CN115507852A - Multi-unmanned aerial vehicle path planning method based on block chain and attention-enhancing learning - Google Patents
Multi-unmanned aerial vehicle path planning method based on block chain and attention-enhancing learning Download PDFInfo
- Publication number
- CN115507852A CN115507852A CN202211086751.0A CN202211086751A CN115507852A CN 115507852 A CN115507852 A CN 115507852A CN 202211086751 A CN202211086751 A CN 202211086751A CN 115507852 A CN115507852 A CN 115507852A
- Authority
- CN
- China
- Prior art keywords
- unmanned aerial
- aerial vehicle
- model
- path
- block chain
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000013439 planning Methods 0.000 title claims abstract description 56
- 238000000034 method Methods 0.000 title claims description 46
- 230000007246 mechanism Effects 0.000 claims abstract description 19
- 230000006870 function Effects 0.000 claims description 31
- 238000012549 training Methods 0.000 claims description 19
- 230000004083 survival effect Effects 0.000 claims description 16
- 230000008569 process Effects 0.000 claims description 13
- 238000005457 optimization Methods 0.000 claims description 8
- 238000013507 mapping Methods 0.000 claims description 6
- 238000012544 monitoring process Methods 0.000 claims description 6
- 238000004088 simulation Methods 0.000 claims description 3
- 230000001186 cumulative effect Effects 0.000 claims description 2
- 238000004422 calculation algorithm Methods 0.000 abstract description 14
- 238000010586 diagram Methods 0.000 description 8
- 238000004891 communication Methods 0.000 description 6
- 230000006399 behavior Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000000873 masking effect Effects 0.000 description 2
- 238000012876 topography Methods 0.000 description 2
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000010485 coping Effects 0.000 description 1
- 238000013523 data management Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 230000006386 memory function Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000002787 reinforcement Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000002922 simulated annealing Methods 0.000 description 1
- 238000005728 strengthening Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 239000013598 vector Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
- G01C21/20—Instruments for performing navigational calculations
-
- 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)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Automation & Control Theory (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Traffic Control Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to the field of unmanned aerial vehicle path planning, and discloses a multi-unmanned aerial vehicle path planning algorithm based on a block chain and enhanced attention. Meanwhile, a safety mechanism of a block chain is introduced, so that the information safety of the multi-computer system is improved, some malicious information attacks are prevented, and meanwhile, the safety and the stability of the system are greatly improved by adding an asymmetric key mechanism. Meanwhile, a block chain knowledge sharing mechanism is used, and a public key and a private key are issued to achieve the function of sharing the multi-unmanned aerial vehicle path planning knowledge. The generalization ability of many unmanned aerial vehicle path planning can be improved.
Description
Technical Field
The invention relates to the field of unmanned aerial vehicle path planning, in particular to a multi-unmanned aerial vehicle path planning method based on a block chain and attention-enhancing learning.
Background
The application of the unmanned aerial vehicle brings brand-new development paths for a plurality of industries, the unmanned aerial vehicle has excellent application prospect, the safety risk coefficient is extremely low in the using process, the consumed resources are less in operation, and tasks can be completed quickly and quickly. When the unmanned aerial vehicle moves, how to plan the path is the core content of the application of the unmanned aerial vehicle, and generally, the path planning of the unmanned aerial vehicle is the use of reducing resources and time to the maximum while reaching the place. Under the condition of an obstacle, the unmanned aerial vehicle finishes the movement to a target point by simulating the behavior of human avoiding the obstacle, and for the condition that a plurality of target points exist, people convert the behavior into a traveler problem to search, and provide an algorithm based on a self-organizing mapping network. The collaborative path planning method for multiple drones can be used as an NP combinatorial optimization problem with multiple constraints. At present, the main classical models of the multi-unmanned aerial vehicle cooperative reconnaissance problem comprise a multi-traveler model, a mixed linear integer programming model and a vehicle scheduling and path programming model. Conventional solving methods solve through heuristic algorithms, such as genetic algorithms, simulated annealing algorithms, and evolutionary algorithms. However, the traditional model cannot fully describe the constraints of the multi-drone cooperative reconnaissance mission. Under complex environmental conditions, in the face of constantly changing terrain data, the heuristic algorithm needs to be re-optimized and solved, so that the adaptability is poor, and a corresponding solution cannot be quickly given. Most of the traditional heuristic learning can not avoid the problem that iteration needs to be optimized again, and the instantaneity is not high. And some important flight route information is often related to in traditional multi-unmanned aerial vehicle path planning, the risk that the unmanned aerial vehicle is attacked by cyber physics in the coming years is increased, and traditional unmanned aerial vehicle crowd is easily attacked by other malicious unmanned aerial vehicles, so that some important path information is lost, the multi-unmanned aerial vehicle path planning method is damaged, and other problems occur in the current multi-unmanned aerial vehicle application, and some aerial collisions often occur due to the coordination of the unmanned aerial vehicles or the communication of the multi-unmanned aerial vehicles is hijacked. In addition, most of the existing systems only rely on local communication between adjacent member unmanned aerial vehicles in member unmanned aerial vehicles of an unmanned aerial vehicle network at present, and lack knowledge sharing capable of trusting unmanned aerial vehicles globally, so that a multi-unmanned aerial vehicle path planning method based on block chains and strengthened attention learning is provided.
