CN105844068A - Distribution method oriented to simulation Q learning attack targets - Google Patents
Distribution method oriented to simulation Q learning attack targets Download PDFInfo
- Publication number
- CN105844068A CN105844068A CN201610427869.3A CN201610427869A CN105844068A CN 105844068 A CN105844068 A CN 105844068A CN 201610427869 A CN201610427869 A CN 201610427869A CN 105844068 A CN105844068 A CN 105844068A
- Authority
- CN
- China
- Prior art keywords
- red
- state
- aircraft
- learning
- action
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/30—Circuit design
- G06F30/36—Circuit design at the analogue level
- G06F30/367—Design verification, e.g. using simulation, simulation program with integrated circuit emphasis [SPICE], direct methods or relaxation methods
Landscapes
- Engineering & Computer Science (AREA)
- Computer Hardware Design (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Microelectronics & Electronic Packaging (AREA)
- Evolutionary Computation (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Feedback Control In General (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a distribution method oriented to simulation Q learning attack targets. The method comprises the following steps: (1) determining an initial state, acquiring two-side air combat situation information of a red side and a blue side, wherein the two-side air combat situation information comprises the number of formation aircrafts and relevant parameters of formation aircrafts, and providing input for target distribution and air combat model calculation of the red side; (2) determining an action set performed by the formation of the red side, strictly stipulating complete state-action pairs, determining a proper probability epsilon value and selecting actions of the red side by using an epsilon-greedy strategy; and (3) stipulating a Q learning algorithm reward function, an ending state and a convergence condition and distributing iteration to attack targets of the red side by using a Q learning algorithm until the convergence condition is met. The method gets rid of the dependence on the priori knowledge; due to the introduction of the epsilon-greedy strategy, the local optimum trap is avoided; by setting the parameter epsilon, the balance between the algorithm efficiency and the local optimum problem can be sought.
Description
Technical field
The present invention relates to war simulation technical field, in particular relate to the Q learning attack mesh of a kind of Simulation-Oriented
Mark distribution method.
Background technology
In no-data region, traditional target assignment method is at two row (task and performance element) be given in advance
The Maximum Value (or consuming minimum) produced after finding a pairing scheme to make it match in element, but due to it
Rely on gradient function thus be easily absorbed in local optimum trap;Ant group algorithm has optimizing ability for Target Assignment
Good advantage, but calculate the longest, especially when complication system, inefficient;Compare ant colony to calculate
Method, particle cluster algorithm has when Target Assignment that speed of searching optimization is fast, the simple advantage of algorithm, but is processing
Local optimum still it is easily trapped into during dispersed problem;When genetic algorithm solves problems, although land productivity very well
By features such as the self-organization of this algorithm, adaptivity, concurrency, uncertainties, but fail to overcome it
The defect of local search ability difference, causes algorithmic statement slow, greatly have impact on search efficiency, and easily
Precocious phenomenon occurs.To this end, a kind of being avoided that need to be proposed be absorbed in local optimum and the higher target of attack of efficiency
Distribution method, for no-data region.
Summary of the invention
For the deficiencies in the prior art, the present invention proposes the Q learning attack Target Assignment of a kind of Simulation-Oriented
Method, for obtaining the optimum results of red fighter formation Target Assignment in no-data region.
The technical scheme is that
The Q learning attack target assignment method of a kind of Simulation-Oriented, comprises the following steps:
(1) determining original state, obtain red blue both sides' air combat situation information, both sides' air combat situation information includes
Both sides form into columns interior number of aircraft and formation aircraft relevant parameter, calculate for red Target Assignment and air battle model and carry
For input;
(2) determine that red is formed into columns the behavior aggregate that can perform, and complete " state-action " of strict regulations is right;
Determine suitable probability ε value and use ε-greedy strategy to carry out red Action Selection;
(3) regulation Q learning algorithm Reward Program, final state and the condition of convergence, application Q learning algorithm is
The red target of attack distribution iteration that carries out is until meeting the condition of convergence.
In the step (1) of the present invention, the formation aircraft relevant parameter calculated for air battle model includes that both sides fly
Machine quantity, commander the factor, opportunity of combat vulnerability, Multi-target Attacking ability, sortie ratio, allow belligerent than, make
War the Radar Cross Section of aircraft, operational aircraft Radar cross-section redaction coefficient, airborne radar maximum detection distance,
On airborne absolutely empty bullet maximum effective range, maximum detection distance, probability of detection, early warning plane to target are effectively sent out
Existing destination probability, guided missile quantity and guided missile scoring.Original state for the study of red Target Assignment is belligerent
Front both sides respectively form into columns interior number of aircraft;Target Assignment study also needs to be returned based on air battle model result of calculation
The relevant parameter of report value and Q learning algorithm such as determines discount factor and the Learning Step of Q learning algorithm.
In the present invention, red is formed into columns the most different target distribution schemes of the action that can perform, with Arab's number
Word table shows;Both sides participate in aircraft kind and quantity perceptually state in the formation of air battle, represent with matrix.
Red Action Selection have employed ε-greedy strategy, i.e. chooses current Q function with a greater probability (1-ε) and reaches
Action to maximum;Attempt other different actions with probability ε simultaneously.
In the step (3) of the present invention, final state is to have each interior number of aircraft of forming into columns of a side in operation both sides
It is 0;The condition of convergence for all " state-actions " to the amplitude of variation of awards sum be respectively less than and give threshold values;When
Arriving final state, algorithm will terminate an iteration circulation, and continue to start new iteration from original state and follow
Ring, until study terminates.
The present invention has versatility, carries out target of attack distribution based on Q learning algorithm and has the advantage that
(1) dependence to priori has been broken away from;
(2) introducing to ε-greedy strategy, it is to avoid be absorbed in local optimum trap;
(3) by the setting to parameter ε, can be at seeking balance in efficiency of algorithm and local optimum problem.
Accompanying drawing explanation
Fig. 1 is that target of attack distributes schematic diagram.
Fig. 2 is the flow chart of the present invention.
Detailed description of the invention
The present invention is described in detail below in conjunction with the accompanying drawings.
Target of attack distribution in air battle refers to, monitors according to combat duty and Theater Air War situation, formulates and make
Fighting spirit figure, the aircraft resource dispatching, controlling a whole group of planes in units of fighter formation carries out corresponding formation point
Joining, its concept is as shown in Figure 1.The present invention proposes the Q learning attack target assignment method of a kind of Simulation-Oriented,
For obtaining the optimum results of red fighter formation Target Assignment in no-data region.
Step of the present invention includes:
(1) determine original state, obtain both sides' air combat situation information, including both sides form into columns interior number of aircraft with
Formation aircraft relevant parameter, calculates for red Target Assignment and air battle model and provides input.
(2) determine that red is formed into columns the behavior aggregate that can perform, and complete " state-action " of strict regulations is right;
Determine suitable ε value and use ε-greedy strategy to carry out red Action Selection.
(3) regulation Q learning algorithm Reward Program, final state and the condition of convergence, apply Q learning algorithm
Target of attack distribution iteration is carried out until meeting the condition of convergence for red.
It is the flow chart of the Q learning attack target assignment method of Simulation-Oriented of the present invention with reference to Fig. 2, Fig. 2, wherein
Q(st,at) represent that red is in state stLower employing action atThe incentive discount obtained and;γ ∈ [0,1] be discount because of
Son, the importance between balance return immediately and long-term return;α is Learning Step, for control algolithm
Learning efficiency;R is Reward Program, rt+1Represent at stLower selection action atReturn immediately;A is all feasible
Behavior aggregate;ε is greedy strategy parameter, introduces and is used for avoiding local optimum trap;Represent at shape
State st+1Lower everything maximal rewards, reflects the mode of action of long-term return.
Below in conjunction with a specific embodiment, after meeting with blue party assault aircraft and jammer with red patrol aircraft,
Target Assignment under red early warning plane is commanded is background, the implementation process of the detailed description present invention:
(1) original state is determined, its form such as table 1 below.Wherein, red early warning plane have learning ability and
The most directly participate in fighting (assuming that will not be shot down).It addition, the action award value of red is all under initial condition
0, the aforementioned relevant parameter of both sides' air combat formation calculated for air battle model preserves with tabular form.
Table 1 state representation form
Determine discount factor and the Learning Step of Q learning algorithm simultaneously, discount factor γ=0.9 is set;Study step
Long α=0.1.
(2) determine that red is formed into columns the behavior aggregate that can perform, and complete " state-action " of strict regulations be right,
Its form such as table 2.
Table 2 " state-action " is to representation
It is simultaneously introduced ε-greedy strategy, goes out suitable ε value according to emulation Data Mining.General scenario takes ε=0.1
Can meet demand, its meaning is: when red carries out Action Selection, chooses current Q function with 0.9 probability and reaches
The action of maximum;Attempt other different actions with the probability of 0.1 simultaneously.
(3) red Reward Program is defined as follows: after using air battle model to calculate air battle result, when judging blue party
Assault aircraft or countermeasure aircraft have a frame to be shot down, then red obtains the award of+1;When judging that red patrols
Patrolling aircraft has a frame to be shot down, then red obtains the return of-1.Self-defence is selected to evade the system i.e. abandoned at red
Sky temporary, gives red punishment-10.
Employing air battle model is calculated by air battle result, and after above-mentioned Reward Program calculates, available red is held
The return immediately that action is made.
Red " state-action " to incentive discount value renewal equation as follows:
rt+1It is the return immediately of red execution action;Represent in state st+1Lower red everything is
Big return.
Definition final state is 0 for a number formulary amount in belligerent both aircraft, each time from arbitrary initial state
The process reaching final state is referred to as one act (also referred to as scene).
Reach final state every time, it is judged that red " state-action " to incentive discount value matrix whether restrain.
If not convergence, go back to original state iteration again, until red " state-action " to incentive discount
Value matrix is restrained.
To red " state-action " to the matrix that is not zero of incentive discount value carry out data analysis, it is known that red
Optimal strategy is incentive discount and the combination of actions of value maximum under this state.
Claims (6)
1. the Q learning attack target assignment method of a Simulation-Oriented, it is characterised in that comprise the following steps:
(1) determining original state, obtain red blue both sides' air combat situation information, both sides' air combat situation information includes
Both sides form into columns interior number of aircraft and formation aircraft relevant parameter, calculate for red Target Assignment and air battle model and carry
For input;
(2) determine that red is formed into columns the behavior aggregate that can perform, and complete " state-action " of strict regulations is right;
Determine suitable probability ε value and use ε-greedy strategy to carry out red Action Selection;
(3) regulation Q learning algorithm Reward Program, final state and the condition of convergence, application Q learning algorithm is
The red target of attack distribution iteration that carries out is until meeting the condition of convergence.
The Q learning attack target assignment method of Simulation-Oriented the most according to claim 1, it is characterised in that
In step (1), formation aircraft relevant parameter includes both aircraft quantity, commander's factor, opportunity of combat vulnerability, many
Target attack ability, sortie ratio, allow belligerent ratio, the Radar Cross Section of operational aircraft, operational aircraft
Radar cross-section redaction coefficient, airborne radar maximum detection distance, airborne absolutely empty bullet maximum effective range, maximum are sent out
Now apart to the probability of detection of target, the effective target detection probability of early warning plane, guided missile quantity and guided missile scoring.
The Q learning attack target assignment method of Simulation-Oriented the most according to claim 1, it is characterised in that
In step (3), final state is that in having a side each to form into columns in operation both sides, number of aircraft is 0;The condition of convergence
For all " state-actions " to the amplitude of variation of awards sum be respectively less than and give threshold values;When arriving final state,
Algorithm will terminate an iteration circulation, and continue to start new iterative cycles from original state, until study knot
Bundle.
The Q learning attack target assignment method of Simulation-Oriented the most according to claim 1, it is characterised in that
In step (1), determining original state, wherein red has s11Frame early warning plane;s12Frame patrol aircraft;Blue
There is s side21Frame jammer;s22Frame assault aircraft;Red early warning plane has learning ability and the most directly participates in fighting
And suppose to be shot down;
In step (2), determine that red is formed into columns the behavior aggregate that can perform, and strict regulations complete " state-
Action " right, its form is as shown in table 2:
Table 2 " state-action " is to representation
Choose ε=0.1, when i.e. red carries out Action Selection, choose current Q function with 0.9 probability and reach maximum
Action;Attempt other different actions with the probability of 0.1 simultaneously;
In step (3), red Reward Program is defined as follows: after using air battle model to calculate air battle result, when
Judge that blue party assault aircraft or countermeasure aircraft have a frame to be shot down, then red obtains the award of+1;When sentencing
Determining red patrol aircraft has a frame to be shot down, then red obtains the return of-1.Self-defence is selected to evade at red
When i.e. abandoning control of the air, give red punishment-10;
Red " state-action " to incentive discount value renewal equation as follows:
rt+1It is the return immediately of red execution action;Represent in state st+1Lower red everything is
Big return.
Definition final state is 0 for a number formulary amount in belligerent both aircraft, each time from arbitrary initial state
The process reaching final state is referred to as one act;
Reach final state every time, it is judged that red " state-action " to incentive discount value matrix whether restrain;
If not convergence, go back to original state iteration again, until red " state-action " to incentive discount value
Matrix is restrained.
The Q learning attack target assignment method of Simulation-Oriented the most according to claim 4, it is characterised in that
Determine discount factor and the Learning Step of Q learning algorithm, discount factor γ=0.9 is set;Learning Step α=0.1.
The Q learning attack target assignment method of Simulation-Oriented the most according to claim 4, it is characterised in that
To red " state-action " to the matrix that is not zero of incentive discount value carry out data analysis, it is known that red is optimal
Strategy is incentive discount and the combination of actions of value maximum under this state.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610427869.3A CN105844068B (en) | 2016-06-16 | 2016-06-16 | A kind of Q learning attack target assignment methods of Simulation-Oriented |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610427869.3A CN105844068B (en) | 2016-06-16 | 2016-06-16 | A kind of Q learning attack target assignment methods of Simulation-Oriented |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105844068A true CN105844068A (en) | 2016-08-10 |
CN105844068B CN105844068B (en) | 2018-11-13 |
Family
ID=56576161
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610427869.3A Active CN105844068B (en) | 2016-06-16 | 2016-06-16 | A kind of Q learning attack target assignment methods of Simulation-Oriented |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105844068B (en) |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106411749A (en) * | 2016-10-12 | 2017-02-15 | 国网江苏省电力公司苏州供电公司 | Path selection method for software defined network based on Q learning |
CN106595671A (en) * | 2017-02-22 | 2017-04-26 | 南方科技大学 | Method and apparatus for planning route of unmanned aerial vehicle based on reinforcement learning |
CN106991500A (en) * | 2017-04-10 | 2017-07-28 | 哈尔滨理工大学 | Inventory allocation method based on multi-Agent network for distributed sales model |
CN108966330A (en) * | 2018-09-21 | 2018-12-07 | 西北大学 | A kind of mobile terminal music player dynamic regulation energy consumption optimization method based on Q-learning |
CN110007688A (en) * | 2019-04-25 | 2019-07-12 | 西安电子科技大学 | A kind of cluster distributed formation method of unmanned plane based on intensified learning |
CN110213262A (en) * | 2019-05-30 | 2019-09-06 | 华北电力大学 | A kind of full-automatic advanced escape technical testing method based on depth Q network |
CN110554707A (en) * | 2019-10-17 | 2019-12-10 | 陕西师范大学 | Q learning automatic parameter adjusting method for aircraft attitude control loop |
CN111046497A (en) * | 2019-12-24 | 2020-04-21 | 中国航空工业集团公司沈阳飞机设计研究所 | Rapid assessment device for high-altitude high-speed airplane penetration viability |
CN113238579A (en) * | 2021-05-18 | 2021-08-10 | 西安电子科技大学 | Multi-unmanned aerial vehicle cluster formation obstacle avoidance method based on Oc-ACO algorithm |
CN113641191A (en) * | 2021-10-14 | 2021-11-12 | 中国人民解放军空军预警学院 | Airspace configuration method and equipment for cooperative operation of early warning machine and jammer |
CN115494831A (en) * | 2021-06-17 | 2022-12-20 | 中国科学院沈阳自动化研究所 | Man-machine autonomous intelligent cooperative tracking method |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105302153A (en) * | 2015-10-19 | 2016-02-03 | 南京航空航天大学 | Heterogeneous multi-UAV (Unmanned Aerial Vehicle) cooperative scouting and striking task planning method |
CN105678030A (en) * | 2016-03-03 | 2016-06-15 | 黄安祥 | Air-combat tactic team simulating method based on expert system and tactic-military-strategy fractalization |
-
2016
- 2016-06-16 CN CN201610427869.3A patent/CN105844068B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105302153A (en) * | 2015-10-19 | 2016-02-03 | 南京航空航天大学 | Heterogeneous multi-UAV (Unmanned Aerial Vehicle) cooperative scouting and striking task planning method |
CN105678030A (en) * | 2016-03-03 | 2016-06-15 | 黄安祥 | Air-combat tactic team simulating method based on expert system and tactic-military-strategy fractalization |
Non-Patent Citations (3)
Title |
---|
EDUARDO 等: "Dynamic Analysis of Multiagent Q-learning with ǫ-greedy Exploration", 《ICML 09 PROCEEDINGS OF THE 26TH ANNUAL INTERNATIONAL CONFERENCE ON MACHINE LEARNING》 * |
STEFAN WENDER 等: "Applying Reinforcement Learning to Small Scale Combat in the Real-Time Strategy Game StarCraft:Broodwar", 《2012 IEEE CONFERENCE ON COMPUTATIONAL INTELLIGENCE AND GAMES(CIG)》 * |
刘波 等: "基于群集智能的协同多目标攻击空战决策", 《航空学报》 * |
Cited By (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106411749B (en) * | 2016-10-12 | 2019-07-30 | 国网江苏省电力公司苏州供电公司 | A kind of routing resource for software defined network based on Q study |
CN106411749A (en) * | 2016-10-12 | 2017-02-15 | 国网江苏省电力公司苏州供电公司 | Path selection method for software defined network based on Q learning |
CN106595671A (en) * | 2017-02-22 | 2017-04-26 | 南方科技大学 | Method and apparatus for planning route of unmanned aerial vehicle based on reinforcement learning |
CN106991500A (en) * | 2017-04-10 | 2017-07-28 | 哈尔滨理工大学 | Inventory allocation method based on multi-Agent network for distributed sales model |
CN108966330A (en) * | 2018-09-21 | 2018-12-07 | 西北大学 | A kind of mobile terminal music player dynamic regulation energy consumption optimization method based on Q-learning |
CN110007688A (en) * | 2019-04-25 | 2019-07-12 | 西安电子科技大学 | A kind of cluster distributed formation method of unmanned plane based on intensified learning |
CN110007688B (en) * | 2019-04-25 | 2021-06-01 | 西安电子科技大学 | Unmanned aerial vehicle cluster distributed formation method based on reinforcement learning |
CN110213262B (en) * | 2019-05-30 | 2022-01-28 | 华北电力大学 | Full-automatic advanced escape technology detection method based on deep Q network |
CN110213262A (en) * | 2019-05-30 | 2019-09-06 | 华北电力大学 | A kind of full-automatic advanced escape technical testing method based on depth Q network |
CN110554707A (en) * | 2019-10-17 | 2019-12-10 | 陕西师范大学 | Q learning automatic parameter adjusting method for aircraft attitude control loop |
CN111046497A (en) * | 2019-12-24 | 2020-04-21 | 中国航空工业集团公司沈阳飞机设计研究所 | Rapid assessment device for high-altitude high-speed airplane penetration viability |
CN111046497B (en) * | 2019-12-24 | 2023-04-07 | 中国航空工业集团公司沈阳飞机设计研究所 | Rapid assessment device for high-altitude high-speed airplane penetration viability |
CN113238579A (en) * | 2021-05-18 | 2021-08-10 | 西安电子科技大学 | Multi-unmanned aerial vehicle cluster formation obstacle avoidance method based on Oc-ACO algorithm |
CN115494831A (en) * | 2021-06-17 | 2022-12-20 | 中国科学院沈阳自动化研究所 | Man-machine autonomous intelligent cooperative tracking method |
CN115494831B (en) * | 2021-06-17 | 2024-04-16 | 中国科学院沈阳自动化研究所 | Tracking method for autonomous intelligent collaboration of human and machine |
CN113641191A (en) * | 2021-10-14 | 2021-11-12 | 中国人民解放军空军预警学院 | Airspace configuration method and equipment for cooperative operation of early warning machine and jammer |
CN113641191B (en) * | 2021-10-14 | 2022-01-14 | 中国人民解放军空军预警学院 | Airspace configuration method and device for cooperative operation of early warning machine and jammer |
Also Published As
Publication number | Publication date |
---|---|
CN105844068B (en) | 2018-11-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105844068A (en) | Distribution method oriented to simulation Q learning attack targets | |
CN110083971B (en) | Self-explosion unmanned aerial vehicle cluster combat force distribution method based on combat deduction | |
Xin et al. | An efficient rule-based constructive heuristic to solve dynamic weapon-target assignment problem | |
CN107886184B (en) | Multi-type air defense weapon mixed-programming fire group target distribution optimization method | |
CN112507622B (en) | Anti-unmanned aerial vehicle task allocation method based on reinforcement learning | |
Xin et al. | Efficient decision makings for dynamic weapon-target assignment by virtual permutation and tabu search heuristics | |
CN108319132B (en) | Decision-making system and method for unmanned aerial vehicle air countermeasure | |
CN106779210B (en) | Algorithm of Firepower Allocation based on ant group algorithm | |
CN111859541B (en) | PMADDPG multi-unmanned aerial vehicle task decision method based on transfer learning improvement | |
CN109408877B (en) | Intelligent shooting decision-making method for anti-tank missile teams | |
CN115222271A (en) | Weapon target distribution method based on neural network | |
CN111797966B (en) | Multi-machine collaborative global target distribution method based on improved flock algorithm | |
CN110986680B (en) | Composite interception method for low-speed small targets in urban environment | |
Ha et al. | A stochastic game-based approach for multiple beyond-visual-range air combat | |
Pan et al. | Radar jamming strategy allocation algorithm based on improved chaos genetic algorithm | |
CN110782062A (en) | Many-to-many packet interception target distribution method and system for air defense system | |
CN112464549B (en) | Dynamic allocation method of countermeasure unit | |
CN114548674A (en) | Multi-agent confrontation scene-oriented threat situation assessment method, device and equipment | |
CN114675262A (en) | Hypersonic aircraft searching method based on guide information | |
CN108871080A (en) | Air defence weapon system target interception ordering designs method | |
CN114047761A (en) | Elastic killer network construction method and device based on formation cross-platform resource scheduling | |
CN110930054A (en) | Data-driven battle system key parameter rapid optimization method | |
CN112464548B (en) | Dynamic allocation device for countermeasure unit | |
Xu et al. | Approach for combat capability requirement generation and evaluation of new naval gun | |
CN116976188A (en) | Improved intelligent algorithm for solving weapon-target distribution problem |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |