CN116451576A - Unmanned plane cluster dynamic collaborative reasoning system and method based on size model switching - Google Patents
Unmanned plane cluster dynamic collaborative reasoning system and method based on size model switching Download PDFInfo
- Publication number
- CN116451576A CN116451576A CN202310377115.1A CN202310377115A CN116451576A CN 116451576 A CN116451576 A CN 116451576A CN 202310377115 A CN202310377115 A CN 202310377115A CN 116451576 A CN116451576 A CN 116451576A
- Authority
- CN
- China
- Prior art keywords
- unmanned aerial
- aerial vehicle
- reasoning
- model
- information
- 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
- 238000000034 method Methods 0.000 title claims abstract description 35
- 230000007613 environmental effect Effects 0.000 claims abstract description 13
- 230000008447 perception Effects 0.000 claims abstract description 7
- 238000004891 communication Methods 0.000 claims description 24
- 238000010801 machine learning Methods 0.000 claims description 14
- 238000004364 calculation method Methods 0.000 claims description 12
- 238000003062 neural network model Methods 0.000 claims description 6
- 238000011161 development Methods 0.000 claims description 4
- 230000018109 developmental process Effects 0.000 claims description 4
- 230000011218 segmentation Effects 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/04—Inference or reasoning models
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Geometry (AREA)
- Computer Hardware Design (AREA)
- Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)
Abstract
The invention discloses a unmanned aerial vehicle cluster dynamic collaborative reasoning system and method based on size model switching, wherein the unmanned aerial vehicle cluster consists of a plurality of heterogeneous unmanned aerial vehicles with computing capability, and each unmanned aerial vehicle comprises: the system comprises a large model, a small model, an environment sensing module and an information feedback module which are deployed on each unmanned aerial vehicle, wherein the environment sensing module is connected with the information feedback module, the information feedback module is connected with the unmanned aerial vehicle, and the unmanned aerial vehicle is respectively connected with the large model and the small model; the environment sensing module is used for sensing environment information; the information feedback module is used for receiving the environment information perceived by the environment perception module and communicating with the unmanned aerial vehicle and transmitting data; the large model is used for carrying out unmanned aerial vehicle collaborative execution reasoning tasks according to the environmental information received by the information feedback module; the small model is used for carrying out the unmanned plane single machine execution reasoning task according to the environment information received by the information feedback module, and the reasoning task is better completed through the switching of the size model.
Description
Technical Field
The invention belongs to the technical field of unmanned aerial vehicle cluster collaborative reasoning, and particularly relates to an unmanned aerial vehicle cluster dynamic collaborative reasoning system and method based on size model switching.
Background
In recent years, with the development of artificial intelligence technology, deep neural networks are being widely used in various fields including voice processing, image recognition, and natural language processing, and have achieved a surprisingly good expression. However, with the complexity of deep neural networks, the computation and storage space required for the model is much greater, thus requiring more resources for both network training and reasoning. Since computing and storage resources of the edge devices are often limited, most artificial intelligence applications on mobile devices typically send data to a cloud server, which performs model training and reasoning tasks, and returns the results to the edge devices. However, the conventional cloud computing mode has the problems of insufficient real-time performance, insufficient bandwidth, high energy consumption and poor privacy.
Along with the improvement of computing capacity and storage space of the edge equipment and the development of model compression and optimization technology, the reasoning task of the deep learning model can be moved from the cloud server to the edge equipment to be realized, so that the problem of a cloud computing mode is effectively relieved. The existing collaborative reasoning method generally only carries out model cutting on a high-precision model and deploys a sub-model on edge equipment to carry out reasoning, however, the reasoning effect is poor when the environment where the unmanned aerial vehicle is located is interfered by the reasoning mode, so that practical application is affected, and robustness is lacking.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides the unmanned plane cluster dynamic collaborative reasoning system and the unmanned plane cluster dynamic collaborative reasoning method based on the size model switching, which select large model collaborative reasoning or small model single machine reasoning according to the current environment information of the unmanned plane at the beginning of reasoning, thereby improving the applicability of model reasoning to the reasoning environment and having practical significance and good application prospect.