Disclosure of Invention
Technical problem to be solved
Aiming at the defects of the prior art, the invention provides a multi-unmanned aerial vehicle path planning method based on a block chain and strengthened attention learning, and solves the problems.
(II) technical scheme
In order to achieve the above purpose, the invention provides the following technical scheme: a multi-unmanned aerial vehicle path planning method based on block chains and enhanced attention learning comprises the following steps:
the first step is as follows: using decision variablesDecision variables representing cooperative reconnaissance tasks of multiple unmanned aerial vehicles;
the second step is that: different proportionality coefficients are used for combining each objective function in the process that the unmanned aerial vehicle executes the reconnaissance mission;
the third step: introducing a node variance to adjust the load of each drone;
the fourth step: establishing an end-to-end training model of an attention mechanism to plan the path of the unmanned aerial vehicle;
the fifth step: training a model;
and a sixth step: a multi-unmanned aerial vehicle system carrying a block chain;
the seventh step: and multi-unmanned aerial vehicle cross-domain knowledge sharing and common learning.
Preferably, the objective function includes: access by each drone to a node;
the unmanned aerial vehicle leaves the access of the node;
the distance constraint of each drone cannot exceed its maximum distance;
total flight mileage of all unmanned aerial vehicles;
survival coverage function of drone.
Preferably, the survival coverage function of the whole path of the multi-drone scout mission is an accumulated sum of the survival coverage functions of the nodes.
Preferably, the path planning in the fourth step is obtained by the following method:
s1: mapping coordinates of the nodes to a 128-dimensional embedding layer by using different parameters, and distinguishing the base and the nodes by using a linear mapping layer of parameter sums;
s2: the embedding of the input node is updated through attention layers, each attention layer consists of two sublayers, a multi-head attention layer and a node full-connection feedforward layer are connected, and all sublayers are connected by adopting a jump connection structure;
s3: calculating graph embedding by calculating an average value of network output layer embedding, and then transmitting the output layer embedding and the graph embedding to a decoder;
s4: the decoder generates a path plan by receiving the embedding, the graph embedding and the context embedding of the encoder output layer nodes.
Preferably, the decoding process is performed step by step, and at each time point t, the decoder generates context embedding based on the input nodes and the generated path plan at the time point t-1.
Preferably, the method for training the model in the fifth step includes the following steps:
the first step is as follows: defining a model and defining an optimization objective;
the second step: generating simulation data, then performing pre-training, judging whether the standard is valid, and if the standard is invalid, performing the pre-training again, and effectively extracting the model;
the third step: inputting terrain data, solving a model, and training the model;
the fourth step: and obtaining the maximum optimal solution of the path, judging whether the effectiveness in the model is improved or not, if not, ending, and if so, effectively extracting the model again.
Preferably, the blockchain multi-drone system includes the following: each unmanned aerial vehicle has a public key and a private key in a multi-unmanned aerial vehicle system, when the unmanned aerial vehicle needs to send information, the public key is needed to encrypt the information, and then other unmanned aerial vehicles need to obtain the information and need to decrypt the encrypted information by using the private keys of the other unmanned aerial vehicles.
