CN109976726A - Vehicle-mounted Edge intelligence computing architecture, method, system and storage medium - Google Patents
Vehicle-mounted Edge intelligence computing architecture, method, system and storage medium Download PDFInfo
- Publication number
- CN109976726A CN109976726A CN201910216185.2A CN201910216185A CN109976726A CN 109976726 A CN109976726 A CN 109976726A CN 201910216185 A CN201910216185 A CN 201910216185A CN 109976726 A CN109976726 A CN 109976726A
- Authority
- CN
- China
- Prior art keywords
- state
- vehicle
- data
- layer
- mounted edge
- 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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F8/00—Arrangements for software engineering
- G06F8/20—Software design
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F8/00—Arrangements for software engineering
- G06F8/70—Software maintenance or management
- G06F8/73—Program documentation
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/12—Protocols specially adapted for proprietary or special-purpose networking environments, e.g. medical networks, sensor networks, networks in vehicles or remote metering networks
Landscapes
- Engineering & Computer Science (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Library & Information Science (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses a kind of vehicle-mounted Edge intelligence computing architecture, method, system and storage medium, vehicle-mounted Edge intelligence computing architecture includes sensing layer, state layer and application layer；Sensing layer is standardized for connecting external sensible equipment, and to the polynary isomeric data that external awareness apparatus is got, and is sent to state layer after obtaining standardization multivariate data；State layer exports corresponding Obj State by state recognition model, and Obj State is sent to application layer for will standardize multivariate data input state identification model when receiving the standardization multivariate data of sensing layer transmission；Application layer, for providing application service to vehicle based on Obj State after receiving Obj State.Vehicular applications exploitation of the invention is no longer necessary solution by the hardware configuration of vehicle until entire framework, it is only necessary to multiple objects based on application layer can direct development and application, greatly reduce the difficulty of vehicular applications exploitation.
Description
Technical field
The present invention relates to automobile technical field more particularly to a kind of vehicle-mounted Edge intelligence computing architecture, method, system and deposit
Storage media.
Background technique
In recent years, the rapid development of intelligent network connection automobile, automobile have become critically important intelligent use carrier, are meeting
Except the function of the vehicles itself, automobile is melted by carrying the hardware systems such as advanced sensor, controller, actuator
Hop communication and network technology, realize V2X (vehicle to everything, automobile is to the external world) intelligent information exchange with
It is shared, the functions such as complex environment perception, intelligent decision, Collaborative Control are also equipped with, safe and efficient, comfortable driving may be implemented
Experience.The exploitation of new opplication is absorbed in more and more researchs, but the development and application on such complicated vehicle end, needs technology
Personnel have the overall process ability that upper layer application is got from bottom data, therefore, requirement of the application and development to technical staff
Height, this forms very big resistance for the application development of net connection automobile.
Summary of the invention
The present invention provides a kind of vehicle-mounted Edge intelligence computing architecture, method, system and storage medium, it is intended to for networking automobile
Application and development friendly access way is provided, reduce technical threshold.
To achieve the above object, the present invention provides a kind of vehicle-mounted Edge intelligence computing architecture, including sensing layer, state
Layer and application layer；
The sensing layer, for connecting external sensible equipment, and to the polynary isomery that the external sensible equipment is got
Data are standardized, and are sent to the state layer after obtaining standardization multivariate data；
The state layer, for when receiving the standardization multivariate data that the sensing layer is sent, by the standardization
Multivariate data input state identification model exports corresponding Obj State by the state recognition model, and by the object shape
State is sent to the application layer；
The application layer, for providing application to vehicle based on the Obj State after receiving the Obj State
Service.
Optionally, the sensing layer includes object end protocol stack and object end microprocessing unit, object end microprocessing unit
Input terminal is connected with external sensible equipment, to obtain the polynary isomeric data of the external sensible equipment；
The output end of object end microprocessing unit is connect with the input terminal of object end microprocessing unit, and the object end
Protocol stack is run in the microprocessing unit of the object end, to be split to the polynary isomeric data and feature extraction, is obtained
To standardization multivariate data；
The output end of object end microprocessing unit is connect with the state layer, to the standardization multivariate data is defeated
Out to the state layer.
Optionally, the external sensible equipment includes CAN bus, non-interference equipment, wearable device, environment information acquisition
Equipment.
Optionally, the state layer is provided with edge calculations machine module, intelligence AI chip, intelligent algorithm module and multiple marks
Quasi- object, and each standard object is provided with a variety of quantization Status Types；The intelligent algorithm module is based on the quantization
Status Type constructs the state recognition model.
Optionally, the standard object includes driver, passenger, vehicle, environment, traffic and network.
The present invention also provides a kind of vehicle-mounted Edge intelligence calculation method, the vehicle-mounted Edge intelligence calculation method includes following
Step:
When sensing layer receives the polynary isomeric data of external sensible equipment, standard is carried out to the polynary isomeric data
Change processing is sent to state layer after obtaining standardization multivariate data；
When the state layer receives the standardization multivariate data, the standardization multivariate data input state is known
Other model exports corresponding Obj State by the state recognition model, and the Obj State is sent to application layer；
After the application layer receives the Obj State, application service is provided to vehicle based on the Obj State.
Optionally, described that the polynary isomeric data is standardized, it is sent after obtaining standardization multivariate data
Include: to the step of state layer
The polynary isomeric data of the external sensible equipment is obtained by object end microprocessing unit；
The polynary isomeric data is split by object end protocol stack and feature extraction, obtains standardization multivariate data
After be sent to the state layer.
Optionally, the polynary isomeric data is driving demand force data；It is described by object end protocol stack to described polynary
Isomeric data is split and the step of feature extraction includes:
Based on the continuous data in the driving demand force data, data length normalized is carried out, multiple windows are obtained
Mouth data；
Frequency domain character extraction is carried out based on the window data, obtains the standardization multivariate data.
