中国智能交通年会暨第七届国际节能与新能源汽车创新发展论坛优秀论文集（智能交通）》 * |
Cited By (29)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107450593A (en) * | 2017-08-30 | 2017-12-08 | 清华大学 | A kind of unmanned plane autonomous navigation method and system |
| CN107450593B (en) * | 2017-08-30 | 2020-06-12 | 清华大学 | Unmanned aerial vehicle autonomous navigation method and system |
| CN109753047A (en) * | 2017-11-08 | 2019-05-14 | 通用汽车环球科技运作有限责任公司 | System and method for autonomous vehicle behaviour control |
| US11643086B2 (en) | 2017-12-18 | 2023-05-09 | Plusai, Inc. | Method and system for human-like vehicle control prediction in autonomous driving vehicles |
| CN111433102A (en) * | 2017-12-18 | 2020-07-17 | 智加科技公司 | Method and system for aggregate vehicle control prediction in autonomous vehicles |
| US11650586B2 (en) | 2017-12-18 | 2023-05-16 | Plusai, Inc. | Method and system for adaptive motion planning based on passenger reaction to vehicle motion in autonomous driving vehicles |
| CN111433102B (en) * | 2017-12-18 | 2024-01-16 | 智加科技公司 | Method and system for collective vehicle control prediction in an autonomous vehicle |
| CN108216233B (en) * | 2017-12-28 | 2019-10-15 | 北京经纬恒润科技有限公司 | A kind of scaling method and device of self-adaption cruise system control parameter |
| CN108216233A (en) * | 2017-12-28 | 2018-06-29 | 北京经纬恒润科技有限公司 | A kind of scaling method and device of self-adaption cruise system control parameter |
| CN108469815A (en) * | 2018-02-27 | 2018-08-31 | 重庆嵩岳贸易服务有限公司 | A kind of self-navigation of computer deep learning and control loop and its method based on intention |
| CN108470460A (en) * | 2018-04-11 | 2018-08-31 | 江苏大学 | A kind of nearby vehicle Activity recognition method based on smart mobile phone and RNN |
| CN110422171A (en) * | 2018-04-27 | 2019-11-08 | 通用汽车环球科技运作有限责任公司 | The autonomous driving learnt using driver neural network based |
| CN108897313A (en) * | 2018-05-23 | 2018-11-27 | 清华大学 | A kind of end-to-end Vehicular automatic driving system construction method of layer-stepping |
| CN110633596A (en) * | 2018-06-21 | 2019-12-31 | 北京京东尚科信息技术有限公司 | Method and device for predicting vehicle direction angle |
| CN109035873A (en) * | 2018-08-14 | 2018-12-18 | 黄芷姗 | A kind of method of adjustment and adjustment device of unmanned aerial vehicle (UAV) control permission |
| US10733510B2 (en) | 2018-08-24 | 2020-08-04 | Ford Global Technologies, Llc | Vehicle adaptive learning |
| CN111258217A (en) * | 2018-11-30 | 2020-06-09 | 百度（美国）有限责任公司 | Real-time object behavior prediction |
| CN112034829A (en) * | 2019-05-15 | 2020-12-04 | 广州汽车集团股份有限公司 | End-to-end automatic driving method and system and vehicle |
| CN110221611A (en) * | 2019-06-11 | 2019-09-10 | 北京三快在线科技有限公司 | A kind of Trajectory Tracking Control method, apparatus and automatic driving vehicle |
| CN110221611B (en) * | 2019-06-11 | 2020-09-04 | 北京三快在线科技有限公司 | Trajectory tracking control method and device and unmanned vehicle |
| CN110823603A (en) * | 2019-11-19 | 2020-02-21 | 上汽通用汽车有限公司 | Full-time monitoring system and method for automobile chassis, vehicle and storage medium |
| CN111231588A (en) * | 2020-01-17 | 2020-06-05 | 新石器慧通（北京）科技有限公司 | Tire pressure monitoring method and device, vehicle-mounted equipment and storage medium |
| CN112092825A (en) * | 2020-08-31 | 2020-12-18 | 英博超算(南京)科技有限公司 | Lane keeping method based on machine learning |
| CN112269314A (en) * | 2020-09-30 | 2021-01-26 | 南京航空航天大学 | Automobile chassis decoupling control method based on LSTM neural network |
| CN113753063A (en) * | 2020-11-23 | 2021-12-07 | 北京京东乾石科技有限公司 | Vehicle driving instruction determination method, device, equipment and storage medium |
| CN113104010A (en) * | 2021-05-07 | 2021-07-13 | 的卢技术有限公司 | Vehicle brake control method, system, computer device and storage medium |
| CN113104010B (en) * | 2021-05-07 | 2022-04-15 | 的卢技术有限公司 | Vehicle brake control method, system, computer device and storage medium |
| US11884287B2 (en) | 2021-08-19 | 2024-01-30 | Ford Global Technologies, Llc | Enhanced vehicle operation |
| CN114237049A (en) * | 2021-12-14 | 2022-03-25 | 西安建筑科技大学 | Intelligent building system prediction control parameter setting method based on LSTM |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106990714A (en) | Adaptive Control Method and device based on deep learning | |
| Li et al. | Shared control driver assistance system based on driving intention and situation assessment | |
| CN110187639B (en) | Trajectory planning control method based on parameter decision framework | |
| Gao et al. | Personalized adaptive cruise control based on online driving style recognition technology and model predictive control | |
| Shadrin et al. | Experimental autonomous road vehicle with logical artificial intelligence | |
| CN111142379B (en) | Dynamics prediction control system and method for automatic driving vehicle | |
| Zhang et al. | Chassis coordinated control for full X-by-wire vehicles-A review | |
| Guo et al. | Design of automatic steering controller for trajectory tracking of unmanned vehicles using genetic algorithms | |
| Plöchl et al. | Driver models in automobile dynamics application | |
| Naranjo et al. | Power-steering control architecture for automatic driving | |
| CN104859650B (en) | A kind of vehicle yaw stability rolling optimization control method of Multiple Time Scales | |
| CN111434554A (en) | Controlling an autonomous vehicle based on passenger and context aware driving style profiles | |
| CN108445750A (en) | Method and system for vehicle movement planning | |
| Chiang et al. | The human-in-the-loop design approach to the longitudinal automation system for an intelligent vehicle | |
| Chiang et al. | Embedded driver-assistance system using multiple sensors for safe overtaking maneuver | |
| Rath et al. | Personalised lane keeping assist strategy: Adaptation to driving style | |
| CN113835421B (en) | Method and device for training driving behavior decision model | |
| CN114379583A (en) | Automatic driving vehicle trajectory tracking system and method based on neural network dynamics model | |
| CN113650609A (en) | Flexible transfer method and system for man-machine co-driving control power based on fuzzy rule | |
| CN114580302A (en) | Decision planning method for automatic driving automobile based on maximum entropy reinforcement learning | |
| CN110879595A (en) | Unmanned mine card tracking control system and method based on deep reinforcement learning | |
| CN110320916A (en) | Consider the autonomous driving vehicle method for planning track and system of occupant's impression | |
| CN114942642A (en) | Unmanned automobile track planning method | |
| Liu et al. | The robustly-safe automated driving system for enhanced active safety | |
| CN115593433A (en) | Remote take-over method for automatic driving vehicle |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170728 |