Model of Internet of Vehicles three-dimensional communication system based on Start-RIS and UAV-mounted RIS cooperation and optimization method thereof

Abstract

The invention provides an Internet of Vehicles three-dimensional communication system model based on Start-RIS and UAV-mounted RIS cooperation and an optimization method, the system comprises a base station BS, a vehicle-mounted user Vi, a star-shaped reconfigurable intelligent surface Start-RIS, a vehicle user pair Vp and Vp ', an unmanned aerial vehicle UAV-mounted RIS equipped with an RIS, a V2I communication link and a V2V communication link, the V2I link performs signal forwarding and optimization through the Start-RIS, the V2V link is enhanced through the UAV-mounted RIS, and the V2V communication link performs signal forwarding and optimization through the Start-RIS. According to the method, the Start-RIS and the UAV-mounted RIS are combined, the OFDM technology is adopted for signal transmission, a system model considers the synergistic effect and signal interference of the Start-RIS and the UAV-mounted RIS, the SINR and the rate of a link are calculated based on the Shannon theorem, the total rate Rsum of the system is the sum of the rates of the V2I link and the V2V link, the optimization target is to maximize the Rsum, and constraint conditions such as power, reflection coefficients, the SINR and bandwidth are met. The optimization method is based on an Actor-Critic framework in deep reinforcement learning (DRL), a continuous action space is optimized through a TD3 algorithm, and the convergence speed and the resource utilization rate are improved in combination with an intelligent trigger mechanism.

Description

Technical Field

[0001] The present invention belongs to the technical field of wireless communication, and in particular relates to a model of a three-dimensional communication system of an Internet of Vehicles based on the collaboration of Star-RIS and UAV-mounted RIS and an optimization method thereof. Background Art

[0002] In the field of vehicle-to-everything (V2X) communications, reliable and efficient communication coverage is key. Traditional reconfigurable intelligent surface (RIS) technology can improve signal transmission in some cases, but its limitation is that it can only reflect signals on one side, with limited coverage, and it is difficult to cope with dynamic and complex communication environments. With the rapid development of intelligent transportation and unmanned driving technology, vehicle-to-everything communications need to cope with higher challenges, such as high-speed movement, multipath interference and occlusion.

[0003] To overcome these limitations, the Star-shaped Reconfigurable Intelligent Surface (Star-RIS) was born. Unlike traditional RIS, Star-RIS can simultaneously reflect and transmit signals, thereby providing 360-degree coverage. However, ground-deployed Star-RIS still has limitations in spatial coverage, especially when facing scenes with complex terrain or many obstacles. Therefore, combining the UAV-mounted RIS in the air (UAV-mountedRIS) becomes an effective solution. UAV-mountedRIS can flexibly move and adjust its position in three-dimensional space, providing V2X systems with a wider range of signal coverage and optimization capabilities.

[0004] However, in the existing dynamic Internet of Vehicles environment, the following major problems exist: first, the decision-making stability is insufficient, which leads to fluctuations and instability in the communication process; second, the redundant calculation amount is too high, resulting in low computing efficiency and unable to meet real-time communication needs; in addition, the resource utilization rate is low, and the potential of existing network resources cannot be fully utilized; finally, the dynamic adaptability is insufficient, and the system is difficult to make timely and effective adjustments in a rapidly changing network environment, affecting the reliability and efficiency of communication.

[0005] Therefore, the present invention proposes a model of a three-dimensional communication system of Internet of Vehicles based on the collaboration of Star-RIS and UAV-mounted RIS and an optimization method thereof. Summary of the invention

[0006] The present invention proposes a model of a three-dimensional communication system of Internet of Vehicles based on the collaboration of Star-RIS and UAV-mounted RIS and an optimization method thereof. By introducing a reinforcement learning algorithm such as TD3 and an improved trigger mechanism combining state and time thresholds, the decision-making accuracy and stability are effectively improved, redundant calculations are reduced, and resource utilization is optimized, thereby achieving efficient and stable communication in a complex Internet of Vehicles environment.

[0007] In order to achieve the above technical effects, the present invention is implemented through the following technical solutions:

[0008] A model of a three-dimensional communication system of a vehicle network based on the collaboration of Star-RIS and UAV-mounted RIS and an optimization method thereof, characterized in that the model includes: a base station BS, a plurality of vehicle-mounted users V i , Star-shaped reconfigurable intelligent surface Star-RIS, multiple pairs of vehicle users V p and V p' , UAV-mounted RIS equipped with a reconfigurable smart surface, multiple V2I communication links between vehicle-mounted users Vi and the base station BS, and multiple V2V communication links between vehicle-user pairs, where i≥1, p≥1, p'≥1.