In order to achieve the above purpose, the invention adopts the following technical scheme: unmanned aerial vehicle cluster developments collaborative reasoning system based on size model switches, unmanned aerial vehicle cluster comprises many heterogeneous unmanned aerial vehicles that have computational power, and every unmanned aerial vehicle includes: the system comprises a large model, a small model, an environment sensing module and an information feedback module which are deployed on each unmanned aerial vehicle, wherein the environment sensing module is connected with the information feedback module, the information feedback module is connected with the unmanned aerial vehicle, and the unmanned aerial vehicle is respectively connected with the large model and the small model;
the environment sensing module is used for sensing environment information;
the information feedback module is used for receiving the environment information perceived by the environment perception module and communicating with the unmanned aerial vehicle and transmitting data;
the large model is used for carrying out unmanned aerial vehicle collaborative execution reasoning tasks according to the environmental information received by the information feedback module, and is a high-precision complex machine learning model;
the small model is used for performing unmanned aerial vehicle single machine execution reasoning tasks according to the environmental information received by the information feedback module, and is a light machine learning model with higher precision.
Further, the environmental information perceived by the environmental perception module includes: the computational power and communication quality of current unmanned aerial vehicles.
Further, the high-precision complex machine learning model is a deep neural network model with a parameter quantity of more than or equal to 10M.
Further, the higher-precision lightweight machine learning model is a deep neural network model with a parameter amount of less than 10M.
Further, the invention also provides an unmanned aerial vehicle cluster dynamic collaborative reasoning method of the unmanned aerial vehicle cluster dynamic collaborative reasoning system based on size model switching, which specifically comprises the following steps:
step S1, in a cluster of N unmanned aerial vehicles, an unmanned aerial vehicle i receives an inference task demand, carries out self information sensing through an environment sensing module on the unmanned aerial vehicle i, and feeds back self information of the unmanned aerial vehicle i and the inference task demand to the rest N-1 unmanned aerial vehicles in the unmanned aerial vehicle cluster through an information feedback module; the self information of the unmanned aerial vehicle i comprises: calculation capability f of unmanned aerial vehicle i i Communication quality C of unmanned aerial vehicle i and remaining N-1 unmanned aerial vehicles i,j ,(j≠i)；
S2, after the rest N-1 unmanned aerial vehicles in the unmanned aerial vehicle cluster receive self information and reasoning task requirements of the unmanned aerial vehicle i, sensing the self current environment information through respective environment sensing modules, and feeding the self current environment information back to the unmanned aerial vehicle i through an information feedback module; each unmanned aerial vehicle in the N-1 unmanned aerial vehicle has own current environmental information comprising: computing power f of unmanned aerial vehicle a a ,a∈[1,N]A not equal i, communication quality C between unmanned aerial vehicle a and unmanned aerial vehicle b a,b ,a,b∈
[1,N],a≠b≠i；
And S3, the unmanned aerial vehicle i judges the received current environment information of the N-1 unmanned aerial vehicle according to the reasoning task requirement and feeds back the current environment information to the large model or the small model to execute the reasoning task.
Further, in step S3, if the reasoning task demand has the minimum delay requirement, the specific process of step S3 is as follows:
step S31, estimating optimal time delay L of unmanned aerial vehicle collaborative reasoning according to own current environment information of N-1 unmanned aerial vehicles c Z E (1, N) of the Z-frame unmanned aerial vehicle needing to carry out reasoning tasks;
step S32, estimating the time delay L of the unmanned aerial vehicle i for single machine reasoning according to the reasoning task requirement s ；
Step S33, if L c ≤L s The large model corresponding to the reasoning task is subjected to model segmentation and is distributed to unmanned aerial vehicles with different Z frames, and the Z frames of unmanned aerial vehicles enter the model segmentationPerforming collaborative reasoning; if L c >L s And carrying out small-model single-machine reasoning by the unmanned aerial vehicle i.
Further, in step S3, if the accuracy of the task demand is a, the specific process of step S3 is as follows: small model precision to be deployed A s Comparing with the accuracy A of the reasoning task demand, if A s If not less than A, carrying out small-model single-machine reasoning by the unmanned aerial vehicle i, and if A s <A, under the condition that the reasoning task demand precision A is met, estimating the optimal time delay L of unmanned aerial vehicle collaborative reasoning according to the current environment information of the N-1 unmanned aerial vehicle c And the Z-frame unmanned aerial vehicle which needs to carry out reasoning tasks carries out collaborative reasoning by the Z-frame unmanned aerial vehicle, wherein Z is E (1, N).