The present invention also provides a kind of vehicle-mounted Edge intelligence computing system, the vehicle-mounted Edge intelligence computing system includes processing
Device, the vehicle-mounted Edge intelligence calculation procedure of memory and storage in the memory, the vehicle-mounted Edge intelligence calculate journey
When sequence is run by the processor, the step of realizing vehicle-mounted Edge intelligence calculation method as described in any one of above-mentioned.
The present invention also provides a kind of computer storage medium, vehicle-mounted Edge intelligence is stored in the computer storage medium
Calculation procedure realizes the vehicle-mounted edge as described in any one of above-mentioned when the vehicle-mounted Edge intelligence calculation procedure is run by processor
The step of intelligence computation method.
Compared with prior art, the present invention provides a kind of vehicle-mounted Edge intelligence computing architecture, method, system and storage medium,
Firstly, by sensing layer, the polynary isomeric data that external sensible equipment is got is standardized；Secondly, passing through shape
State layer carries out state recognition to standardization multivariate data and obtains Obj State, in order to which relevant staff can be absorbed in algorithm
Research export to promote precision and reduce power consumption, also, since sensing layer given input is standardization multivariate data and be
Obj State avoids the research of algorithm from falling into concrete application scene；Furthermore vehicular applications exploitation no longer needs the hardware by vehicle
Structure is until entire framework, it is only necessary to multiple objects based on application layer can direct development and application, greatly reduce vehicle-mounted answer
With the difficulty of exploitation.
Detailed description of the invention
Fig. 1 is the hardware structural diagram for the vehicle-mounted Edge intelligence computing system that various embodiments of the present invention are related to；
Fig. 2 is the structural schematic diagram of the vehicle-mounted Edge intelligence computing architecture of the present invention；
Fig. 3 is the flow diagram that the vehicle-mounted Edge intelligence computing architecture of the present invention handles polynary isomeric data；
Fig. 4 is the structural schematic diagram of object end microprocessing unit in the vehicle-mounted Edge intelligence computing architecture of the present invention；
Fig. 5 is the flow diagram of the state layer modeling of the vehicle-mounted Edge intelligence computing architecture of the present invention；
Fig. 6 is the standard object structure chart of the state layer of the vehicle-mounted Edge intelligence computing architecture of the present invention；
Fig. 7 be the vehicle-mounted Edge intelligence computing architecture of the present invention identity validation by when flow diagram；
Fig. 8 is the flow diagram of the first embodiment of the vehicle-mounted Edge intelligence calculation method of the present invention；
Fig. 9 is the flow diagram of the second embodiment of the vehicle-mounted Edge intelligence calculation method of the present invention；
Figure 10 is the flow diagram of the 3rd embodiment of the vehicle-mounted Edge intelligence calculation method of the present invention；
Figure 11 is the structural schematic diagram of the 3rd embodiment of the vehicle-mounted Edge intelligence calculation method of the present invention；
Figure 12 is the data segmentation schematic diagram of the 3rd embodiment of the vehicle-mounted Edge intelligence calculation method of the present invention；
Figure 13 is the feature extraction schematic diagram of the 3rd embodiment of the vehicle-mounted Edge intelligence calculation method of the present invention.
Drawing reference numeral explanation:
| Label | Title | Label | Title |
| 10 | Sensing layer | 20 | State layer |
| 30 | Application layer | 101 | Object end protocol stack |
| 102 | Object end microprocessing unit |
The embodiments will be further described with reference to the accompanying drawings for the realization, the function and the advantages of the object of the present invention.
Specific embodiment
It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, it is not intended to limit the present invention.
As shown in Figure 1, Fig. 1 is the system structure diagram for the hardware running environment that the embodiment of the present invention is related to.
As shown in Figure 1, the system may include: processor 1001, such as CPU, network interface 1004, user interface
1003, memory 1005, communication bus 1002.Wherein, communication bus 1002 is for realizing the connection communication between these components.
User interface 1003 may include display screen (Display), input unit such as keyboard (Keyboard), optional user interface
1003 can also include standard wireline interface and wireless interface.Network interface 1004 optionally may include that the wired of standard connects
Mouth, wireless interface (such as WI-FI interface).Memory 1005 can be high speed RAM memory, be also possible to stable memory
(non-volatile memory), such as magnetic disk storage.Memory 1005 optionally can also be independently of aforementioned processor
1001 storage device.
Optionally, system can also include camera, RF (Radio Frequency, radio frequency) circuit, sensor, audio
Circuit, WiFi module etc..Certainly, system can also configure gyroscope, barometer, hygrometer, thermometer, infrared sensor etc.
Other sensors, details are not described herein.
It will be understood by those skilled in the art that the restriction of the not structure paired systems of system structure shown in Fig. 1, can wrap
It includes than illustrating more or fewer components, perhaps combines certain components or different component layouts.
As shown in Figure 1, as may include that operating system, network are logical in a kind of memory 1005 of computer storage medium
Believe module, user interface section and vehicle-mounted Edge intelligence calculation procedure.
In the system shown in figure 1, network interface 1004 is mainly used for connecting background server, carries out with background server
Data communication；User interface 1003 is mainly used for connecting client (user terminal), carries out data communication with client；And processor
1001 can be used for calling the vehicle-mounted Edge intelligence calculation procedure stored in memory 1005, and execute following operation:
When sensing layer receives the polynary isomeric data of external sensible equipment, standard is carried out to the polynary isomeric data
Change processing is sent to state layer after obtaining standardization multivariate data；
When the state layer receives the standardization multivariate data, the standardization multivariate data input state is known
Other model exports corresponding Obj State by the state recognition model, and the Obj State is sent to application layer；
After the application layer receives the Obj State, application service is provided to vehicle based on the Obj State.
Further, processor 1001 can call the vehicle-mounted Edge intelligence calculation procedure stored in memory 1005, also
Execute following operation:
The polynary isomeric data of the external sensible equipment is obtained by object end microprocessing unit；
The polynary isomeric data is split by object end protocol stack and feature extraction, obtains standardization multivariate data
After be sent to the state layer.