[0009] Furthermore, the V2I communication link performs signal forwarding and optimization through Star-RIS; the V2V communication link performs signal forwarding and enhancement through UAV-mounted RIS; orthogonal frequency division multiplexing (OFDM) technology is used in the V2I communication link and the V2V communication link; the V2I communication link and the V2V communication link share the same spectrum resources.

[0010] Furthermore, in the V2V communication link, the received signal y of the pth pair of V2V users p (t) is assisted by the reflected and transmitted signals from Star-RIS, and is also interfered by the reflected path of the V2I signal through the UAV-mounted RIS. The received signal model is:

[0011]

[0012] Explanation of symbols:

[0013] y p (t): The signal received by the receiving end of the Pth pair of V2V users.

[0014] n p (t): The noise at the receiving end of the Pth pair of V2V users, obeying

[0015] Furthermore, in the V2I communication link, the received signal y of the i-th vehicle user i (t) is assisted by the reflected signal from the UAV-mountedRIS and is also interfered by the reflection and transmission path of the V2V signal through Star-RIS. The received signal model is:

[0016]

[0017] Explanation of symbols:

[0018] y i (t): signal received by the i-th vehicle user;

[0019] n i (t): The noise at the receiving end of the i-th vehicle user, obeying

[0020] Further, in the SINR and rate of the V2V communication link, the SINR of the Pth pair of V2V users is defined as follows:

[0021]

[0022] in:

[0023] P p : The transmission power of the pth V2V user transmitter;

[0024] P B : The transmission power of the base station;

[0025] Noise power at the receiving end of the pth pair of V2V users;

[0026] According to Shannon's theorem, the rate R of the Pth pair of V2V users V2V,P for:

[0027] R V2V，p =B V2V log2(1+SINRV 2V，p ), (4)

[0028] Among them, B V2V Indicates the bandwidth allocated to the V2V link.

[0029] Furthermore, in the SINR and rate of the V2I communication link, the SINR of the i-th vehicle user is defined as follows:

[0030]

[0031] in:

[0032] P B : The transmission power of the base station;

[0033] P P : The transmission power of the Pth pair of V2V users;

[0034] Noise power at the receiving end of the i-th vehicle user;

[0035] According to Shannon's theorem, the rate R of the i-th vehicle userV2I,i for:

[0036] R V2I,i =BV 2I log2(1+SINR V2I,i ), (6)

[0037] Among them, B V2I Indicates the bandwidth allocated to the V2I link.

[0038] Furthermore, the sum of the V2V link and V2I link rates is the total rate R of the system. sum ,

[0039]

[0040] Another object of the present invention is to propose an optimization method for a model of a three-dimensional communication system of an Internet of Vehicles based on the collaboration of Star-RIS and UAV-mounted RIS, the optimization method comprising:

[0041] (1) System total rate R sum Optimization, the total system rate R sum The optimization is to maximize R sum The goal is to ensure that the relevant constraints of the communication system are met and an intelligent trigger mechanism is introduced;

[0042] (2) The DRL-based optimization algorithm adopts the Actor-Critic framework of deep reinforcement learning (DRL) and processes the continuous action space through the TD3 algorithm.

[0043] Furthermore, the constraints in step (1) include power constraints, reflection and transmission coefficient constraints, UAV-mountedRIS reflection coefficient constraints, SINR minimum constraints and bandwidth allocation constraints.

[0044] Furthermore, the Actor-Critic framework is based on problem modeling, including state space, action space, reward function and optimization objective.

[0045] The beneficial effects of the present invention are:

[0046] The present invention realizes three-dimensional signal coverage and optimization from the ground to the air by combining Star-RIS and UAV-mounted RIS; Star-RIS can simultaneously reflect and transmit signals to achieve 360-degree all-round coverage, while UAV-mountedRIS can further expand the coverage by flexibly adjusting its position; this ground-air collaborative working mechanism not only significantly improves the communication stability and flexibility of the system, but also can effectively deal with multipath interference and signal shielding in complex Internet of Vehicles environments, meeting the needs of efficient communication;