Preferably, the cross-domain knowledge sharing includes the following:
the first step is as follows: finding out a newest block in the channel A, wherein the newest block stores the global model, and initializing parameters of the model;
the second step: the regional multi-unmanned aerial vehicle path planning system needs to update model parameters, firstly a channel application needs to be sent to a blockchain network, then the blockchain network returns a channel list to the regional multi-unmanned aerial vehicle path planning system, a monitoring system needs to register on a corresponding channel, then the regional multi-unmanned aerial vehicle path planning system obtains a public key and a private key, and the regional multi-unmanned aerial vehicle path planning system sends the model parameters and needs to use the public key to encrypt the parameters.
The third step: updating global model parameters of each channel, and downloading the latest block model of the channel by the local monitoring system to update the model parameters;
the fourth step: when the model loss function in the channel tends to be stable, updating the global model state Trie, and generating blocks;
the fifth step: the latest block is written in each channel, and the Trie is updated.
(III) advantageous effects
Compared with the prior art, the invention provides a multi-unmanned aerial vehicle path planning method based on a block chain and strengthened attention learning, which has the following beneficial effects:
1. this many unmanned aerial vehicle route planning method based on block chain and intensive attention study, this patent are in order to solve the learning method that exists in the many unmanned aerial vehicle route planning of tradition and mostly can't avoid optimizing iterative problem again, and adaptability is relatively poor, can't give corresponding solution fast. By using a novel end-to-end model enhanced attention mechanics training method in deep learning, the automatic learning capability of path planning of multiple unmanned aerial vehicles can be enhanced aiming at terrain data which changes continuously.
2. According to the multi-unmanned-plane path planning method based on the block chain and the attention-enhancing learning, the block chain is adopted for data storage and data management. Meanwhile, an asymmetric encryption mechanism in the block chain is introduced to provide safe one-to-one communication in the unmanned aerial vehicle network, and a knowledge sharing mechanism in the block chain is used to share knowledge in data collected by a plurality of unmanned aerial vehicles, so that the problem that the unmanned aerial vehicles only communicate locally can be well solved.
3. According to the multi-unmanned aerial vehicle path planning method based on the block chain and the attention-enhanced learning, the sequence model of the attention mechanism is added, so that the learning capacity of the unmanned aerial vehicle for coping with the data of the changed terrain can be enhanced, and the changeable terrain can be coped with more effectively. Meanwhile, a safety mechanism of a block chain is introduced, so that the information safety of a multi-computer system is improved, malicious information attack is prevented, and meanwhile, an asymmetric key mechanism is added, so that the safety and the stability of the system are greatly improved. Meanwhile, a block chain knowledge sharing mechanism is used, and the function of sharing the path planning knowledge of the multiple unmanned aerial vehicles is achieved by issuing a public key and a private key. The generalization ability of many unmanned aerial vehicle route planning can be improved.
Drawings
Fig. 1 is a schematic diagram of a multi-drone path planning model;
FIG. 2 is a schematic diagram of an encoder structure;
FIG. 3 is a schematic diagram of a decoder network;
fig. 4 is a schematic diagram of a multi-drone path training process;
fig. 5 is a schematic diagram of a multi-drone system based on a blockchain;
FIG. 6 is a block chain based asymmetric encryption information exchange algorithm;
fig. 7 is a schematic diagram of a multi-drone blockchain knowledge sharing model;
fig. 8 is a timing diagram illustrating knowledge sharing of a blockchain.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1 to 8, a method for planning paths of multiple drones based on block chains and enhanced attention learning includes the following steps:
for the multi-unmanned aerial vehicle cooperative reconnaissance mission, the threat and relevant position information of a target area are roughly evaluated before investigation, and the multi-unmanned aerial vehicle needs to reconnaissance on a plurality of target areas of an affected area in detail. Suppose N V Unmanned aerial vehicle needs detailed detection with multiple N T The target area of (1). At present, the cruising distance of the unmanned aerial vehicle and the threat degree of a target are comprehensively considered, and a model is established.