Further, the optimal time delay L of the unmanned plane collaborative reasoning c The calculation process of (2) is as follows: combining N-1 unmanned aerial vehicles, acquiring reasoning time delay according to the computing capacity of each unmanned aerial vehicle, and acquiring communication quality C between the N-1 unmanned aerial vehicles a,b ,a,b∈[1,N]The communication delay is obtained by a not equal to b not equal to i, and the minimum sum of the reasoning delay and the communication delay is taken as the optimal delay L c 。
Further, the communication quality C between the unmanned aerial vehicle i and the rest N-1 unmanned aerial vehicles i,j The calculation process of (j+.i) is:
wherein B is i,j For the bandwidth between unmanned aerial vehicle i and unmanned aerial vehicle j, S i,j For the average power, M, of the signals transmitted by the channel between unmanned aerial vehicle i and unmanned aerial vehicle j i,j Is the gaussian noise power inside the channel between drone i and drone j.
Further, the communication quality C between the unmanned aerial vehicle a and the unmanned aerial vehicle b a,b The calculation process of (1) is as follows:
wherein B is a,b For the bandwidth between unmanned aerial vehicle a and unmanned aerial vehicle b, S a,b Is the average power of the signals transmitted in the channel between the unmanned aerial vehicle a and the unmanned aerial vehicle b, M a,b Is the gaussian noise power inside the channel between drone a and drone b.
Compared with the prior art, the invention has the beneficial effects that: the invention is different from the conventional complex model collaborative reasoning or a single machine execution light model reasoning method, the large model and the small model are simultaneously arranged on the unmanned aerial vehicle, the environment information received by the information feedback module and the reasoning task demand are judged, and the environment information and the reasoning task demand are fed back to the large model or the small model to execute the reasoning task, so that the dynamic switching of the large model and the small model is realized, and the proper model is flexibly selected according to the actual situation, so that the reasoning task can be better completed. The unmanned aerial vehicle cluster dynamic collaborative reasoning method based on size model switching can improve the reliability and the robustness of an unmanned aerial vehicle cluster dynamic collaborative reasoning system.
Drawings
FIG. 1 is a flow chart of unmanned aerial vehicle carrying information and method of the unmanned aerial vehicle cluster dynamic collaborative reasoning method based on size model switching;
FIG. 2 is a scene diagram of acquiring unmanned plane cluster environment information by the unmanned plane i;
FIG. 3 is a scene diagram of unmanned aerial vehicle cluster collaborative reasoning and unmanned aerial vehicle i single machine reasoning under the condition of considering time delay;
fig. 4 is a scene diagram of unmanned aerial vehicle cluster collaborative reasoning and unmanned aerial vehicle i single machine reasoning under the condition of considering accuracy.
Detailed Description
The technical scheme of the invention is further explained below with reference to the accompanying drawings.
As shown in fig. 1, the invention provides an unmanned aerial vehicle cluster dynamic collaborative reasoning system based on size model switching, wherein an unmanned aerial vehicle cluster is composed of a plurality of heterogeneous unmanned aerial vehicles with computing capability, and each unmanned aerial vehicle comprises: the system comprises a large model, a small model, an environment sensing module and an information feedback module which are deployed on each unmanned aerial vehicle, wherein the environment sensing module is connected with the information feedback module, the information feedback module is connected with the unmanned aerial vehicle, and the unmanned aerial vehicle is respectively connected with the large model and the small model;
the environment sensing module is used for sensing environment information, and comprises: the computing power and communication quality of the current unmanned aerial vehicle;
the information feedback module is used for receiving the environment information perceived by the environment perception module and communicating with the unmanned aerial vehicle and transmitting data;
the large model is used for carrying out unmanned aerial vehicle collaborative execution reasoning tasks according to the environmental information received by the information feedback module, the large model is a high-precision complex machine learning model, and the high-precision complex machine learning model is a deep neural network model with parameter quantity more than or equal to 10M, such as: VGG, resNet.
The small model is used for performing unmanned aerial vehicle single machine execution reasoning tasks according to the environmental information received by the information feedback module, the small model is a light machine learning model with higher precision, and the light machine learning model with higher precision is a deep neural network model with parameter less than 10M, such as: squeezeNet, mobileNet, shuffleNet.
The invention also provides an unmanned aerial vehicle cluster dynamic collaborative reasoning method based on the unmanned aerial vehicle cluster dynamic collaborative reasoning system switched by the size model, which comprises the following steps:
step S1, as shown in FIG. 2, in a cluster of N unmanned aerial vehicles, an unmanned aerial vehicle i receives an inference task demand, carries out self information perception through an environment perception module on the unmanned aerial vehicle i, and feeds back self information of the unmanned aerial vehicle i and the inference task demand to the rest N-1 unmanned aerial vehicles in the unmanned aerial vehicle cluster through an information feedback module; the self information of the unmanned aerial vehicle i in the invention comprises: calculation capability f of unmanned aerial vehicle i i Communication quality C of unmanned aerial vehicle i and remaining N-1 unmanned aerial vehicles i,j ,(j≠i)；
Wherein, the communication quality C between the unmanned aerial vehicle i and the rest N-1 unmanned aerial vehicles i,j The calculation process of (j+.i) is:
wherein B is i,j For the bandwidth between unmanned aerial vehicle i and unmanned aerial vehicle j, S i,j For the average power, M, of the signals transmitted by the channel between unmanned aerial vehicle i and unmanned aerial vehicle j i,j Is the gaussian noise power inside the channel between drone i and drone j.
S2, after the rest N-1 unmanned aerial vehicles in the unmanned aerial vehicle cluster receive self information and reasoning task requirements of the unmanned aerial vehicle i, sensing the self current environment information through respective environment sensing modules, and feeding the self current environment information back to the unmanned aerial vehicle i through an information feedback module; the current environment information of each unmanned aerial vehicle in the N-1 unmanned aerial vehicles comprises the following information: computing power f of unmanned aerial vehicle a a ,a∈[1,N]A not equal i, communication quality C between unmanned aerial vehicle a and unmanned aerial vehicle b a,b ,a,b∈
[1,N],a≠b≠i；
Wherein, communication quality C between unmanned aerial vehicle a and unmanned aerial vehicle b a,b The calculation process of (1) is as follows:
wherein B is a,b For the bandwidth between unmanned aerial vehicle a and unmanned aerial vehicle b, S a,b Is the average power of the signals transmitted in the channel between the unmanned aerial vehicle a and the unmanned aerial vehicle b, M a,b Is the gaussian noise power inside the channel between drone a and drone b.
And S3, the unmanned aerial vehicle i judges the received current environment information of the N-1 unmanned aerial vehicle according to the reasoning task requirement, feeds back the current environment information to the large model or the small model to execute the reasoning task, and can select a proper model to execute the reasoning task according to the current environment information of the unmanned aerial vehicle cluster and the reasoning task requirement, so that resources can be saved or higher reasoning precision can be obtained while the reasoning task requirement is met. Specifically:
in step S3, if the reasoning task requirement has the minimum delay requirement, as shown in fig. 3, the specific process of step S3 is as follows:
s31, estimating the unmanned aerial vehicle coordination according to the current environment information of the N-1 unmanned aerial vehicleOptimal time delay L by the same reasoning c Z E (1, N) is the unmanned aerial vehicle which needs to carry out the reasoning task, wherein the unmanned aerial vehicle i is the unmanned aerial vehicle which receives the reasoning task, so the Z unmanned aerial vehicle comprises the unmanned aerial vehicle i;
step S32, estimating the time delay L of the unmanned aerial vehicle i for single machine reasoning according to the reasoning task requirement s ；
Step S33, if L c ≤L s Carrying out model segmentation on a large model corresponding to an reasoning task, distributing the large model to different unmanned aerial vehicles of a Z frame, and carrying out collaborative reasoning by the Z frame unmanned aerial vehicles; if L c >L s The unmanned plane i performs small-model single-machine reasoning, and by the method, if large-model collaborative reasoning is selected, higher reasoning precision can be ensured while the time delay requirement is met, and if small-model single-machine reasoning is selected, resources can be saved while the time delay requirement is met.
In step S3, if the accuracy of the task demand is a, as shown in fig. 4, the specific process in step S3 is as follows: small model precision to be deployed A s Comparing with the accuracy A of the reasoning task demand, if A s If not less than A, carrying out small-model single-machine reasoning by the unmanned aerial vehicle i, and if A s <A, under the condition that the reasoning task demand precision A is met, estimating the optimal time delay L of unmanned aerial vehicle collaborative reasoning according to the current environment information of the N-1 unmanned aerial vehicle c And the Z-frame unmanned aerial vehicle which needs to carry out the reasoning task carries out cooperative reasoning by the Z-frame unmanned aerial vehicle, and a subset of unmanned aerial vehicles with the lowest reasoning time delay can be selected under the condition of meeting the precision requirement, so that the best cooperative reasoning effect is achieved, namely, the precision requirement is met and the time delay is lower.
Optimal time delay L of unmanned plane collaborative reasoning in the invention c The calculation process of (2) is as follows: combining N-1 unmanned aerial vehicles, acquiring reasoning time delay according to the computing capacity of each unmanned aerial vehicle, and acquiring communication quality C between the N-1 unmanned aerial vehicles a,b ,a,b∈[1,N]The communication delay is obtained by a not equal to b not equal to i, and the minimum sum of the reasoning delay and the communication delay is taken as the optimal delay L c . And the unmanned aerial vehicle combination with the optimal delay and the unmanned aerial vehicle i are jointly used as a Z-frame unmanned aerial vehicle for completing the collaborative reasoning task.
According to the unmanned plane cluster dynamic collaborative reasoning method based on size model switching, a high-precision complex machine learning model and a high-precision light machine learning model can be dynamically switched, so that the unmanned plane cluster dynamic collaborative reasoning environment can be better adapted to a reasoning environment and a reasoning task can be performed. The invention can also select an reasoning mode according to the requirements of the application on time delay and precision, so that the unmanned aerial vehicle cluster dynamically selects and executes a large model or a small model according to the application requirements to obtain a reasoning result meeting the requirements.
The above is only a preferred embodiment of the present invention, and the scope of the present invention is not limited to the above embodiment, and all technical solutions belonging to the concept of the present invention are within the scope of the present invention. It should be noted that modifications and adaptations to the invention without departing from the principles thereof are intended to be within the scope of the invention as set forth in the following claims.
Claims (10)
1. Unmanned aerial vehicle cluster developments collaborative reasoning system based on size model switches, its characterized in that, unmanned aerial vehicle cluster comprises many heterogeneous unmanned aerial vehicles that have computational capability, and every unmanned aerial vehicle includes: the system comprises a large model, a small model, an environment sensing module and an information feedback module which are deployed on each unmanned aerial vehicle, wherein the environment sensing module is connected with the information feedback module, the information feedback module is connected with the unmanned aerial vehicle, and the unmanned aerial vehicle is respectively connected with the large model and the small model;
the environment sensing module is used for sensing environment information;
the information feedback module is used for receiving the environment information perceived by the environment perception module and communicating with the unmanned aerial vehicle and transmitting data;
the large model is used for carrying out unmanned aerial vehicle collaborative execution reasoning tasks according to the environmental information received by the information feedback module, and is a high-precision complex machine learning model;
the small model is used for performing unmanned aerial vehicle single machine execution reasoning tasks according to the environmental information received by the information feedback module, and is a light machine learning model with higher precision.
2. The unmanned aerial vehicle cluster dynamic collaborative reasoning system based on size model switching of claim 1, wherein the environmental information perceived by the environmental awareness module comprises: the computational power and communication quality of current unmanned aerial vehicles.
3. The unmanned aerial vehicle cluster dynamic collaborative reasoning system based on size model switching of claim 1, wherein the high-precision complex machine learning model is a deep neural network model with a parameter quantity of more than or equal to 10M.
4. The unmanned aerial vehicle cluster dynamic collaborative reasoning system based on size model switching according to claim 1, wherein the high-precision lightweight machine learning model is a deep neural network model with a parameter quantity of less than 10M.