[0047] 2. The present invention introduces an improved trigger mechanism that combines network status and time threshold to intelligently determine when to update the communication strategy, thereby effectively reducing redundant calculations, improving resource utilization, accelerating algorithm convergence, and improving learning efficiency; by combining the advanced delayed deep deterministic policy gradient (TD3) algorithm, the system exhibits unique advantages in signal coverage, resource allocation optimization, and communication efficiency improvement, which is significantly better than traditional methods, and provides a high-performance solution for the highly dynamic and resource-limited Internet of Vehicles environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0049] Figure 1 It is a system model diagram of the present invention;

[0050] Figure 2 It is a normalized reward function change curve diagram simulated by the present invention;

[0051] Figure 3 It is a normalized loss function change curve diagram simulated by the present invention; DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] Example 1

[0054] See also Figures 1 to 3As shown in FIG. 1 , a model of a three-dimensional communication system of a vehicle network based on the collaboration of Star-RIS and UAV-mounted RIS and an optimization method thereof are provided, wherein the model includes: a base station BS, W vehicle-mounted users V i (i∈{1,2,···,W}), star-shaped reconfigurable intelligent surface Star-RIS, M for vehicle users V p and V p' (P, p'∈{1,2,…,M}), UAV-mounted RIS equipped with reconfigurable smart surface, multiple V2I communication links between vehicle-mounted users Vi and base station BS, and multiple V2V communication links between vehicle-user pairs.

[0055] The UAV-mounted RIS equipped with a reconfigurable smart surface can dynamically adjust its position and optimize signal reflection in the air to provide channel optimization and enhancement for each pair of V2V communication links in a non-direct path (NLoS) environment.

[0056] The Star-RIS is a new type of reconfigurable smart surface. Unlike the traditional RIS that can only reflect signals, Star-RIS supports both signal reflection and transmission, achieving 360° all-round coverage. Its flexible control capability significantly improves communication quality in complex and obstructed scenes.

[0057] In this system, there are a total of W+M links consisting of channel links, W V2I links from the base station to the vehicle-mounted user. Since there is no direct path (LoS), the V2I links are forwarded and optimized through Star-RIS. In addition, there are M V2V links. Due to the lack of LoS path, the V2V links are forwarded and enhanced through UAV-mounted RIS.

[0058] In order to avoid interference within the V2I communication link and the V2V communication link, the system uses orthogonal frequency division multiplexing (OFDM) technology in the V2I and V2V links respectively. However, in order to save spectrum resources, the V2I and V2V communication links share the same spectrum resources. This design brings the following characteristics: there is no interference between the W V2I links, there is no interference between the M V2V links, and there is co-frequency interference between the V2I and V2V communication links.

[0059] In the established channel link mathematical model, the meaning of each symbol definition is as follows:

[0060] x(t): signal sent by the base station.

[0061] s p (t): The signal sent by the transmitter of the pth pair of V2V users.

[0062] S B，n : Channel coefficient from the base station (BS) to the nth unit of STAR-RIS.

[0063] From the nth STAR-RIS unit to the i-th vehicle user V i The channel coefficient of .

[0064] From the sending end v of the p-th group of V2V users p to the channel coefficients of the lth unit on the UAV-mounted RIS.

[0065] From the lth unit on the UAV-mounted RIS to the receiving end of the pth group of V2V users p The channel coefficient of .

[0066] The reflection coefficient matrix of STAR-RIS is represents the reflection gain of the nth unit, represents the corresponding phase shift.

[0067] The transmission coefficient matrix of STAR-RIS is represents the transmission gain of the nth unit, represents the corresponding phase shift.

[0068] The reflection coefficient matrix of the UAV-mounted RIS, where al∈[0,1] represents the reflection gain of the lth unit, represents the corresponding phase shift.

[0069] In the V2V communication link, the received signal y of the pth pair of V2V users p (t) is assisted by the reflected and transmitted signals from Star-RIS, and is also interfered by the reflected path of the V2I signal through the UAV-mounted RIS. The received signal model is:

[0070]

[0071] Explanation of symbols:

[0072] y p (t): The signal received by the receiving end of the Pth pair of V2V users.

[0073] n p (t): The noise at the receiving end of the Pth pair of V2V users, obeying

[0074] In the V2I communication link, the received signal y of the i-th vehicle user i (t) is assisted by the reflected signal from the UAV-mountedRIS and is also interfered by the reflection and transmission path of the V2V signal through Star-RIS. The received signal model is:

[0075]

[0076] Explanation of symbols:

[0077] y i (t): signal received by the i-th vehicle user;

[0078] n i (t): The noise at the receiving end of the i-th vehicle user, obeying

[0079] In the SINR and rate of the V2V communication link, the SINR of the Pth pair of V2V users is defined as follows:

[0080]

[0081] in:

[0082] P p : The transmission power of the pth V2V user transmitter;

[0083] P B : The transmission power of the base station;

[0084] Noise power at the receiving end of the pth pair of V2V users;

[0085] According to Shannon's theorem, the rate R of the Pth pair of V2V users V2V,P for:

[0086] R V2V，p =B V2V log2(1+SINR V2V，p ), (4)

[0087] Among them, B V2V Indicates the bandwidth allocated to the V2V link.

[0088] Furthermore, in the SINR and rate of the V2I communication link, the SINR of the i-th vehicle user is defined as follows:

[0089]

[0090] in:

[0091] P B : The transmission power of the base station;

[0092] P P : The transmission power of the Pth pair of V2V users;

[0093] Noise power at the receiving end of the i-th vehicle user;

[0094] According to Shannon's theorem, the rate R of the i-th vehicle user V2I,i for:

[0095] R V2I，i =B V2I log2(1+SINR V2I，i ), (6)

[0096] Among them, B V2I Indicates the bandwidth allocated to the V2I link.

[0097] Furthermore, the sum of the V2V link and V2I link rates is the total rate R of the system. sum ,

[0098]

[0099] Example 2

[0100] Another object of the present invention is to propose an optimization method for a model of a three-dimensional communication system of an Internet of Vehicles based on the collaboration of Star-RIS and UAV-mounted RIS, the optimization method comprising:

[0101] Total system rate R sum Optimization, the total system rate R sum The optimization is to maximize R sum The goal is to maximize R sum for:

[0102]

[0103] in:

[0104] The maximum transmission power limit for V2V users is P.

[0105] Maximum transmit power limit of the base station.

[0106] Minimum SINR requirement for V2V users.

[0107] Minimum SINR requirement for V2I users.

[0108] B total : The total bandwidth of the system.

[0109] At the same time, ensure that the relevant constraints of the communication system are met, including:

[0110] Power Constraints:

[0111] The transmission power of a V2V user cannot exceed its maximum power:

[0112]

[0113] The total transmit power of a base station cannot exceed its maximum power:

[0114]

[0115] Reflection and transmission coefficient constraints:

[0116] The reflection and transmission gains of Star-RIS must satisfy energy constraints:

[0117]

[0118] Range of reflection coefficient:

[0119]

[0120] Phase shift range:

[0121]

[0122] UAV-mounted RIS reflection coefficient constraints:

[0123] The reflection gain of the UAV-mounted RIS must satisfy the energy constraints:

[0124]

[0125] Phase shift range:

[0126]

[0127] SINR minimum constraint:

[0128] Ensure that the SINR of each V2V user and V2I user meets the minimum quality requirements:

[0129]

[0130]

[0131] Bandwidth allocation constraints:

[0132] Bandwidth allocation satisfies the total bandwidth limit:

[0133] BV2V +B V2I ≤B total ; (18)

[0134] This optimization problem belongs to a non-convex mixed integer nonlinear programming problem. The goal is to jointly adjust the IRS configuration (Φ r ,Φ u ,Ψ) and power allocation (P P ,P B ) to maximize the total rate of V2V and V2I links; the problem has high-dimensional joint optimization, complex constraints (such as power limit, IRS configuration, QoS requirements, etc.), and is affected by the dynamic changes of channel status and vehicle position. At the same time, the huge state space further increases the computational complexity and real-time challenges; due to the shortcomings of traditional optimization methods in computational complexity and adaptability, the present invention adopts a deep reinforcement learning method based on TD3 to achieve efficient strategy learning and dynamic environment adaptation.

[0135] To improve the optimization efficiency, the present invention introduces an event triggering mechanism to decide when to activate the DRL-based optimizer. This mechanism monitors the system state changes and triggers learning only when a significant state update is detected or a predefined time interval is reached, thereby reducing redundant calculations and improving efficiency. Compared with the traditional time triggering mechanism (learning is activated at each time step) or the threshold-based triggering method (learning is triggered only when the state difference exceeds the threshold), this mechanism combines the advantages of both, and is defined as follows:

[0136]

[0137] Among them, δ learning To learn the trigger indicator, δ learning =1 means starting a new DRL process, δ learning =0 means reuse the action generated by the previous learning S t is the system state at time step t, is the state when the DRL process was last triggered; ρ is the threshold of state change, and T is the fixed time interval. This mechanism ensures that DRL can be triggered periodically even if the state change is small, balancing the real-time requirements and computing overhead.

