A method for designing a broadband digital coding metasurface unit based on DDPG

Abstract

The application discloses a kind of wideband digital coding metasurface unit design method based on DDPG, it is related to digital coding metasurface technical field, digital coding metasurface unit designed by the application has greater bandwidth, can improve the regulation and control range of digital coding metasurface, so that it can realize accurate electromagnetic wave regulation and control in more extensive frequency range.In addition, the application is based on the DDPG algorithm in deep reinforcement learning (DRL), without training label, with the characteristics of online learning and sample generation, higher storage efficiency, realizes the design of wideband digital coding metasurface unit.The phase difference of the designed 1-bit digital coding metasurface unit is 180±20°, and the relative bandwidth in the frequency band of 8.96 to 11.39GHz is 23.88%;The phase difference of the designed 2-bit digital coding metasurface unit is 90±20°, and the relative bandwidth in the frequency band of 9.51 to 10.46GHz is 9.51%.

Description

Technical Field

[0001] This invention relates to the field of digital coded metasurface technology, specifically to a broadband digital coded metasurface unit design method based on DDPG. Background Technology

[0002] Digitally coded metasurface units can be designed to encode digital information on each unit, and by adjusting the structure and distribution of the units that make up the digitally coded metasurface, flexible manipulation of electromagnetic waves can be achieved. Existing metasurface units typically operate only within a narrow frequency band, thus preventing precise electromagnetic wave control over a wide frequency range.

[0003] For a 16×16 random dot matrix unit with eightfold symmetry, composed of two types of sub-units, different combinations of sub-units correspond to different reflection coefficients. Since these 16×16 square sub-blocks are subject to eightfold symmetry, only one-eighth of the pattern module needs to be studied, thus the eightfold symmetry unit will have 2... 36 The existing technology typically uses full-wave simulation software CST Microwave Studio to manually arrange these combinations for simulation and post-processing to obtain the reflection coefficient. This method is time-consuming, labor-intensive, and inefficient.

[0004] Discrete action space and continuous action space are two important concepts in reinforcement learning, used to describe the types of actions an agent can choose during the decision-making process. The number of actions in the discrete action space is countable, while the number of actions in the continuous action space is uncountable. The DDPG (Deep Deterministic Policy Gradient) algorithm based on the Actor-Critic framework was proposed by Lillicrap et al. in 2016. It is not only suitable for solving problems in high-dimensional continuous action spaces, but also has great advantages in handling problems in large-scale discrete action spaces. Summary of the Invention

[0005] The purpose of this invention is to address the problem that existing metasurface units can typically only operate within a narrow frequency band, thus preventing them from achieving precise electromagnetic wave control over a wide frequency range. This invention proposes a broadband digitally coded metasurface unit design method based on DDPG.

[0006] The technical solution adopted by the present invention to solve the above-mentioned technical problems is as follows:

[0007] A broadband digitally coded metasurface unit design method based on DDPG, the method comprising the following steps:

[0008] Step 1: Use a square module as the basic pattern of the digital encoding unit, and use PyCharm to randomly generate a basic pattern. Then call CST to perform co-simulation, modeling and post-processing to obtain the reflection coefficient corresponding to the basic pattern. Finally, combine the DDPG algorithm to obtain 1-bit encoding units element0 and element1.

[0009] Step 2: Obtain the reflection phase from the reflection coefficient, and based on the reflection phase from the reflection coefficient, obtain the phase s of the digital encoding unit. ran and s ran +180°, and the phase difference θ(s) between digital encoding units ran ,s ran +180°), and according to the phase s of the digital encoding unit ran s ran +180° and the phase difference θ(s) between digital encoding units ran ,s ran +180°) to obtain the initial state s0 of the coding units element0 and element1 < s ran ,s ran +180°,θ(s ran ,s ran +180°)>, then obtain the phase s of coding units element0 and element1 in the current state within the frequency band. t0 s t1 And in the current state, the phase difference θ(s) between coding units element0 and element1 within the frequency band. t0 ,s t1 This leads to the state space s of the 1-bit digital encoding unit. t =<s t0 ,s t1 ,θ(s t0 ,s t1 >

[0010] Step 3: Define the action space as an array of length 72. The array of length 72 includes one-eighth of the basic pattern of encoding unit element0, which is a 36-length array, and one-eighth of the basic pattern of encoding unit element1, which is a 36-length array.