The patent describes the path planning problem of multi-unmanned aerial vehicle cooperation as the base, wherein the coordinates of the position where the unmanned aerial vehicle leaves and the position returned after completing the reconnaissance task are (x) 0 ,y 0 ) (ii) a Unmanned aerial vehicle assuming N exists v Unmanned plane, through set V =To show that the maximum cruising distance of the unmanned aerial vehicle is L. Detecting a target: suppose there is N T Objects to be detected, using setsIndicating that the set T = {0} < u > T is used 0 Where 0 represents the initial position and landing position of the drone and the set T is used to represent nodes used locally for detection. The coordinate of any one target node i in the set T is (x) i ,y i ) D when the object i is detected,survival probability of unmanned aerial vehicle is P i ∈[0.7,0.9](ii) a Set of paths, using pi v And { (i, j) | i, j ∈ T } represents the unmanned aerial vehicle reconnaissance path, wherein v represents the basic principle of safe flight.
Using decision variablesDecision variables representing a multi-drone cooperative reconnaissance mission:
the cruising distance, the survival probability and the execution capacity of the unmanned aerial vehicle are considered. This patent uses a relatively simple linear combination method, and this method uses different proportionality coefficients to combine each objective function in the unmanned aerial vehicle carries out the reconnaissance task, and the objective function is as follows:
min L(π)=a*f 1 +b*f 2 +c*f 3 ,0≤a,b,c≤a+b+c=1 (2)
wherein f is 1 ,f 2 And f and 3 the method is characterized in that three optimization objective functions in an unmanned aerial vehicle reconnaissance model are adopted, and a, b and c respectively correspond to loss function proportionality coefficients of the three objective functions.
Where equation (3) represents one visit by each drone to the node and equation (4) represents one visit by a drone to leave the node. When multiple drones are flying at high altitude for a long time, many limiting factors can occur. For example, if the cruising distance is too long, the mission may directly fail, first due to the limited power of the on-board battery. It may also cause communication failures due to interference from various external uncertainties, resulting in the inability to transmit valid data to the ground station. Therefore, to reduce the flight risk, the shorter the total flight distance of the drone, the better. Equation (5) indicates that the distance constraint of each drone cannot exceed its maximum distance.
Equation (6) describes the total flight range of all drones, which is one of the optimization objectives of the reconnaissance mission
Since the threat level of each node to be detected to the drone is not consistent, the drone must select the safest flight path when detecting multiple nodes, and equation (7) defines the survival coverage function of the drone
Wherein a is v Is the starting node of the unmanned aerial vehicle v reconnaissance path, b v Is the end node, P i Is the survival probability of drone v when detecting node i. Assuming that drones pass through nodes 1, 2 and 3, their survival probabilities are 0.9, 0.8 and 0.7, the survival coverage function of the drone at node 1 is 0.9, the survival coverage function of node 2 is 0.9 x 0.8=0.72, node 3 is 0.9 x 0.8 x 0.7=0.5, and the survival coverage function of the entire path of the multi-drone scout task is the cumulative sum of the survival coverage functions of the nodes, as shown in equation (8)
Equation (9) represents the survival coverage function f 2 The optimization is to take the maximum value.
In the process of executing the actual task, in order to avoid the overload of the whole unmanned aerial vehicle, the load of each unmanned aerial vehicle is adjusted by introducing the node variance, as shown in formula (10).
min f 3 =var(π v ) (10)
This patent establishes an end-to-end training model based on attention mechanism and solves many unmanned aerial vehicle path planning problem.
Whereinπ t E T pi is a path plan given by constraints on the local terrain, s is an example of the current plan. In the end-to-end model, an encoder processes encoding of an input coordinate node, a decoder receives the input of the encoder, and the decoding process is constrained by a masking mechanism to give a reasonable path plan pi.
The encoder structure that this patent adopted is shown in fig. 2:
mapping the coordinates of node e to the 128-dimensional embedding layer using different parameters, by using parameter w x And b x To distinguish between bases and nodes.
The embedding of the input nodes is updated by N layers of attention layers, each consisting of two sublayers: the multi-head attention layer and the nodes are all connected with the feedforward layer. The sub-layers are connected by adopting a hop connection structure, so that the outstanding problem that the learning efficiency is reduced along with the increase of the number of network layers is solved. The calculation formulas are shown in formula (13) and formula (14).