[0138] Based on the DRL optimization algorithm, the Actor-Critic framework of deep reinforcement learning (DRL) is adopted to solve the problem, and the TD3 algorithm is used to process the continuous action space.

[0139] The Actor-Critic framework is based on problem modeling, including state space, action space, reward function and optimization objectives.

[0140] The system state space S is defined as:

[0141] S = {location of vehicle and base station, current power allocation (P p , P B ), IRS configuration (Φ r , Φ u ,Ψ)}, (20) Action space To optimize the combination of variables, include:

[0142]

[0143] The reward function R aims to maximize the total rate R sum And satisfy the constraints:

[0144]

[0145] Among them: λ1,λ2 are constraint penalty coefficients.

[0146] The optimization goal is to maximize the cumulative reward:

[0147]

[0148] Where γ is the discount factor, Solve for the symbol for mathematical expectation.

[0149] TD3 algorithm process:

[0150] 1. Initialization:

[0151] Initialize the Actor network μ(s|θ μ ), dual critic network and And their target networks μ′ and Q'_1, Q'_2. Initialize the experience replay pool D. Initialize the noise parameter N (for noise in action selection).

[0152] 2. Start training each episode

[0153] For each episodei=1,2,...,max_episodes: initialize the environment state s0.

[0154] 3. Select an action and execute it

[0155] For each time step t=1,2,...,max_steps: according to the current state s t and Actor network select action: a t =μ(s t |θ μ )+N. Execute action a t , observe the reward R t and the new state st+1 .

[0156] Will (s t ,a t ,R t ,s t+1 ) is stored in the experience replay pool D.

[0157] 4. Sampling from the Experience Replay Pool

[0158] Randomly sample a batch of experience (s) from the experience replay pool D i ,a i ,R i ,s i+1 ).

[0159] 5. Add noise to the target action

[0160] Calculate the target action a′:

[0161] a′=μ′(s i+1 |θ μ′ )+clip(N(0,σ 2 ),-c,c).

[0162] The noise of the target action is generated by Gaussian noise N(0,σ 2 ), and cut off the noise to limit the amplitude of the noise.

[0163] 6. Update the Critic network

[0164] Calculate the target value y i :

[0165]

[0166] Calculate the loss function of the Critic network:

[0167]

[0168] Update Critic network parameters and To minimize the loss function L.

[0169] 7. Update Actor Network and Target Network

[0170] Every d, perform the following updates:

[0171] (1) Update the Actor network:

[0172]

[0173] (2) Update target network parameters:

[0174]

[0175] θ μ′ ←τθ μ +(1-τ)θ μ′

[0176] Among them, τ is a soft update parameter that ensures that the update of the target network is smooth.

[0177] 8. End the current episode

[0178] When the maximum number of steps T is completed, the current episode ends and the next training begins.

[0179] The main symbols used in the algorithm and their meanings are as follows:

[0180]

[0181] Simulation and result analysis:

[0182] Figure 2 and Figure 3 The changes of the normalized reward function and the normalized loss function are shown respectively. As can be seen from the figure, the TD3 algorithm based on the trigger mechanism performs best in terms of convergence speed, with the fastest decrease in loss value and the fastest increase in reward value; followed by the TD3 algorithm without a trigger mechanism; and the worst performance is the DRL method based on the Deep Deterministic Policy Gradient (DDPG). This is mainly because TD3 effectively alleviates the over-estimation problem through a dual Critic network and a delayed update strategy, while the trigger mechanism further reduces redundant calculations and improves learning efficiency. In contrast, DDPG is more susceptible to over-estimation and insufficient exploration in high-dimensional continuous control scenarios because it only uses one Critic network, resulting in its performance being inferior to TD3.

Claims

1. A model of a three-dimensional communication system for Internet of Vehicles based on the collaboration of Star-RIS and UAV-mounted RIS and its optimization method, characterized in that: The model includes: base station BS, multiple vehicle-mounted users V i , Star-shaped reconfigurable intelligent surface Star-RIS, multiple pairs of vehicle users V p and V p' , UAV-mounted RIS equipped with a reconfigurable smart surface, multiple V2I communication links between vehicle-mounted users Vi and the base station BS, and multiple V2V communication links between vehicle-user pairs, where i≥1, p≥1, p'≥1.