[0011] Step 4: Define the target phase difference curve, i.e. Where x represents the frequency point, and then according to θ(s) t0 ,s t1The phase difference curve is obtained, and the root mean square error between the phase difference curve and the target phase difference curve in the frequency band is calculated. Then, the root mean square error is negative, and finally, the negative result is used as the reward in the current state.

[0012] Step 5: Define the termination condition as follows: the reflection amplitude of coding units element0 and element1 in 8-12GHz is greater than 0.95, the maximum phase difference between coding units is less than 200°, and the frequency point corresponding to the maximum phase difference between coding units is between 9-11GHz.

[0013] Step 6: Based on DDPG, and combined with the state space in Step 2, the action space in Step 3, the reward in Step 4, and the termination condition in Step 5, obtain the arrangement of the 1-bit encoding units element0 and element1.

[0014] Step 7: Use the 1-bit encoding units element0 and element1 as element00 and element10 in the 2-bit encoding unit. Then, obtain the reflection phase in element00 and obtain the phase s of the digital encoding unit based on the reflection phase in element00. 00 s 00 +90°, s 00 +270°, phase difference θ(s) between digital encoding units 00 ,s 00 +90°) and the phase difference θ(s) between digital encoding units 00 ,s 00 +270°), and according to s 00 s 00 +90° and θ(s) 00 ,s 00 +90°) to obtain the initial state s0 of the coding unit element01. 00 ,s 00 +90°,θ(s 00 ,s 00 +90°)>,

[0015] Then according to s 00 s 00 +270° and θ(s) 00 ,s 00 +270°) to obtain the initial state s0 of the coding unit element11. 00 ,s 00 +270°,θ(s 00 ,s 00 +270°)>, then obtain the phase s of element01 and element11 in the current frequency band.​​t01 s t11 And in the current state, the phase difference θ(s) between coding units element00 and element10, or between coding units element00 and element11 within the frequency band. 00 ,s t01 / 11 ), and according to s t01 s t11 and θ(s) 00 ,s t01 / 11 The state space s of the 2-bit digital encoding unit is obtained. t = 00 ,s t01 / 11 ,θ(s 00 ,s t01 / 11 )>;

[0016] Step 8: Define the action space as an array of length 36, wherein the array of length 36 includes one-eighth of the centrally symmetrical part of the basic pattern of encoding unit element01 or element11;

[0017] Step 9: Define the target phase difference curve function between element00 and element11 as follows: Then according to θ(s) 00 ,s t01 / 11 The phase difference curve between element00 and element11 is obtained, and the root mean square error between the phase difference curve and the target phase difference curve in the frequency band is calculated. The root mean square error is negative, and the negative result is used as the reward for element11 in the current state.

[0018] The target phase difference curve function between element00 and element01 is defined as follows: Then according to θ(s) 00 ,s t01 / 11 The phase difference curve between element00 and element01 is obtained, and the root mean square error between the phase difference curve and the target phase difference curve in the frequency band is calculated. The root mean square error is negative, and the negative result is used as the reward for element01 in the current state.

[0019] Step 10: Define the termination condition as follows: the reflection amplitude of the coding unit in 8-12GHz is greater than 0.95, the maximum phase difference between element00 and element01 is less than 110°, and the frequency point corresponding to the maximum phase difference between coding units is between 9-11GHz.

[0020] ​Step 11: Based on DDPG, and combined with the state space in Step 7, the action space in Step 8, the element01 reward in Step 9, and the termination condition in Step 10, the arrangement of the 2-bit encoding unit element01 is obtained.

[0021] Step 12: Define the termination condition as follows: the reflection amplitude of the coding unit in 8-12GHz is greater than 0.95, the maximum phase difference between element00 and element11 is less than 290°, and the frequency point corresponding to the maximum phase difference between coding units is between 9-11GHz.

[0022] Step 13: Based on DDPG, and combined with the state space in Step 7, the action space in Step 8, the element11 reward in Step 9, and the termination condition in Step 12, the arrangement of the 2-bit encoding unit element11 is obtained.

[0023] Furthermore, the square module has a size of 16×16.

[0024] Furthermore, the basic pattern includes an air module marked 0 and a metal module marked 1.

[0025] Furthermore, the metal module is made of copper with a thickness of 0.018 mm and a conductivity of 5.8e+007 S / m.

[0026] Furthermore, the basic pattern covers one side of a dielectric substrate with a thickness of 2 mm, while the other side of the dielectric substrate is covered by metal.

[0027] Furthermore, the dielectric substrate is made of F4B, and the dielectric constant of F4B is 2.65(1+0.003i).

[0028] Furthermore, the side length L of the square module is 0.5mm.

[0029] The beneficial effects of this invention are:

[0030] The digitally coded metasurface unit designed in this application has a large bandwidth, which can improve the control range of the digitally coded metasurface and enable precise electromagnetic wave control over a wider frequency range. Furthermore, this application is based on the DDPG algorithm in Deep Reinforcement Learning (DRL), which does not require training labels, features online learning and sample generation, and has higher storage efficiency, thus realizing the design of a broadband digitally coded metasurface unit. The designed 1-bit digitally coded metasurface unit has a phase difference of 180±20° and a relative bandwidth of 23.88% in the 8.96 to 11.39 GHz frequency band; the designed 2-bit digitally coded metasurface unit has a phase difference of 90±20° and a relative bandwidth of 9.51% in the 9.51 to 10.46 GHz frequency band. Attached Figure Description

[0031] Figure 1 This is a schematic diagram of the eight-fold symmetrical structure of the digital coding unit in this application;

[0032] Figure 2 This is a side view of the digital encoding unit in this application;

[0033] Figure 3 This is a flowchart of the application;

[0034] Figure 4 This is a schematic diagram of the structure of element0 in this application;

[0035] Figure 5 This is a schematic diagram of the structure of element1 in this application;

[0036] Figure 6 This is a phase diagram of the broadband 1-bit digitally encoded metasurface unit in this application;

[0037] Figure 7 This is an amplitude diagram of the broadband 1-bit digitally encoded metasurface unit in this application;

[0038] Figure 8 This is a schematic diagram of the structure of element00 in this application;

[0039] Figure 9 This is a schematic diagram of the structure of element01 in this application;

[0040] Figure 10 This is a schematic diagram of the structure of element 10 in this application;

[0041] Figure 11 This is a schematic diagram of the structure of element11 in this application;

[0042] Figure 12 This is a phase diagram of the broadband 2-bit digitally encoded metasurface unit in this application;

[0043] Figure 13 This is a phase difference diagram of the broadband 2-bit digitally encoded metasurface unit in this application;

[0044] Figure 14 This is the amplitude diagram of the broadband 2-bit digital coded metasurface unit in this application. Detailed Implementation

[0045] It should be noted that, where there is no conflict, the various embodiments disclosed in this application can be combined with each other.

[0046] Specific implementation method one: Refer to Figure 3This embodiment describes a broadband digitally coded metasurface unit design method based on DDPG, which includes the following steps:

[0047] Step 1: Use a square module as the basic pattern of the digital encoding unit, and use PyCharm to randomly generate a basic pattern. Then call CST to perform co-simulation, modeling and post-processing to obtain the reflection coefficient corresponding to the basic pattern. Finally, combine the DDPG algorithm to obtain 1-bit encoding units element0 and element1.

[0048] Step 2: Obtain the reflection phase from the reflection coefficient, and based on the reflection phase from the reflection coefficient, obtain the phase s of the digital encoding unit. ran and s ran +180°, and the phase difference θ(s) between digital encoding units ran ,s ran +180°), and according to the phase s of the digital encoding unit ran s ran +180° and the phase difference θ(s) between digital encoding units ran ,s ran +180°) to obtain the initial state s0 of encoding units element0 and element1. ran ,s ran +180°,θ(s ran ,s ran +180°)>, then obtain the phase s of coding units element0 and element1 in the current state within the frequency band. t0 s t1 And in the current state, the phase difference θ(s) between coding units element0 and element1 within the frequency band. t0 ,s t1 This leads to the state space s of the 1-bit digital encoding unit. t = t0 ,s t1 ,θ(s t0 ,s t1 )>;

[0049] Step 3: Define the action space as an array of length 72. The array of length 72 includes one-eighth of the basic pattern of encoding unit element0, which is a 36-length array, and one-eighth of the basic pattern of encoding unit element1, which is a 36-length array.

[0050] Step 4: Define the target phase difference curve, i.e. Where x represents the frequency point, and then according to θ(s)​​t0 ,s t1 The phase difference curve is obtained, and the root mean square error between the phase difference curve and the target phase difference curve in the frequency band is calculated. Then, the root mean square error is negative, and finally, the negative result is used as the reward in the current state.

[0051] Step 5: Define the termination condition as follows: the reflection amplitude of coding units element0 and element1 in 8-12GHz is greater than 0.95, the maximum phase difference between coding units is less than 200°, and the frequency point corresponding to the maximum phase difference between coding units is between 9-11GHz.

[0052] Step 6: Based on DDPG, and combined with the state space in Step 2, the action space in Step 3, the reward in Step 4, and the termination condition in Step 5, obtain the arrangement of the 1-bit encoding units element0 and element1.

[0053] Step 7: Use the 1-bit encoding units element0 and element1 as element00 and element10 in the 2-bit encoding unit. Then, obtain the reflection phase in element00 and obtain the phase s of the digital encoding unit based on the reflection phase in element00. 00 s 00 +90°, s 00 +270°, phase difference θ(s) between digital encoding units 00 ,s 00 +90°) and the phase difference θ(s) between digital encoding units 00 ,s 00 +270°), and according to s 00 s 00 +90° and θ(s) 00 ,s 00 +90°) to obtain the initial state s0 of the coding unit element01. 00 ,s 00 +90°,θ(s 00 ,s 00 +90°)>,

[0054] Then according to s 00 s 00 +270° and θ(s) 00 ,s 00 +270°) to obtain the initial state s0 of the coding unit element11. 00 ,s 00 +270°,θ(s 00 ,s 00 ​​+270°)>, then obtain the phase s of element01 and element11 in the current frequency band. t01 s t11 And in the current state, the phase difference θ(s) between coding units element00 and element10, or between coding units element00 and element11 within the frequency band. 00 ,s t01 / 11 ), and according to s t01 s t11 and θ(s) 00 ,s t01 / 11 The state space s of the 2-bit digital encoding unit is obtained. t = 00 ,s t01 / 11 ,θ(s 00 ,s t01 / 11 )>;

[0055] Step 8: Define the action space as an array of length 36, wherein the array of length 36 includes one-eighth of the centrally symmetrical part of the basic pattern of encoding unit element01 or element11;

[0056] Step 9: Define the target phase difference curve function between element00 and element11 as follows: Then according to θ(s) 00 ,s t01 / 11 The phase difference curve between element00 and element11 is obtained, and the root mean square error between the phase difference curve and the target phase difference curve in the frequency band is calculated. The root mean square error is negative, and the negative result is used as the reward for element11 in the current state.

[0057] The target phase difference curve function between element00 and element01 is defined as follows: Then according to θ(s) 00 ,s t01 / 11 The phase difference curve between element00 and element01 is obtained, and the root mean square error between the phase difference curve and the target phase difference curve in the frequency band is calculated. The root mean square error is negative, and the negative result is used as the reward for element01 in the current state.

[0058] Step 10: Define the termination condition as follows: the reflection amplitude of the coding unit in 8-12GHz is greater than 0.95, the maximum phase difference between element00 and element01 is less than 110°, and the frequency point corresponding to the maximum phase difference between coding units is between 9-11GHz.

[0059] ​Step 11: Based on DDPG, and combined with the state space in Step 7, the action space in Step 8, the element01 reward in Step 9, and the termination condition in Step 10, the arrangement of the 2-bit encoding unit element01 is obtained.

[0060] Step 12: Define the termination condition as follows: the reflection amplitude of the coding unit in 8-12GHz is greater than 0.95, the maximum phase difference between element00 and element11 is less than 290°, and the frequency point corresponding to the maximum phase difference between coding units is between 9-11GHz.

[0061] Step 13: Based on DDPG, and combined with the state space in Step 7, the action space in Step 8, the element11 reward in Step 9, and the termination condition in Step 12, the arrangement of the 2-bit encoding unit element11 is obtained.

[0062] This application addresses the problem that existing digitally coded metasurface units can only operate within a narrow frequency band, and that obtaining the reflection coefficient through extensive manual modeling using CST Microwave Studio is difficult. To achieve control over electromagnetic waves across a wider frequency band, this application employs a co-simulation method using CST Studio Suite and PyCharm (an integrated development environment for Python) to design digitally coded units.

[0063] The specific technical solution of this application is as follows:

[0064] The basic pattern of the digital encoding unit of this invention is a 16×16 square module composed of two modules. This basic pattern consists of an air module marked "0" and a metal module marked "1". The metal module is made of copper with a thickness of 0.018 mm and a conductivity of 5.8e+007 S / m. The basic pattern covers one side of a 2 mm thick dielectric substrate made of F4B with a dielectric constant of 2.65(1+0.003i). The other side is entirely covered by metal. The side length of the unit structure is L1 = 10 mm, and the side length of the pattern structure is L2 = 8 mm. The side length of the square module in the unit is L = 0.5 mm. However, due to the large number of combinations of metal and air modules, manual arrangement is time-consuming, labor-intensive, and inefficient. Therefore, this application uses a co-simulation method combining CST Studio Suite and PyCharm (an integrated development environment for Python) to design the digital encoding unit.

[0065] A Markov chain (MDP) consists of an environment and an agent. We consider the digitally coded metasurface unit as the agent and the process of co-simulating with PyCharm and CST Studio Suite as the environment. The MDP comprises a state space, an available action space, state transition functions, and a reward function. CST Studio Suite can be viewed as a state transition function. In the above steps, PyCharm generates an array containing only "0"s and "1"s to represent the combination of patterns on the structural surface. Then, CST Studio Suite is invoked for automatic modeling, simulation, and post-processing. The reflection coefficient is obtained by calling CST Studio Suite from PyCharm.

[0066] Digitally coded metasurface units consist of multiple sub-unit structures with a certain phase difference, represented by different digital states. A 1-bit digitally coded unit has a 180° phase difference, with digital states "0" and "1". A 2-bit digitally coded unit has a 90° phase difference, with digital states "00", "01", "10", and "11". To design the reflective units of 1-bit and 2-bit digitally coded metasurfaces, the phase difference between units, besides the reflection amplitude, should be our focus. The DDPG algorithm aims to obtain the combinations of '0' and '1' on the surface of each unit structure. By carefully designing the state space, action space, reward function, and termination conditions for each generation in the DDPG algorithm, 1-bit and 2-bit digitally coded units were obtained respectively.

[0067] In order to design a 1-bit digital encoding unit (a 1-bit encoding unit such as...) Figure 1 and Figure 2 As shown in the diagram, the digital states are "0" and "1", which we define as element0 (structure diagram as shown in the diagram). Figure 4 (as shown) and element1 (structure diagram as shown) Figure 5 (As shown). Next, the state space, action space, immediate reward, and termination condition are carefully designed as follows.

[0068] 1) State Space: The state space is defined as the phase difference between digital coding units. The phase difference is the reflection phase difference of the coding unit at different frequency points: s t = t0 ,s t1 ,θ(s t0 ,s t1 )>, where s t0 and s t1 It represents the phase of elements 0 and 1 within the current frequency band. θ(s) t0 ,s t1 ​() represents the phase difference between two coding units within the frequency band at the current moment. The phase diagram of a 1-bit digital coding unit is shown below. Figure 6 As shown, the amplitude diagram of a 1-bit digital encoding unit is as follows: Figure 7 As shown.

[0069] 2) Action Space: The action space is defined as an array containing only "0" and "1" to represent the combination of the two modules on the surface, element0 and element1. The length of the array is 72.

[0070] 3) Reward Function: Phase difference is a key aspect of coding unit design. First, a target phase difference curve is defined. The goodness of fit between the target and target phase difference curves is measured by calculating the root mean square error (RMSE) of their respective phase difference curves within the frequency band. The negative value of the calculated result is then defined as the reward at the current moment. We define the target phase difference curve as a Gaussian function: Where x represents the frequency point. Taking TE (transverse electric field) wave as the incident wave, the frequency band ranges from 8 to 12 GHz. The difference between the maximum and minimum values ​​of the target phase difference curve within the entire frequency band is 20°, which is the error of the phase difference between 1 digital coding unit.

[0071] 4) Termination conditions: If the reflection amplitude of the coding unit in 8-12GHz is greater than 0.95, the maximum phase difference between coding units is less than 180+20°, and the corresponding frequency point is between 9-11GHz, then this generation is terminated.

[0072] To design a 2-bit digital encoding unit, the digital states are "00", "01", "10", and "10". We define them as element00 in sequence (the structure diagram is shown below). Figure 8 As shown), element01 (structure diagram as shown) Figure 9 (as shown in the diagram), element10 (structure diagram as shown in the diagram) Figure 10 (as shown) and element11 (structure diagram as shown) Figure 11 (As shown). Based on a pair of defined 1-bit digital encoding units, element00 and element10 correspond to element0 and element1, respectively. Therefore, the phase difference between element01 and element00 is 90°, and the phase difference between element11 and element00 is 270°. Next, the state space, action space, immediate reward, and termination condition are carefully designed as follows.

[0073] 1) State space: s t = t00 / 10 ,s t01 / 11 ,θ(s t00 / 10 ,s t01 / 11 )>, where s t00 s​t01 s t10 and s t11 These represent the phases of elements 00, 01, 10, and 11 within the current frequency band. θ(s) t00 / 10 ,s t01 / 11 This represents the phase difference between two coding units within the frequency band at the current moment. The phase diagram of a 2-bit digital coding unit is shown below. Figure 12 As shown, the phase difference diagram of the 2-bit digital encoding unit is as follows: Figure 13 As shown, the amplitude diagram of the 2-digit encoding unit is as follows: Figure 14 As shown.

[0074] 2) Action Space: The action space is defined as an array containing only "0" and "1" to represent the combination of two modules on the surface of element01 or element11. The length of the array is 36.

[0075] 3) Reward Function: Similarly, we define the target phase difference curve between element11 and element00 as a Gaussian function: The target phase difference curve function between element01 and element00 is defined as:

[0076] 4) Termination Criteria: Similarly, design element01 and element11 separately. If the reflection amplitude of element00 and element01 in the 8-12 GHz range is greater than 0.95, the maximum phase difference between coding units is less than 90 + 20°, and the corresponding frequency point is between 9-11 GHz, then this generation is terminated. If the reflection amplitude of element00 and element11 in the 8-12 GHz range is greater than 0.95, the maximum phase difference between coding units is less than 270 + 20°, and the corresponding frequency point is between 9-11 GHz, then this generation is terminated.

[0077] It should be noted that the specific embodiments are merely explanations and illustrations of the technical solution of the present invention and should not be used to limit the scope of protection. Any modifications made in accordance with the claims and specification of the present invention that are only partial should still fall within the protection scope of the present invention.