Further, processor 1001 can call the vehicle-mounted Edge intelligence calculation procedure stored in memory 1005, also
Execute following operation:
Based on the continuous data in the driving demand force data, data length normalized is carried out, multiple windows are obtained
Mouth data；
Frequency domain character extraction is carried out based on the window data, obtains the standardization multivariate data.
Referring to Fig. 2, Fig. 2 is the structure chart of the vehicle-mounted Edge intelligence computing architecture first embodiment of the present invention.
In the first embodiment, the vehicle-mounted Edge intelligence computing architecture includes sensing layer 10, state layer 20 and application
Layer 30；
Sensing layer 10, for connecting external sensible equipment, and the polynary isomeric data that external awareness apparatus is got into
Row standardization is sent to state layer 20 after obtaining standardization multivariate data；
In the present embodiment, more external sensible equipment can be linked into vehicle intelligent computer in sensing layer 10,
Access way can be by the wireless modes such as CAN bus, ethernet line or Wi-Fi, bluetooth, also, sensing layer 10 also provides
Polynary isomeric data can be carried out unification processing by standard data interface agreement.
Closed orthodox car in the prior art, the mode used realize vehicle by CAN bus mode for automotive interior
The transmission communication of all status datas of body and control data, generally requires to have from bottom when increasing application function on automobile
Data acquisition, and the ability of top layer application and development is arrived, the collaboration that so huge system function generally requires many people can
It completes, therefore, the exploitation threshold of vehicular applications is often too high；However, numerous applications needs with the rapid development of artificial intelligence
Want engineer that there is good development ability, it is desirable that excessively high.
By the framework of the application, sensing layer 10 is used as bottom, and sensing layer 10 can get external awareness apparatus
Polynary isomeric data be standardized, obtain standardization multivariate data after be sent to state layer 20 so that developer without
Framework bottom need to be understood, need to only deep learning be carried out to state layer 20, effectively reduce the threshold of vehicle intelligent research.
State layer 20, for multivariate data will to be standardized when receiving the standardization multivariate data of the transmission of sensing layer 10
Input state identification model exports corresponding Obj State by state recognition model, and Obj State is sent to application layer 30；
In the present embodiment, state layer 20 defines multiple standard objects, such as driver, passenger, vehicle, environment, Ke Yi great
Range includes vehicle-mounted related scope.State recognition model construction is carried out for these standard objects, is receiving sensing layer 10
When the standardization multivariate data of transmission, standardization multivariate data input state identification model is obtained by state recognition model
Corresponding Obj State.
For example, driver, that is, driver attention degree is to influence driving safety by taking the driving demand power of driver as an example
Critically important factor.The one aspect that more fatigue driving is actually influence attention degree is studied at present, and in non-fatigue
Under state, as attention degree is not high (bow and see the mobile phone, operate console and passenger's chat, making and receiving calls etc.)
In the case where cause dangerous driving even to cause accident.And the attention monitoring under these different situations is difficult to by single
Information (such as being based purely on camera) Lai Shixian.Therefore, the present invention realizes the monitoring of attention by introducing multi-source fusion.
It is specific as follows:
Polynary isomeric data is obtained by heterogeneous sensor in sensing layer 10, polynary isomeric data includes: the physiological parameter (heart
Rate HR, heart rate interphase HRI, skin pricktest GSR), driving states (speed V, throttle depth AD, brake depth B D, steering wheel angle of rotation
Spend WA, gear G), video information (front camera acquires driver head's movement and information Vi), audio-frequency information is (in collecting vehicle
Ambient sound Au), limb motion (be worn on 9 axis (acceleration A cc, gyroscope Gyr, the magnetometer Mag) motion sensor on head,
Obtain the posture and motion information on head), above-mentioned data are pre-processed by sensing layer 10, pretreatment work includes continuous
The segmentation and feature extraction of data, compared to normal driving state, abnormal driving state is unique, and developer can be with
Intend combining by the neural network model based on attention intensified learning with intelligent interaction in state layer 20, automatic study choosing
Optimal strategy is selected, learning process is taken, obtains state recognition model, therefore, is known by above-mentioned polynary isomeric data input state
When other model, polynary isomeric data is learnt by state recognition model, so that corresponding Obj State is obtained, for example,
Obj State includes attention height, higher, moderate, lower and low, therefore, in the polynary isomeric data for detecting driver
When, by polynary isomeric data input state identification model, if driver is the lower state of attention, the output note of state layer 20
The lower state of power of anticipating drives so that application layer 30 feeds back this lower state of driver attention in order to remind
Member.
Application layer 30, for providing application service to vehicle based on Obj State after receiving Obj State.
In the present embodiment, after the Obj State that application layer 30 receives the output of state layer 20, based on Obj State to vehicle
Provide application service, including developer be directed to Obj State exploration project, for example, Obj State be attention it is lower,
Then developer for attention tell somebody what one's real intentions are this state carry out exploration project, may include intelligent reminding driver, as voice is broadcast
Put prompt driver: your driving demand power is lower, please suitably reduces speed！It can also include that intelligence sends the items such as remote terminal
Mesh.
Compared with prior art, the present invention provides a kind of vehicle-mounted Edge intelligence computing architecture, method, system and storage medium,
Firstly, by sensing layer 10, the polynary isomeric data that external sensible equipment is got is standardized；Secondly, passing through
20 pairs of standardization multivariate datas of state layer carry out state recognition and obtain Obj State, in order to which relevant staff can be absorbed in
The research of algorithm, with promoted precision and reduce power consumption, also, due to sensing layer 10 given input for standardization multivariate data,
Output is Obj State, and the research of algorithm is avoided to fall into concrete application scene；Furthermore vehicular applications exploitation is no longer needed by vehicle
Hardware configuration until entire framework, it is only necessary to multiple objects based on application layer 30 can direct development and application, substantially reduce
The difficulty of vehicular applications exploitation.