WhereinDefining as node embedding, wherein l belongs to { 1.,. N }; calculating image embedding by calculating the average value of the network output layer embedding, and transmitting the output layer embedding and the image embedding to a decoder;
the decoder generates a path plan pi by receiving the embedding, graph embedding and context embedding of the encoder output layer nodes, the structure is shown in fig. 3:
the decoding process is performed step by step, and as shown in fig. 3, at each time point t, the decoder generates context embedding based on the input nodes and the generated path plan at the time point t-1:
whereinIs the current cruising distance of the drone, whereinIs represented at pi t The distance, the symbol, that the time performs this task.]Describing a horizontal join operator, this means to join three vectors horizontally into one and then use a multi-headed attention tier, where q is (c) ByCalculation of k i And v i ByComputing
Similarity u (c)i Can be calculated by the above formula, wherein d k To set the similarity distance to 16 while masking nodes that do not satisfy the constraint and setting compatibility to- ∞, a single attention tier is used at the last tier of the network in order to derive the policy in equation (11)
And finally, obtaining the access probability of each node, and carrying out probability sampling on the current best-fit node through the following equation in the decoding process.
The model is pre-trained, and the strengthening algorithm used in the patent is a strategy gradient algorithm represented by the following formula:
π θ =P(a|s,θ)≈π(a|s) (19)
to find the optimal strategy, a gradient formula is obtained by differentiating the objective function
WhereinIs a scoring function, Q π And (s, a) are operation state variable values. After the model is constructed, a loss function equation (21) is defined and trained. And modifying the target function of reinforcement learning to ensure that L (pi) achieves the purpose of optimizing planning:
in the enhancement algorithm, the base line b(s) can reduce the variance and improve the training speed of the model. In this patent, the baseline of the enhancement algorithm is M ← β M + (1- β) L (pi). The training process is shown in fig. 4:
multi-drone path training first defines a model and an optimization objective in a first step, and then trains it using a large amount of simulation data so that the model can mine the connection between the data and the optimization path. If the criteria are valid, the model can be extracted and its parameters fixed, then the real data of the terrain can be input to solve the unmanned aerial vehicle reconnaissance problem in real time. Meanwhile, an online training mechanism is introduced into the model, and the model is trained by using real data of the terrain, so that the effect is continuously improved, and the self-evolution of the model is realized.
Many unmanned aerial vehicle systems based on block chain
Some air collisions can occur in the current multi-unmanned aerial vehicle application, which is often caused by problems occurring in the coordination of unmanned aerial vehicles or hijacking of multi-unmanned aerial vehicle communication. Moreover, most current existing systems rely on local communication between adjacent member drones at the member drones of the drone network, and lack the global knowledge sharing that can trust drones.
To this end, this patent proposes a multi-drone system based on block chain storage, as shown in fig. 5:
the system uses the block chain as the unit for storing the information of unmanned aerial vehicle path planning, and simultaneously uses the asymmetric encryption method in the block chain as the encryption algorithm for unmanned aerial vehicle information exchange, as shown in fig. 6:
each unmanned aerial vehicle has a public key and a private key in a multi-unmanned aerial vehicle system, when the unmanned aerial vehicle needs to send information, the information needs to be encrypted by using the public key, and then other unmanned aerial vehicles need to obtain the information and need to decrypt the encrypted information by using the private keys of the other unmanned aerial vehicles. Such encryption mechanism can effectively prevent malicious unmanned aerial vehicle to obtain the information, has improved the security performance of whole unmanned aerial vehicle system, can prevent that unmanned aerial vehicle path planning information from being maliciously stolen, uses the memory function of block chain with, can effectively trace to the source to the information.
Meanwhile, a block chain knowledge sharing learning method is established, and the problem that the traditional multi-unmanned aerial vehicle lacks the functions of cross-domain knowledge sharing and common learning is solved.
The process of the multi-unmanned aerial vehicle path planning model for knowledge sharing is as follows:
the first step is as follows: and finding the latest block in the channel A, wherein the latest block stores the global model and initializes the parameters of the model.
The second step is that: the regional multi-unmanned aerial vehicle path planning system needs to update model parameters, firstly needs to send a channel application to the blockchain network, and then the blockchain network returns a channel list to the regional multi-unmanned aerial vehicle path planning system. The monitoring system needs to register on a corresponding channel, then the regional multi-unmanned-aerial-vehicle path planning system obtains a public key and a private key, and the regional multi-unmanned-aerial-vehicle path planning system sends model parameters and needs to use public key encryption parameters for the parameters.
The third step: and each channel updates the global model parameters, and the local monitoring system downloads the latest block model of the channel to update the model parameters.
The fourth step: and when the model loss function in the channel tends to be stable, updating the global model state Trie and generating the block.
The fifth step: the latest block is written in each channel, and the Trie is updated. The specific timing diagram is shown in fig. 8:
the patent provides a many unmanned aerial vehicle path planning algorithm based on block chain and intensive attention, through the sequence model who adds the attention mechanism, can strengthen the learning ability that unmanned aerial vehicle reply changes the topography data, can effectively reply changeable topography more. Meanwhile, a safety mechanism of a block chain is introduced, so that the information safety of the multi-computer system is improved, some malicious information attacks are prevented, and meanwhile, the safety and the stability of the system are greatly improved by adding an asymmetric key mechanism. Meanwhile, a block chain knowledge sharing mechanism is used, and the function of sharing the path planning knowledge of the multiple unmanned aerial vehicles is achieved by issuing a public key and a private key. The generalization ability of many unmanned aerial vehicle route planning can be improved.
Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (8)
1. A multi-unmanned aerial vehicle path planning method based on a block chain and attention-enhancing learning is characterized by comprising the following steps:
the first step is as follows: using decision variablesA decision variable representing a multi-drone cooperative scout mission;
the second step: different proportionality coefficients are used for combining each objective function in the process of executing the reconnaissance task by the unmanned aerial vehicle;
the third step: introducing a node variance to adjust the load of each unmanned aerial vehicle;
the fourth step: establishing an end-to-end training model of an attention mechanism to plan the path of the unmanned aerial vehicle;
the fifth step: training a model;
and a sixth step: a multi-unmanned aerial vehicle system carrying a block chain;
the seventh step: and sharing cross-domain knowledge of multiple unmanned aerial vehicles and learning together.
2. The method for planning the path of multiple unmanned aerial vehicles based on the block chain and the enhanced attention learning of claim 1, characterized in that: the objective function includes:
access by each drone to a node;
the unmanned aerial vehicle leaves the access of the node;
the distance constraint of each drone cannot exceed its maximum distance;
the total flight mileage of all unmanned aerial vehicles;
survival coverage function of drone.
3. The method for planning the path of the multiple unmanned aerial vehicles based on the block chain and the enhanced attention learning of claim 2, wherein: the survival coverage function of the whole path of the multi-unmanned aerial vehicle reconnaissance task is the cumulative sum of the survival coverage functions of the nodes.
4. The method for planning the path of multiple unmanned aerial vehicles based on the block chain and the enhanced attention learning of claim 1, characterized in that: the path planning in the fourth step is obtained by the following method:
s1: mapping coordinates of the nodes to a 128-dimensional embedding layer by using different parameters, and distinguishing bases and nodes by using a linear mapping layer of parameter sums;
s2: the embedding of the input node is updated through attention layers, each attention layer consists of two sublayers, a multi-head attention layer and a node full-connection feedforward layer are connected, and all sublayers are connected by adopting a jump connection structure;
s3: calculating graph embedding by calculating an average value of network output layer embedding, and then transmitting the output layer embedding and the graph embedding to a decoder;
s4: the decoder generates a path plan by receiving the embedding, the graph embedding and the context embedding of the encoder output layer nodes.
5. The method for planning the path of the multiple unmanned aerial vehicles based on the block chain and the enhanced attention learning of claim 1, wherein: the decoding process is carried out step by step, and at each time point t, the decoder generates context embedding based on the input nodes and the generated path plan at the time point t-1.
6. The method for planning the path of the multiple unmanned aerial vehicles based on the block chain and the enhanced attention learning of claim 1, wherein: the method for training the model in the fifth step comprises the following steps:
the first step is as follows: defining a model and defining an optimization objective;
the second step is that: generating simulation data, then performing pre-training, judging whether the standard is valid, and if the standard is invalid, performing the pre-training again, and effectively extracting the model;
the third step: inputting terrain data, solving a model, and training the model;
the fourth step: and obtaining the maximum optimal solution of the path, judging whether the effectiveness in the model is improved or not, if not, ending, and if so, effectively extracting the model again.