5. The unmanned aerial vehicle cluster dynamic collaborative reasoning method of the unmanned aerial vehicle cluster dynamic collaborative reasoning system based on size model switching of any one of claims 1-4, which is characterized by comprising the following steps:
step S1, in a cluster of N unmanned aerial vehicles, an unmanned aerial vehicle i receives an inference task demand, carries out self information sensing through an environment sensing module on the unmanned aerial vehicle i, and feeds back self information of the unmanned aerial vehicle i and the inference task demand to the rest N-1 unmanned aerial vehicles in the unmanned aerial vehicle cluster through an information feedback module; the self information of the unmanned aerial vehicle i comprises: calculation capability f of unmanned aerial vehicle i i Communication quality C of unmanned aerial vehicle i and remaining N-1 unmanned aerial vehicles i，j ，(j≠i)；
S2, after the rest N-1 unmanned aerial vehicles in the unmanned aerial vehicle cluster receive self information and reasoning task requirements of the unmanned aerial vehicle i, sensing the self current environment information through respective environment sensing modules, and feeding the self current environment information back to the unmanned aerial vehicle i through an information feedback module; each unmanned aerial vehicle in the N-1 unmanned aerial vehicle has own current environmental information comprising:computing power f of unmanned aerial vehicle a a ，a∈[1，N]A not equal i, communication quality C between unmanned aerial vehicle a and unmanned aerial vehicle b a，b ，a，b∈[1，N]，a≠b≠i；
And S3, the unmanned aerial vehicle i judges the received current environment information of the N-1 unmanned aerial vehicle according to the reasoning task requirement and feeds back the current environment information to the large model or the small model to execute the reasoning task.
6. The unmanned aerial vehicle cluster dynamic collaborative reasoning method of the unmanned aerial vehicle cluster dynamic collaborative reasoning system based on size model switching according to claim 5, wherein if the reasoning task demand has the minimum delay requirement in step S3, the specific process of step S3 is as follows:
step S31, estimating optimal time delay L of unmanned aerial vehicle collaborative reasoning according to own current environment information of N-1 unmanned aerial vehicles c Z E (1, N) of the Z-frame unmanned aerial vehicle needing to carry out reasoning tasks;
step S32, estimating the time delay L of the unmanned aerial vehicle i for single machine reasoning according to the reasoning task requirement s ；
Step S33, if L c ≤L s Carrying out model segmentation on a large model corresponding to an reasoning task, distributing the large model to different unmanned aerial vehicles of a Z frame, and carrying out collaborative reasoning by the Z frame unmanned aerial vehicles; if L c >L s And carrying out small-model single-machine reasoning by the unmanned aerial vehicle i.
7. The unmanned aerial vehicle cluster dynamic collaborative reasoning method of the unmanned aerial vehicle cluster dynamic collaborative reasoning system based on size model switching according to claim 5, wherein in step S3, if the precision of the reasoning task demand is a, the specific process of step S3 is: small model precision to be deployed A s Comparing with the accuracy A of the reasoning task demand, if A s If not less than A, carrying out small-model single-machine reasoning by the unmanned aerial vehicle i, and if A s <A, under the condition that the reasoning task demand precision A is met, estimating the optimal time delay L of unmanned aerial vehicle collaborative reasoning according to the current environment information of the N-1 unmanned aerial vehicle c And a Z-frame unmanned aerial vehicle required to perform reasoning tasks, the Z-frame unmanned aerial vehicle being used forAnd carrying out collaborative reasoning, wherein Z is E (1, N).
8. The unmanned aerial vehicle cluster dynamic collaborative reasoning method of the unmanned aerial vehicle cluster dynamic collaborative reasoning system based on size model switching according to claim 6 or 7, wherein the optimal time delay L of unmanned aerial vehicle collaborative reasoning is c The calculation process of (2) is as follows: combining N-1 unmanned aerial vehicles, acquiring reasoning time delay according to the computing capacity of each unmanned aerial vehicle, and acquiring communication quality C between the N-1 unmanned aerial vehicles a，b ，a，b∈[1，N]The communication delay is obtained by a not equal to b not equal to i, and the minimum sum of the reasoning delay and the communication delay is taken as the optimal delay L c 。
9. The unmanned aerial vehicle cluster dynamic collaborative reasoning method based on the unmanned aerial vehicle cluster dynamic collaborative reasoning system switched by the size model according to claim 5, wherein the communication quality Ci of the unmanned aerial vehicle i and the rest of N-1 unmanned aerial vehicles is that ，j， The calculation process of (j+.i) is:
wherein B is i，j For the bandwidth between unmanned aerial vehicle i and unmanned aerial vehicle j, S i，j For the average power, M, of the signals transmitted by the channel between unmanned aerial vehicle i and unmanned aerial vehicle j i，j Is the gaussian noise power inside the channel between drone i and drone j.