Claims

1. A design method for broadband digitally coded metasurface units based on DDPG, characterized in that, The method includes the following steps: Step 1: Use a square module as the basic pattern of the digital encoding unit, and use PyCharm to randomly generate a basic pattern. Then call CST to perform co-simulation, modeling and post-processing to obtain the reflection coefficient corresponding to the basic pattern. Finally, combine the DDPG algorithm to obtain 1-bit encoding units element0 and element1. Step 2: Obtain the reflection phase from the reflection coefficient, and based on the reflection phase from the reflection coefficient, obtain the phase of the digital encoding unit. and and the phase difference between digital encoding units And according to the phase of the digital encoding unit , and the phase difference between digital encoding units Obtain the initial states of encoding units element0 and element1. Then, the phase of coding units element0 and element1 within the frequency band in the current state is obtained. , And the phase difference between coding units element0 and element1 within the frequency band in the current state. This leads to the state space of the 1-bit digital encoding unit. ; Step 3: Define the action space as an array of length 72. The array of length 72 includes one-eighth of the basic pattern of encoding unit element0, which is a 36-length array, and one-eighth of the basic pattern of encoding unit element1, which is a 36-length array. Step 4: Define the target phase difference curve, i.e. ,in Indicate the frequency point, and then according to The phase difference curve is obtained, and the root mean square error between the phase difference curve and the target phase difference curve in the frequency band is calculated. Then, the root mean square error is negative, and finally, the negative result is used as the reward in the current state. Step 5: Define the termination condition as follows: the reflection amplitude of coding units element0 and element1 in 8-12GHz is greater than 0.95, the maximum phase difference between coding units is less than 200º, and the frequency point corresponding to the maximum phase difference between coding units is between 9-11GHz. Step 6: Based on DDPG, and combined with the state space in Step 2, the action space in Step 3, the reward in Step 4, and the termination condition in Step 5, obtain the arrangement of the 1-bit encoding units element0 and element1. Step 7: Use the 1-bit encoding units element0 and element1 as element00 and element10 in the 2-bit encoding unit. Then, obtain the reflection phase in element00 and obtain the phase of the digital encoding unit based on the reflection phase in element00. , , Phase difference between digital encoding units and the phase difference between digital encoding units and according to , and Obtain the initial state of the encoding unit element01 , Then according to , and Obtain the initial state of the encoding unit element11 Then, obtain the phase of element01 and element11 within the frequency band in the current state. , And the phase difference between coding units element00 and element10, or coding units element00 and element11, within the frequency band under the current state. and according to , as well as Obtain the state space of the 2-bit digital encoding unit ; Step 8: Define the action space as an array of length 36, wherein the array of length 36 includes one-eighth of the centrally symmetrical part of the basic pattern of the encoding unit element01 or element11; Step 9: Define the target phase difference curve function between element00 and element11 as follows: Then according to Obtain the phase difference curve between element00 and element11, and use it to calculate the root mean square error of the phase difference curve and the target phase difference curve in the frequency band. Then, take the negative value of the root mean square error and use the negative value as the reward of element11 in the current state. The target phase difference curve function between element00 and element01 is defined as follows: Then according to Obtain the phase difference curve between element00 and element01, and use it to calculate the root mean square error of the phase difference curve and the target phase difference curve in the frequency band. Then, take the negative value of the root mean square error and use the negative value as the reward of element01 in the current state. Step 10: Define the termination condition as follows: the reflection amplitude of the coding unit in 8-12GHz is greater than 0.95, the maximum phase difference between element00 and element01 is less than 110º, and the frequency point corresponding to the maximum phase difference between coding units is between 9-11GHz. Step 11: Based on DDPG, and combined with the state space in Step 7, the action space in Step 8, the element01 reward in Step 9, and the termination condition in Step 10, the arrangement of the 2-bit encoding unit element01 is obtained. Step 12: Define the termination condition as follows: the reflection amplitude of the coding unit in 8-12GHz is greater than 0.95, the maximum phase difference between element00 and element11 is less than 290º, and the frequency point corresponding to the maximum phase difference between coding units is between 9-11GHz. Step 13: Based on DDPG, and combined with the state space in Step 7, the action space in Step 8, the element11 reward in Step 9, and the termination condition in Step 12, the arrangement of the 2-bit encoding unit element11 is obtained.

2. The broadband digitally coded metasurface unit design method based on DDPG according to claim 1, characterized in that... The square module is 16×16 in size.

3. The broadband digitally coded metasurface unit design method based on DDPG according to claim 1, characterized in that... The basic pattern includes an air module marked 0 and a metal module marked 1.

4. The broadband digitally coded metasurface unit design method based on DDPG according to claim 3, characterized in that... The metal module is made of copper with a thickness of 0.018 mm and a conductivity of 5.8e+007 S / m.

5. The broadband digitally coded metasurface unit design method based on DDPG according to claim 1, characterized in that... The basic pattern is covered on one side of a dielectric substrate with a thickness of 2 mm, and the other side of the dielectric substrate is covered by metal.

6. The broadband digitally coded metasurface unit design method based on DDPG according to claim 5, characterized in that... The dielectric substrate is made of F4B, and the dielectric constant of F4B is 2.65 (1+0.003i).

7. The broadband digitally coded metasurface unit design method based on DDPG according to claim 1, characterized in that... The side length of the square module =0.5mm.

Images