Further, sensing layer 10 includes object end protocol stack 101 and object end microprocessing unit 102, object end microprocessing unit
102 input terminal is connected with external sensible equipment, to obtain the polynary isomeric data of external sensible equipment；
The output end of object end microprocessing unit 102 is connect with the input terminal of object end microprocessing unit 102, and object end protocol stack
101 run in object end microprocessing unit 102, to be split to polynary isomeric data and feature extraction, are standardized
Multivariate data；
The output end of object end microprocessing unit 102 is connect with state layer 20, is exported will standardize multivariate data to shape
State layer 20.
In the present embodiment, since the data perceived on automobile are polynary isomeries, data processing and application are being carried out
When be difficult effectively to be merged.Therefore, the invention proposes object end protocol stack 101, which is to realize polynary isomery number
According to consistency criterionization, as shown in Figure 3.So that data have unified format.A kind of feasible json format of being defined as
Key-value form, such as { " UnixTime ": " 1543813327 "；"DateType":"Audio"；"Feq":"50"；
"Dim":"1"；"Max":"89"；"Min":"12"；"Mean":"56"；"Var":"7.8"；…；"RawData":"15 89
32 46…"}.Original data are standardized by the data definition of standard, in addition, for later newly-increased data
With preferable scalability.Also, unlike the arrangement of general data format, the standardized data proposed here is not
It is only to be unified raw data format, and directly provide the feature extraction of data, such as in above-mentioned example most
Big value, minimum value, mean value, variance etc..The extraction of feature can be done directly by such mode on a sensor, be reduced
Centralized processing in layer application, and such protocol stack needs hardware device to support, therefore, it is necessary to object end of the invention is micro-
Processing unit 102.
Object end microprocessing unit 102 is present in the end of awareness apparatus, can complete the data filtering etc. to initial data
Work.The object end microprocessing unit 102 only needs to have lower processing capacity, therefore low-power consumption may be implemented.His function
Object end protocol stack 101 is mainly run, the collected initial data of external awareness apparatus is standardized, by data
The work such as pretreatment, feature extraction is completed in object end microprocessing unit 102.A kind of feasible object end microprocessing unit 102 can
To be low-power consumption object end chip, as shown in figure 4, wherein AI Core is artificial intelligence core, the algorithm mould of lightweight can be run
Type.RISC-V Core is the core module based on RISC-V instruction set, can store and run protocol instructions, I-RAM and D-
RAM is storage unit.
Further, external sensible equipment includes that CAN bus, non-interference equipment, wearable device, environment information acquisition are set
It is standby.
In the present embodiment, vehicles itself relevant device such as CAN bus and brake, throttle, steering wheel, gear and speed connects
It connects；Non-interference equipment includes the smart machines such as camera, ultrasonic sensor and microphone；Wearable device includes motion-sensing
The human bodies information acquisition apparatus such as device, skin resistance tester, heart rate sensor, electromyogram evoked potentuial measuring system；Environment information acquisition
Equipment includes environment temperature sensor, Alcohol mental disorders instrument, baroceptor, humidity sensor, air quality detector etc.
Environment information acquisition equipment.
Said external awareness apparatus is connect with object end chip, and the data combination that said external awareness apparatus is got respectively
Form polynary isomeric data.
Further, as shown in figure 5, state layer 20 is provided with edge calculations machine module, intelligence AI chip, intelligent algorithm mould
Block and multiple standard objects, and each standard object is provided with a variety of quantization Status Types；Intelligent algorithm module is based on quantization shape
State type constructs state recognition model.
Multivariate data is standardized by 101 input state layer 20 of object end protocol stack, also, by edge calculations machine module,
The cooperation of intelligent AI chip is melted using AI algorithms library, and by intelligent algorithm module progress feature extraction, feature selecting, feature
The series of computation such as conjunction construct state recognition model based on quantization Status Type, and standardization multivariate data input state is identified
After in model, exported after being handled by state recognition model standardization multivariate data.
Further, as shown in fig. 6, standard object includes driver, passenger, vehicle, environment, traffic and network.
In the present embodiment, since the application and development on the automobile of the prior art is generally required from bottom data acquisition, model
Building and the full link of upper layer application exploitation are all familiar with.But existing application and development is all relatively simple, often in initial data
On the basis of, it can be realized as applying by simple threshold decision model.Such as remind and fasten the safety belt, it is by below seat
One pressure sensor is installed, by experiencing pressure when people takes, then it is known that someone, at this time if safe
The buckle of band is not inserted into, then can remind and not fasten seat belts.But more and more applications now be cannot be by simple
Sensor or threshold triggers realize, such as the detection of fatigue driving, it is necessary to pass through camera and vehicle body state
The fusion of data is realized, and algorithm ability that the needs such as identification of facial expression are very deep, these are needed based on polynary number
Often exploitation threshold is too high for the application realized according to fusion and intelligent algorithm.
In order to solve this problem, the present invention defines shape on the sensing layer 10 of vehicle-mounted Edge intelligence computer architecture
State layer 20.Six basic objects are defined in state layer 20, comprising: 1) driver；2) passenger；3) vehicle；4) environment；5) traffic；6)
Network.Application demand under the more complete covering vehicle environment of this 6 class object, except can be expanded in special circumstances
Exhibition.And be directed to each object, then many states of the object can be defined, as shown in Figure 5.Such as this object of driver,
Driving demand power state, alcohol screen test state, fatigue state, authentication state etc. can be defined；It, can for each passenger
To define emotional state, by bus comfort level state etc.；The information such as speed, oil consumption, throttle, brake, gear are defined for vehicle；Needle
To Environment Definition temperature, humidity, illuminance, noise level etc.；Road segment classification, congestion etc. can be defined for traffic
Deng；Network type, equipment in network access situation, network delay, network security level etc. can be defined for network.
The quantizating index of different states is not quite similar, such as the state of attention of driver, can be according to 5 grades
It is other: 1- intensive concentration, 2- higher attention, 3- moderate attention, the weaker attention of 4-, 5- minuent attention；Passenger's multiplies
Vehicle comfort level can then characterize etc. according to 1 to 10 points of marking.The completion of state layer 20 is based primarily upon powerful algorithm model,
The standardization multivariate data of the given output of sensing layer 10 and the Status Type for defining object, then developer can be absorbed in building
Powerful algorithm model realizes high-precision state output.
As shown in fig. 7, when the state of state layer output passes through for identity validation, and when credit rating is excellent, then application layer
The information that output payment verification passes through.
The embodiment of the invention provides a kind of vehicle-mounted Edge intelligence calculation methods.
It is the flow diagram of the vehicle-mounted Edge intelligence calculation method first embodiment of the present invention referring to Fig. 8, Fig. 8.
The vehicle-mounted Edge intelligence calculation method the following steps are included:
Step S10, when sensing layer receives the polynary isomeric data of external sensible equipment, to the polynary isomeric data
It is standardized, is sent to state layer after obtaining standardization multivariate data；
In the present embodiment, more external sensible equipment can be linked into vehicle intelligent computer, be connect in sensing layer
Entering mode can be by the wireless mode such as CAN bus, ethernet line or Wi-Fi, bluetooth, also, sensing layer additionally provides mark
Polynary isomeric data can be carried out unification processing by quasi- data interface protocol.
Closed orthodox car in the prior art, the mode used realize vehicle by CAN bus mode for automotive interior
The transmission communication of all status datas of body and control data, generally requires to have from bottom when increasing application function on automobile
Data acquisition, and the ability of top layer application and development is arrived, the collaboration that so huge system function generally requires many people can
It completes, therefore, the exploitation threshold of vehicular applications is often too high；However, numerous applications needs with the rapid development of artificial intelligence
Want engineer that there is good development ability, it is desirable that excessively high.
By the framework of the application, sensing layer as bottom, and sensing layer external awareness apparatus can be got it is more
First isomeric data is standardized, and is sent to state layer after obtaining standardization multivariate data, so that developer is without understanding
Framework bottom only need to carry out deep learning to state layer, effectively reduce the threshold of vehicle intelligent research.
Step S20, it is when the state layer receives the standardization multivariate data, the standardization multivariate data is defeated
Enter state recognition model, corresponding Obj State is exported by the state recognition model, and the Obj State is sent to and is answered
With layer；
In the present embodiment, state layer defines multiple standard objects, such as driver, passenger, vehicle, environment, can big model
That encloses includes vehicle-mounted related scope.State recognition model construction is carried out for these standard objects, is sent receiving sensing layer
Standardization multivariate data when, standardization multivariate data input state identification model is corresponded to by state recognition model
Obj State.
For example, driver, that is, driver attention degree is to influence driving safety by taking the driving demand power of driver as an example
Critically important factor.The one aspect that more fatigue driving is actually influence attention degree is studied at present, and in non-fatigue
Under state, as attention degree is not high (bow and see the mobile phone, operate console and passenger's chat, making and receiving calls etc.)
In the case where cause dangerous driving even to cause accident.And the attention monitoring under these different situations is difficult to by single
Information (such as being based purely on camera) Lai Shixian.Therefore, the present invention realizes the monitoring of attention by introducing multi-source fusion.
It is specific as follows:
Polynary isomeric data is obtained by heterogeneous sensor in sensing layer, polynary isomeric data includes: physiological parameter (heart rate
HR, heart rate interphase HRI, skin pricktest GSR), driving states (speed V, throttle depth AD, brake depth B D, steering wheel rotational angle
WA, gear G), video information (front camera acquire driver head movement and information Vi), audio-frequency information (collecting vehicle inner ring
Border sound Au), limb motion (is worn on 9 axis (acceleration A cc, gyroscope Gyr, the magnetometer Mag) motion sensor on head, obtains
To the posture and motion information on head), above-mentioned data are pre-processed by sensing layer, pretreatment work includes continuous data
Segmentation and feature extraction, compared to normal driving state, abnormal driving state is unique, and developer can be in shape
State layer is quasi- to be combined by the neural network model based on attention intensified learning with intelligent interaction, and automatic study selection is best
Strategy takes learning process, obtains state recognition model, therefore, by above-mentioned polynary isomeric data input state identification model
When, polynary isomeric data is learnt by state recognition model, so that corresponding Obj State is obtained, for example, object shape
State includes attention height, higher, moderate, lower and low, therefore, will in the polynary isomeric data for detecting driver
Polynary isomeric data input state identification model, if driver be the lower state of attention, state layer export attention compared with
Low state, so that application layer feeds back this lower state of driver attention, in order to remind driver.
Step S30 is answered based on the Obj State to vehicle offer after the application layer receives the Obj State
With service.
In the present embodiment, after the Obj State that application layer receives state layer output, mentioned based on Obj State to vehicle
For application service, it is directed to the exploration project of Obj State including developer, for example, Obj State is that attention is lower, then opens
Hair personnel for attention tell somebody what one's real intentions are this state carry out exploration project, may include intelligent reminding driver, as voice broadcasting mention
Show driver: your driving demand power is lower, please suitably reduces speed！It can also include that intelligence sends the projects such as remote terminal.
Compared with prior art, the present invention provides a kind of vehicle-mounted Edge intelligence computing architecture, method, system and storage medium,
Firstly, by sensing layer, the polynary isomeric data that external sensible equipment is got is standardized；Secondly, passing through shape
State layer carries out state recognition to standardization multivariate data and obtains Obj State, in order to which relevant staff can be absorbed in algorithm
Research export to promote precision and reduce power consumption, also, since sensing layer given input is standardization multivariate data and be
Obj State avoids the research of algorithm from falling into concrete application scene；Furthermore vehicular applications exploitation no longer needs the hardware by vehicle
Structure is until entire framework, it is only necessary to multiple objects based on application layer can direct development and application, greatly reduce vehicle-mounted answer
With the difficulty of exploitation.
Based on first embodiment, the second embodiment of vehicle-mounted Edge intelligence calculation method of the invention is proposed, such as Fig. 9 institute
Show, step S10 includes:
Step S100 obtains the polynary isomeric data of the external sensible equipment by object end microprocessing unit；
Step S101, is split the polynary isomeric data by object end protocol stack and feature extraction, and standard is obtained
The state layer is sent to after changing multivariate data.
In the present embodiment, since the data perceived on automobile are polynary isomeries, data processing and application are being carried out
When be difficult effectively to be merged.Therefore, the invention proposes object end protocol stack, which is to realize polynary isomeric data
Consistency criterionization, so that data have unified format.A kind of feasible key-value form for being defined as json format,
Such as { " UnixTime ": " 1543813327 "；"DateType":"Audio"；"Feq":"50"；"Dim":"1"；"Max":
"89"；"Min":"12"；"Mean":"56"；"Var":"7.8"；…；"RawData":"1589 32 46…"}.Pass through standard
Data definition original data are standardized, in addition, there is preferable scalability for later newly-increased data.
Also, unlike the arrangement of general data format, the standardized data proposed here is not only raw data format
Unified, and directly provide the feature extraction of data, such as the maximum value in above-mentioned example, minimum value, mean value, side
Difference etc..The extraction of feature can be done directly by such mode on a sensor, reduce the centralized processing in upper layer application,
And such protocol stack needs hardware device to support, therefore, it is necessary to object end microprocessing units of the invention.
Object end microprocessing unit is present in the end of awareness apparatus, can complete to works such as the data filterings of initial data
Make.The object end microprocessing unit only needs to have lower processing capacity, therefore low-power consumption may be implemented.His function is mainly
Object end protocol stack is run, the collected initial data of external awareness apparatus is standardized, by the pretreatment of data, spy
The work such as sign extraction are completed in the microprocessing unit of object end.A kind of feasible object end microprocessing unit can be low-power consumption object end core
Piece, wherein AI Core is artificial intelligence core, can run the algorithm model of lightweight.RISC-V Core is referred to based on RISC-V
The core module for enabling collection, can store and run protocol instructions, I-RAM and D-RAM are storage units.
Based on second embodiment, proposes a kind of 3rd embodiment of vehicle-mounted Edge intelligence calculation method of the present invention, please refer to
Figure 10-13, the polynary isomeric data are driving demand force data；It is described by object end protocol stack to the polynary isomeric data
It is split and the step of feature extraction includes:
Step S1010 carries out data length normalized based on the continuous data in the driving demand force data,
Obtain multiple window datas；
Step S1011 carries out frequency domain character extraction based on the window data, obtains the standardization multivariate data.
In the present embodiment, multi-source perception data is obtained by heterogeneous sensor in sensing layer first, comprising:
1) physiological parameter: heart rate HR, heart rate interphase HRI, skin pricktest GSR；
2) driving states: speed V, throttle depth AD, brake depth B D, steering wheel rotational angle WA, gear G；
3) video information: front camera acquires driver head's movement and information Vi；
4) audio-frequency information: acquisition environment inside car sound Au；
5) limb motion: being worn on 9 axis (acceleration A cc, gyroscope Gyr, the magnetometer Mag) motion sensor on head,
Obtain the posture and motion information on head.
It is primarily based on object end chip and object end protocol stack pre-processes these isomeric datas, pretreated saddlebag
Include the segmentation and feature extraction of continuous data:
Data segmentation: please referring to Figure 12, for continuous data, data segmentation carried out usually in the way of sliding window,
Assuming that sliding window time window size is T, window moving step length is S, and gives the data instance number in each window and be
N, then all data need to carry out length normalization method in the way of up-sampling or down-sampling.After the normalization of example number,
Then the data in i-th of window can be expressed as datai={ HRi,HRIi,GSRi,Vi,ADi,BDi,WAi,Gi,Vii,Aui,
Acci,Gyri,Magi, whereinOther data are also similar.
Feature extraction: please referring to Figure 13, after completing data segmentation, then needs to carry out feature on raw data set to mention
It takes, feature extraction is based primarily upon experience, such as the acceleration A cc in motion sensori, gyroscope GyriAnd magnetometer
Data Magi, then the maximum value max, minimum value min, mean value mean in a window, standard deviation std, mode mode, model are extracted
Range is enclosed, the frequency domain character that the kens features such as average point number above_mean and data are extracted after FFT variation is crossed
Such as DC component DC, amplitude A, power spectral density PSD etc..In addition, video data can be extracted as image, sound according to normal size
Frequency is sound spectrograph according to then processing.
The data for completing above-mentioned sensing layer then can be in the building of sensing layer progress state recognition model.
Compared to normal driving state, abnormal driving state has it unique.The present invention is quasi- by strong based on attention
The neural network model that chemistry is practised, combines with intelligent interaction, and automatic study selection optimal strategy takes learning process.Specifically
For, characteristic pattern is obtained by networks such as CNN, k attention region is generated centered on the lime light of last time estimation, and mention
Take the feature of corresponding fixed size.By LSTM using the hidden state of feature and last iteration as input to attention region
It gives a mark, estimates next iteration point.
In order to integrate attention mechanism, need to store the history feature from depth integration network during model training
Cx, automatically generate comprising weight WxWith weighted sum Zc, calculating process is as follows:
Cx=[Zt-w+1；……；Zt-1]；
Wx=soft max (ex)；
Zc=(WxCx)T；
Wherein w indicates that number, r is arranged in weightxAnd QxRespectively indicate weight vectors and matrix based on attention, CxPass through
Depth integration network is continuously updated, and enough inputs is not included when sequence most starts to train, because the method is gone using null vector
Fill history context information matrix, ZcResult is exported as the context based on attention.
It is distributed finally, all prediction scores carry out fusion as final label.Since the model is that iteratively prediction is current
Region obtains label score, searches for the position of next suboptimum, therefore can be trained by extensive chemical acquisition ways, overcomes sample
This is insufficient to obtain problem.
Here we define ZcValue range be { 1,2,3,4,5 }, respectively correspond state of attention are as follows: height, it is higher,
Moderate, it is lower and low.
In application layer, based on the output valve of above-mentioned attention, example provides one based on driving demand power in this application
Intelligent reminding and be served by.
In addition, the embodiment of the present invention also proposes a kind of readable storage medium storing program for executing, it is stored on the readable storage medium storing program for executing vehicle-mounted
Edge intelligence calculation procedure realizes vehicle-mounted edge as described above when the vehicle-mounted Edge intelligence calculation procedure is executed by processor
The step of intelligence computation method.
Each embodiment of the specific embodiment of readable storage medium storing program for executing of the present invention and above-mentioned vehicle-mounted Edge intelligence calculation method
Essentially identical, this will not be repeated here.
It should be noted that, in this document, the terms "include", "comprise" or its any other variant are intended to non-row
His property includes, so that the process, method, article or the device that include a series of elements not only include those elements, and
And further include other elements that are not explicitly listed, or further include for this process, method, article or device institute it is intrinsic
Element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that including being somebody's turn to do
There is also other identical elements in the process, method of element, article or device.
The serial number of the above embodiments of the invention is only for description, does not represent the advantages or disadvantages of the embodiments.
Through the above description of the embodiments, those skilled in the art can be understood that above-described embodiment side
Method can be realized by means of software and necessary general hardware platform, naturally it is also possible to by hardware, but in many cases
The former is more preferably embodiment.Based on this understanding, technical solution of the present invention substantially in other words does the prior art
The part contributed out can be embodied in the form of software products, which is stored in one as described above
In storage medium (such as ROM/RAM, magnetic disk, CD), including some instructions are used so that appliance arrangement (it can be mobile phone,
Computer, server, air conditioner or network equipment etc.) execute method described in each embodiment of the present invention.
The above is only a preferred embodiment of the present invention, is not intended to limit the scope of the invention, all to utilize this hair
Equivalent structure or equivalent flow shift made by bright specification and accompanying drawing content is applied directly or indirectly in other relevant skills
Art field, is included within the scope of the present invention.
Claims (10)
1. a kind of vehicle-mounted Edge intelligence computing architecture, which is characterized in that the vehicle-mounted Edge intelligence computing architecture include sensing layer,
State layer and application layer；
The sensing layer, for connecting external sensible equipment, and to the polynary isomeric data that the external sensible equipment is got
It is standardized, is sent to the state layer after obtaining standardization multivariate data；
The state layer, it is for when receiving the standardization multivariate data that the sensing layer is sent, the standardization is polynary
Data input state identification model exports corresponding Obj State by the state recognition model, and the Obj State is sent out
It send to the application layer；
The application layer, for providing application service to vehicle based on the Obj State after receiving the Obj State.
2. vehicle-mounted Edge intelligence computing architecture according to claim 1, which is characterized in that the sensing layer includes object end association
Stack and object end microprocessing unit are discussed, the input terminal of object end microprocessing unit is connected with external sensible equipment, to obtain
State the polynary isomeric data of external sensible equipment；
The output end of object end microprocessing unit is connect with the input terminal of object end microprocessing unit, and object end agreement
Stack is run in the microprocessing unit of the object end, to be split to the polynary isomeric data and feature extraction, is marked
Standardization multivariate data；
The output end of object end microprocessing unit is connect with the state layer, to by the standardization multivariate data export to
The state layer.
3. vehicle-mounted Edge intelligence computing architecture according to any one of claim 1 to 2, which is characterized in that the outside
Awareness apparatus includes CAN bus, non-interference equipment, wearable device, environment information acquisition equipment.
4. vehicle-mounted Edge intelligence computing architecture according to claim 1, which is characterized in that the state layer is provided with edge
Computer module, intelligence AI chip, intelligent algorithm module and multiple standard objects, and there are many each standard object settings
Quantify Status Type；The intelligent algorithm module is based on the quantization Status Type and constructs the state recognition model.
5. vehicle-mounted Edge intelligence computing architecture according to claim 4, which is characterized in that the standard object includes department
Machine, passenger, vehicle, environment, traffic and network.
6. a kind of vehicle-mounted Edge intelligence calculation method, which is characterized in that the vehicle-mounted Edge intelligence calculation method includes following step
It is rapid:
When sensing layer receives the polynary isomeric data of external sensible equipment, place is standardized to the polynary isomeric data
Reason is sent to state layer after obtaining standardization multivariate data；
When the state layer receives the standardization multivariate data, the standardization multivariate data input state is identified into mould
Type exports corresponding Obj State by the state recognition model, and the Obj State is sent to application layer；
After the application layer receives the Obj State, application service is provided to vehicle based on the Obj State.
7. vehicle-mounted Edge intelligence calculation method according to claim 6, which is characterized in that described to the polynary isomery number
According to being standardized, the step of being sent to state layer, includes: after obtaining standardization multivariate data
The polynary isomeric data of the external sensible equipment is obtained by object end microprocessing unit；
The polynary isomeric data is split by object end protocol stack and feature extraction, is sent out after obtaining standardization multivariate data
It send to the state layer.
8. vehicle-mounted Edge intelligence calculation method according to claim 7, which is characterized in that the polynary isomeric data is to drive
Sail attention force data；It is described the polynary isomeric data to be split by object end protocol stack and packet the step of feature extraction
It includes:
Based on the continuous data in the driving demand force data, data length normalized is carried out, multiple window numbers are obtained
According to；
Frequency domain character extraction is carried out based on the window data, obtains the standardization multivariate data.
9. a kind of vehicle-mounted Edge intelligence computing system, which is characterized in that the vehicle-mounted Edge intelligence computing system includes processor,
The vehicle-mounted Edge intelligence calculation procedure of memory and storage in the memory, the vehicle-mounted Edge intelligence calculation procedure quilt
When the processor is run, the step of realizing vehicle-mounted Edge intelligence calculation method as described in any one of claim 6-8.
10. a kind of computer storage medium, which is characterized in that be stored with vehicle-mounted Edge intelligence meter in the computer storage medium
Program is calculated, the vehicle as described in any one of claim 6-8 is realized when the vehicle-mounted Edge intelligence calculation procedure is run by processor
The step of carrying Edge intelligence calculation method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910216185.2A CN109976726A (en) | 2019-03-20 | 2019-03-20 | Vehicle-mounted Edge intelligence computing architecture, method, system and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910216185.2A CN109976726A (en) | 2019-03-20 | 2019-03-20 | Vehicle-mounted Edge intelligence computing architecture, method, system and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109976726A true CN109976726A (en) | 2019-07-05 |
Family
ID=67079948
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910216185.2A Pending CN109976726A (en) | 2019-03-20 | 2019-03-20 | Vehicle-mounted Edge intelligence computing architecture, method, system and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109976726A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111414884A (en) * | 2020-03-27 | 2020-07-14 | 南京工业大学 | Facial expression recognition method based on edge calculation |
| CN113491520A (en) * | 2020-04-07 | 2021-10-12 | 广州汽车集团股份有限公司 | Driving fatigue detection method and device |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1034984A2 (en) * | 1999-03-12 | 2000-09-13 | Navigation Technologies Corporation | Method and system for an in-vehicle computing architecture |
| CN105025077A (en) * | 2015-05-28 | 2015-11-04 | 广州番禺职业技术学院 | Vehicular Internet of Things operation system based on cloud computing |
| CN108551491A (en) * | 2018-05-24 | 2018-09-18 | 广西电网有限责任公司 | The mobile Internet of Things cloud system of heterogeneous network spatial multi time based on artificial intelligence |
| CN109109863A (en) * | 2018-07-28 | 2019-01-01 | 华为技术有限公司 | Smart machine and its control method, device |
-
2019
- 2019-03-20 CN CN201910216185.2A patent/CN109976726A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1034984A2 (en) * | 1999-03-12 | 2000-09-13 | Navigation Technologies Corporation | Method and system for an in-vehicle computing architecture |
| CN105025077A (en) * | 2015-05-28 | 2015-11-04 | 广州番禺职业技术学院 | Vehicular Internet of Things operation system based on cloud computing |
| CN108551491A (en) * | 2018-05-24 | 2018-09-18 | 广西电网有限责任公司 | The mobile Internet of Things cloud system of heterogeneous network spatial multi time based on artificial intelligence |
| CN109109863A (en) * | 2018-07-28 | 2019-01-01 | 华为技术有限公司 | Smart machine and its control method, device |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111414884A (en) * | 2020-03-27 | 2020-07-14 | 南京工业大学 | Facial expression recognition method based on edge calculation |
| CN113491520A (en) * | 2020-04-07 | 2021-10-12 | 广州汽车集团股份有限公司 | Driving fatigue detection method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20200057487A1 (en) | Methods and systems for using artificial intelligence to evaluate, correct, and monitor user attentiveness | |
| US11155141B2 (en) | Facilitating personalized vehicle occupant comfort | |
| El Mekki et al. | Improving driver identification for the next-generation of in-vehicle software systems | |
| CN108688676A (en) | Vehicle drive support system and vehicle drive support method | |
| CN108688677A (en) | Vehicle drive support system and vehicle drive support method | |
| CN105189241B (en) | Assess and manage the emotional state of vehicle operators | |
| US20190225232A1 (en) | Passenger Experience and Biometric Monitoring in an Autonomous Vehicle | |
| US20170151930A1 (en) | Vehicle with wearable for identifying one or more vehicle occupants | |
| CN109131167A (en) | Method for controlling a vehicle and device | |
| CN108327722A (en) | System and method for identifying vehicle driver by Move Mode | |
| CN109416733A (en) | Portable personalization | |
| KR20180040092A (en) | Mobile sensor platform | |
| CN107531236A (en) | Wagon control based on occupant | |
| CN110517687A (en) | The system for controlling its function using the voice command outside automotive | |
| JP2022552860A (en) | Passenger Condition Adjustment System for Passenger Cars Based on Predictive and Anticipatory Control | |
| CN108688673A (en) | Vehicle drive support system | |
| CN107521485A (en) | Driving behavior analysis based on vehicle braking | |
| CN110371132A (en) | Driver's adapter tube appraisal procedure and device | |
| US11605379B2 (en) | Artificial intelligence server | |
| CN110450783A (en) | For running the control unit and method of autonomous vehicle | |
| CN108688675A (en) | Vehicle drive support system | |
| CN110458604A (en) | A kind of net about driver's evaluation method, device, equipment and storage medium | |
| CN108688593A (en) | System and method for identifying at least one passenger of vehicle by Move Mode | |
| Lozoya-Santos et al. | Survey on biometry for cognitive automotive systems | |
| CN109976726A (en) | Vehicle-mounted Edge intelligence computing architecture, method, system and storage medium |
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 | ||
| CB02 | Change of applicant information |
Address after: 518000 706, building 3b, hongrongyuan Shangjun phase II, Longping community, Dalang street, Longhua District, Shenzhen City, Guangdong Province Applicant after: Jintu computing technology (Shenzhen) Co.,Ltd. Address before: 518000 area a, 18 / F, building 1, creative park, broadcasting group, No.1 Qingqing Road, Qinghu community, Longhua sub district office, Longhua District, Shenzhen City, Guangdong Province Applicant before: SHENZHEN SEMISKY TECHNOLOGY Co.,Ltd. |
|
| CB02 | Change of applicant information | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190705 |
|
| RJ01 | Rejection of invention patent application after publication |