2. According to claim 1, a model of a three-dimensional communication system for Internet of Vehicles based on the collaboration of Star-RIS and UAV-mounted RIS, characterized in that: The V2I communication link performs signal forwarding and optimization through Star-RIS; the V2V communication link performs signal forwarding and enhancement through UAV-mounted RIS; the V2I communication link and the V2V communication link adopt orthogonal frequency division multiplexing (OFDM) technology; the V2I communication link and the V2V communication link share the same spectrum resources.

3. According to claim 1, a model of a three-dimensional communication system for Internet of Vehicles based on the collaboration of Star-RIS and UAV-mounted RIS, characterized in that: In the V2V communication link, the received signal y of the pth pair of V2V users p (t) is assisted by the reflected and transmitted signals from Star-RIS, and is also interfered by the reflected path of the V2I signal through the UAV-mounted RIS. The received signal model is: Explanation of symbols: y p (t): signal received by the receiving end of the Pth pair of V2V users; n p (t): The noise at the receiving end of the Pth pair of V2V users, obeying n p (t)~CN(0,σ 2 p ).

4. The model of the three-dimensional communication system of the Internet of Vehicles based on the collaboration of Star-RIS and UAV-mountedRIS according to claim 1, characterized in that: In the V2I communication link, the received signal y of the i-th vehicle user i (t) is assisted by the reflected signal from the UAV-mountedRIS and is also interfered by the reflection and transmission path of the V2V signal through Star-RIS. The received signal model is: Explanation of symbols: y i (t): signal received by the i-th vehicle user; n i (t): The noise at the receiving end of the i-th vehicle user, which obeys n i (t)~CN(0,σ i 2 ).

5. The model of the three-dimensional communication system of the Internet of Vehicles based on the collaboration of Star-RIS and UAV-mountedRIS according to claim 1, characterized in that: In the SINR and rate of the V2V communication link, the SINR of the Pth pair of V2V users is defined as follows: in: P p : The transmission power of the pth V2V user transmitting end; P B : The transmission power of the base station; Noise power at the receiving end of the pth pair of V2V users; According to Shannon's theorem, the rate R of the pth pair of V2V users V2V,P for: R V2V,p =B V2V log2(1+SINR V2V,p ), (4) Among them, B V2V Indicates the bandwidth allocated to the V2V link.

6. The model of the three-dimensional communication system of the Internet of Vehicles based on the collaboration of Star-RIS and UAV-mountedRIS according to claim 1, characterized in that: In the SINR and rate of the V2I communication link, the SINR of the i-th vehicle user is defined as follows: in: P B : The transmission power of the base station; P P : The transmission power of the Pth pair of V2V users; Noise power at the receiving end of the i-th vehicle user; According to Shannon's theorem, the rate R of the i-th vehicle user V2I,i for: R V2I,i =B V2I log2(1+SINR V2I,i ), (6) Among them, B V2I Indicates the bandwidth allocated to the V2I link.

7. The model of the three-dimensional communication system of the Internet of Vehicles based on the collaboration of Star-RIS and UAV-mountedRIS according to claim 1, characterized in that: The sum of the V2V link and V2I link rates is the total system rate R sum , 8. The optimization method of the model of the three-dimensional communication system of the Internet of Vehicles based on the collaboration of Star-RIS and UAV-mountedRIS according to claim 1 is characterized in that: The optimization method includes: (1) System total rate R sum Optimization, the total system rate R sum The optimization is to maximize R sum The goal is to ensure that the relevant constraints of the communication system are met and an intelligent trigger mechanism is introduced; (2) The DRL-based optimization algorithm adopts the Actor-Critic framework of deep reinforcement learning (DRL) and processes the continuous action space through the TD3 algorithm.

9. The optimization method of the model of the three-dimensional communication system of the Internet of Vehicles based on the collaboration of Star-RIS and UAV-mountedRIS according to claim 8, characterized in that: The constraints in step (1) include power constraints, reflection and transmission coefficient constraints, UAV-mountedRIS reflection coefficient constraints, SINR minimum constraints and bandwidth allocation constraints.

10. The method for optimizing the model of the three-dimensional communication system of the Internet of Vehicles based on the collaboration of Star-RIS and UAV-mountedRIS according to claim 8, characterized in that: The Actor-Critic framework is based on problem modeling, including state space, action space, reward function and optimization objectives.

Images