10. The unmanned aerial vehicle cluster dynamic collaborative reasoning method based on the unmanned aerial vehicle cluster dynamic collaborative reasoning system switched by the size model according to claim 5, wherein the communication quality C between the unmanned aerial vehicle a and the unmanned aerial vehicle b is that a，b The calculation process of (1) is as follows:
wherein B is a，b For the bandwidth between unmanned aerial vehicle a and unmanned aerial vehicle b, S a，b Is the average power of the signals transmitted in the channel between the unmanned aerial vehicle a and the unmanned aerial vehicle b, M a，b Is the gaussian noise power inside the channel between drone a and drone b.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310377115.1A CN116451576A (en) | 2023-04-11 | 2023-04-11 | Unmanned plane cluster dynamic collaborative reasoning system and method based on size model switching |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310377115.1A CN116451576A (en) | 2023-04-11 | 2023-04-11 | Unmanned plane cluster dynamic collaborative reasoning system and method based on size model switching |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116451576A true CN116451576A (en) | 2023-07-18 |
Family
ID=87127132
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310377115.1A Pending CN116451576A (en) | 2023-04-11 | 2023-04-11 | Unmanned plane cluster dynamic collaborative reasoning system and method based on size model switching |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116451576A (en) |
-
2023
- 2023-04-11 CN CN202310377115.1A patent/CN116451576A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Hanyao et al. | Edge-assisted online on-device object detection for real-time video analytics | |
| CN112348172B (en) | Deep neural network collaborative reasoning method based on end edge cloud architecture | |
| CN110839184B (en) | Method and device for adjusting bandwidth of mobile fronthaul optical network based on flow prediction | |
| CN110941667A (en) | Method and system for calculating and unloading in mobile edge calculation network | |
| CN110837842A (en) | Video quality evaluation method, model training method and model training device | |
| CN112616126B (en) | Sensor information processing method based on distributed cooperation | |
| CN110968426A (en) | Edge cloud collaborative k-means clustering model optimization method based on online learning | |
| CN111915142A (en) | Unmanned aerial vehicle auxiliary resource allocation method based on deep reinforcement learning | |
| CN113282352A (en) | Energy-saving unloading method based on multi-unmanned aerial vehicle cooperative auxiliary edge calculation | |
| WO2023020502A1 (en) | Data processing method and apparatus | |
| Chien et al. | A lightweight model with spatial–temporal correlation for cellular traffic prediction in Internet of Things | |
| CN115665869A (en) | Multi-user collaboration platform and method based on edge calculation and directed acyclic graph | |
| CN116451576A (en) | Unmanned plane cluster dynamic collaborative reasoning system and method based on size model switching | |
| CN117114113A (en) | Collaborative reasoning acceleration method based on queuing theory | |
| CN110647396A (en) | Method for realizing intelligent application of end cloud cooperative low-power consumption and limited bandwidth | |
| An et al. | Air-ground integrated mobile edge computing in vehicular visual sensor networks | |
| CN114327689B (en) | Strategy scheduling method and system for complex edge computing environment | |
| CN115150288B (en) | Distributed communication system and method | |
| CN116419325A (en) | Task unloading, resource allocation and track planning method and system for collaborative calculation of multiple unmanned aerial vehicles | |
| CN112738225B (en) | Edge calculation method based on artificial intelligence | |
| CN116017570A (en) | Edge computing system resource management method based on block chain | |
| Martini et al. | Challenges and perspectives in neuromorphic-based visual IoT systems and networks | |
| CN112363844B (en) | Convolutional neural network vertical segmentation method for image processing | |
| CN115080219A (en) | Data processing method, electronic device and computer program product | |
| CN117939635B (en) | Power distribution method suitable for industrial Internet of things communication learning performance balance |
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 |