7. The method for planning the path of the multiple unmanned aerial vehicles based on the block chain and the enhanced attention learning of claim 1, wherein: the multi-drone system of the blockchain comprises the following: each unmanned aerial vehicle has a public key and a private key in a multi-unmanned aerial vehicle system, when the unmanned aerial vehicle needs to send information, the information needs to be encrypted by using the public key, and then other unmanned aerial vehicles need to obtain the information and need to decrypt the encrypted information by using the private keys of the other unmanned aerial vehicles.
8. The method for planning the path of multiple unmanned aerial vehicles based on the block chain and the enhanced attention learning of claim 7, wherein: cross-domain knowledge sharing includes the following:
the first step is as follows: finding out a newest block in the channel A, wherein the newest block stores the global model, and initializing parameters of the model;
the second step is that: the regional multi-unmanned aerial vehicle path planning system needs to update model parameters, firstly needs to send a channel application to a blockchain network, then the blockchain network returns a channel list to the regional multi-unmanned aerial vehicle path planning system, a monitoring system needs to register on a corresponding channel, then the regional multi-unmanned aerial vehicle path planning system obtains a public key and a private key, and the regional multi-unmanned aerial vehicle path planning system sends the model parameters and needs to use public key encryption parameters for the parameters.
The third step: updating global model parameters of each channel, and downloading the latest block model of the channel by the local monitoring system to update the model parameters;
the fourth step: when the model loss function in the channel tends to be stable, updating the global model state Trie, and generating blocks;
the fifth step: the latest block is written in each channel, and the Trie is updated.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211086751.0A CN115507852B (en) | 2022-09-07 | 2022-09-07 | Multi-unmanned aerial vehicle path planning method based on blockchain and enhanced attention learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211086751.0A CN115507852B (en) | 2022-09-07 | 2022-09-07 | Multi-unmanned aerial vehicle path planning method based on blockchain and enhanced attention learning |
Publications (2)
Publication Number | Publication Date |
---|---|
CN115507852A true CN115507852A (en) | 2022-12-23 |
CN115507852B CN115507852B (en) | 2023-11-03 |
Family
ID=84503288
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211086751.0A Active CN115507852B (en) | 2022-09-07 | 2022-09-07 | Multi-unmanned aerial vehicle path planning method based on blockchain and enhanced attention learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115507852B (en) |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112256056A (en) * | 2020-10-19 | 2021-01-22 | 中山大学 | Unmanned aerial vehicle control method and system based on multi-agent deep reinforcement learning |
US20210200212A1 (en) * | 2019-12-31 | 2021-07-01 | Uatc, Llc | Jointly Learnable Behavior and Trajectory Planning for Autonomous Vehicles |
CN113095439A (en) * | 2021-04-30 | 2021-07-09 | 东南大学 | Heterogeneous graph embedding learning method based on attention mechanism |
CN113703482A (en) * | 2021-08-30 | 2021-11-26 | 西安电子科技大学 | Task planning method based on simplified attention network in large-scale unmanned aerial vehicle cluster |
CN114462664A (en) * | 2021-12-09 | 2022-05-10 | 武汉长江通信智联技术有限公司 | Short-range branch flight scheduling method integrating deep reinforcement learning and genetic algorithm |
CN114565053A (en) * | 2022-03-10 | 2022-05-31 | 天津大学 | Deep heterogeneous map embedding model based on feature fusion |
CN114578860A (en) * | 2022-03-28 | 2022-06-03 | 中国人民解放军国防科技大学 | Large-scale unmanned aerial vehicle cluster flight method based on deep reinforcement learning |
-
2022
- 2022-09-07 CN CN202211086751.0A patent/CN115507852B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20210200212A1 (en) * | 2019-12-31 | 2021-07-01 | Uatc, Llc | Jointly Learnable Behavior and Trajectory Planning for Autonomous Vehicles |
CN112256056A (en) * | 2020-10-19 | 2021-01-22 | 中山大学 | Unmanned aerial vehicle control method and system based on multi-agent deep reinforcement learning |
CN113095439A (en) * | 2021-04-30 | 2021-07-09 | 东南大学 | Heterogeneous graph embedding learning method based on attention mechanism |
CN113703482A (en) * | 2021-08-30 | 2021-11-26 | 西安电子科技大学 | Task planning method based on simplified attention network in large-scale unmanned aerial vehicle cluster |
CN114462664A (en) * | 2021-12-09 | 2022-05-10 | 武汉长江通信智联技术有限公司 | Short-range branch flight scheduling method integrating deep reinforcement learning and genetic algorithm |
CN114565053A (en) * | 2022-03-10 | 2022-05-31 | 天津大学 | Deep heterogeneous map embedding model based on feature fusion |
CN114578860A (en) * | 2022-03-28 | 2022-06-03 | 中国人民解放军国防科技大学 | Large-scale unmanned aerial vehicle cluster flight method based on deep reinforcement learning |
Non-Patent Citations (4)
Title |
---|
ABEGAZ MOHAMMED SEID等: "Blockchain-Enabled Task Offloading With Energy Harvesting in Multi-UAV-Assisted IoT Networks:A Multi-Agent DRL Approach", IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS * |
PRABHAT KUMAR等: "Blockchain and Deep Learning Empowered Secure Data Sharing Framework for Softwarized UAVs", ICC WORKSHOPS * |
李鸿一等: "基于随机采样的高层消防无人机协同搜索规划", 中国科学, vol. 52, no. 9 * |
罗傲等: "基于强化学习的一类具有输入约束非线性系统最优控制", 控制理论与应用 * |
Also Published As
Publication number | Publication date |
---|---|
CN115507852B (en) | 2023-11-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Wang et al. | Learning in the air: Secure federated learning for UAV-assisted crowdsensing | |
Jin et al. | Coride: joint order dispatching and fleet management for multi-scale ride-hailing platforms | |
Yuan et al. | Evolutionary multitasking in permutation-based combinatorial optimization problems: Realization with TSP, QAP, LOP, and JSP | |
Huang et al. | Resilient routing mechanism for wireless sensor networks with deep learning link reliability prediction | |
Pan et al. | How You Act Tells a Lot: Privacy-Leaking Attack on Deep Reinforcement Learning. | |
CN112329348A (en) | Intelligent decision-making method for military countermeasure game under incomplete information condition | |
CN113472419B (en) | Safe transmission method and system based on space-based reconfigurable intelligent surface | |
Kang et al. | Securing data sharing from the sky: Integrating blockchains into drones in 5G and beyond | |
Yan et al. | Optimal routes and aborting strategies of trucks and drones under random attacks | |
Sun et al. | A cooperative target search method based on intelligent water drops algorithm | |
Gao et al. | Multi-UAV task allocation based on improved algorithm of multi-objective particle swarm optimization | |
CN116861239A (en) | Federal learning method and system | |
Ahmed et al. | 5G-empowered drone networks in federated and deep reinforcement learning environments | |
Xu et al. | Security and privacy in artificial intelligence-enabled 6g | |
Peng et al. | Modeling and solving the dynamic task allocation problem of heterogeneous UAV swarm in unknown environment | |
Liu et al. | A hybrid mobile robot path planning scheme based on modified gray wolf optimization and situation assessment | |
Zhang et al. | Backtracking search algorithm with dynamic population for energy consumption problem of a UAV-assisted IoT data collection system | |
CN115507852A (en) | Multi-unmanned aerial vehicle path planning method based on block chain and attention-enhancing learning | |
Benfriha et al. | Insiders detection in the uncertain IoD using fuzzy logic | |
Wang et al. | A Vertical Heterogeneous Network (VHetNet)–Enabled Asynchronous Federated Learning-Based Anomaly Detection Framework for Ubiquitous IoT | |
CN115119215B (en) | Optimal path repairing method for fence coverage holes in natural protected area | |
Lv et al. | Guest Editorial Introduction to the Special Issue on Internet of Things in Intelligent Transportation Infrastructure | |
An et al. | Robust Topology Generation of Internet of Things Based on PPO Algorithm Using Discrete Action Space | |
Beghriche | An adaptive secure and efficient bio-inspired routing protocol for effective cooperation in FANETs | |
CN112215414B (en) | Multi-machine collaborative route planning method and system based on similarity model |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |