Macro-micro base station cooperative sensing system and method
By combining deep reinforcement learning and multi-source information cross-validation, the problems of low resource allocation efficiency and insufficient security in the integrated communication and sensing of macro and micro base stations are solved, and the intelligent resource management and sensing performance of the system are realized.
Patent Information
- Application Number
- CN202511652067.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-12
- Publication Date
- 2026-02-06
AI Technical Summary
Existing technologies suffer from low resource allocation efficiency, lack of effective security protection, and a single collaborative mechanism in the integration of macro and micro base station communication and sensing, resulting in limited system performance and severe interference.
By employing a dynamic resource allocation mechanism based on deep reinforcement learning and a security module that cross-validates multi-source information, combined with an intelligent collaborative strategy, we can achieve real-time optimization of resources such as power, spectrum, and beam, as well as the identification and defense against spoofing attacks.
It improves resource utilization efficiency, enhances the credibility and security of sensing data, and achieves synergistic improvement in communication and sensing performance, adapting to different task requirements.
Smart Images
Figure CN121486832A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of mobile communication technology, and in particular to a system and method for collaborative sensing between macro and micro base stations. Background Technology
[0002] With the large-scale commercialization of 5G mobile communication technology and the in-depth research on 6G mobile communication technology, integrated communication and sensing has become an important evolution direction for future wireless networks. Macro and micro base station heterogeneous networking, as the basic architecture of mobile communication networks, provides an effective solution for balancing communication coverage and capacity. Macro base stations are responsible for wide-area coverage, while micro base stations focus on high-capacity services in hotspot areas. Integrating sensing functions into this type of communication network enables base stations not only to provide communication services but also to perceive the surrounding environment through the analysis of wireless signals, providing fundamental support for applications such as smart transportation and the Industrial Internet of Things.
[0003] However, existing technologies face numerous challenges in achieving integrated communication and sensing. Traditional resource allocation schemes often employ static or simple rule-based methods, making it difficult to adapt to dynamic changes in network traffic and sensing demands, resulting in low resource utilization efficiency. Simultaneously, existing systems lack effective security mechanisms and insufficient verification of the reliability of sensing data, making them vulnerable to spoofing attacks. Regarding coordination mechanisms, macro base stations and micro base stations often adopt fixed coordination modes, failing to flexibly adjust their operating strategies according to different sensing task requirements, thus hindering further improvements in sensing performance. Furthermore, communication and sensing functions typically operate independently, leading to resource contention and mutual interference, making it difficult to optimize overall system performance.
[0004] To address the aforementioned problems, this invention proposes an intelligent macro-micro base station collaborative sensing system and method. By employing a dynamic resource allocation mechanism based on deep reinforcement learning, real-time optimization of resources such as power, spectrum, and beamwidth is achieved. An innovative security module with multi-source information cross-validation is introduced to effectively identify and resist spoofing attacks. The system can also intelligently select the optimal collaborative mode based on the characteristics of the sensing task, fully leveraging the complementary advantages of macro and micro base stations. Finally, through integrated communication and sensing design, a unified resource optimization framework is constructed, achieving a synergistic improvement in communication capabilities and sensing performance. Summary of the Invention
[0005] To overcome the problems mentioned in the background art, the present invention proposes a system and method for collaborative sensing of macro and micro base stations.
[0006] The technical solution of this invention is: a system for collaborative sensing between macro and micro base stations, comprising: Macro base station clusters are responsible for wide-area communication coverage and wide-area coarse-grained environmental perception; Micro base station clusters are deployed in hotspot areas within the coverage area of macro base station clusters, responsible for providing high-capacity data services and high-precision local environment perception; The collaborative control and processing unit communicates with the macro base station cluster and the micro base station cluster. It is used to schedule the sensing tasks of macro base stations and micro base stations, and to fuse the sensing data from macro base stations and micro base stations to generate environmental sensing information.
[0007] Preferably, the collaborative control and processing unit includes: A11: Dynamic resource allocation module, used to dynamically allocate sensing resources based on sensing task requirements and network status; A12: Perception security module, used to identify and defend against deception attacks on the perception system through cross-verification of multi-source information.
[0008] Preferably, the dynamic resource allocation module employs a deep reinforcement learning model. Specifically, the dynamic resource allocation module, during its operation, includes: S11: State observation step, real-time observation and acquisition of network state information and perception task requirements, which together constitute the state vector of the deep reinforcement learning model at time t. S12: Decision generation step, input the state vector into the trained deep reinforcement learning policy network, and the policy network outputs the action to allocate communication and sensing resources to macro base stations and micro base stations in the current state. The action includes the allocation scheme of power, spectrum, beam and time resources. S13: Action execution steps, execute the actions output by the policy network, and send the resource allocation instructions to the corresponding macro base stations and micro base stations; S14: Benefit evaluation step: After the action is performed, observe the environmental feedback, calculate the immediate reward, and enter a new state; S15: Model update step, using the obtained empirical tuples to update the parameters of the deep reinforcement learning model and optimize the strategy.
[0009] Preferably, after the action is performed, when observing environmental feedback and calculating the immediate reward, the immediate reward is calculated using the following formula: ; in, This indicates that at time step t, the action was performed. Then, the immediate rewards obtained from the environment, , , and The weighting coefficients are used to balance different optimization objectives. Indicates the system is in action The overall communication throughput benefit, Indicates the system is in action The overall perception accuracy benefits. A penalty term characterizing the communication-sensing or sensing-sensing interference that may occur due to resource allocation. A penalty term characterizing the total power consumption cost of the system. The communication service load status at time step t. For the resource allocation action selected by the agent at time step t, To perceive the task requirements and priority status at time step t.
[0010] As a preferred approach, the deep reinforcement learning model employs an actor-critic architecture, comprising an actor network that generates decision actions and a critic network that evaluates state values. During the model update step, the parameters of both networks are updated simultaneously. When updating the parameters of the actor network, the objective is to maximize the expected cumulative reward in the future; when updating the parameters of the critic network, the objective is to minimize the error in its prediction of state values.
[0011] As a preferred embodiment, when the perception security module identifies and defends against deception attacks on the perception system through cross-verification of multi-source information, it specifically includes: S21: Receive sensing observation data of the same potential target from at least one macro base station and multiple micro base stations, including the target position, radial velocity and radar cross section characteristics estimated by each base station; S22: Correlate and register the observation data in a unified spatiotemporal coordinate system; S23: Calculate multiple consistency metrics; S24: Construct a residual vector containing multiple consistent residuals and calculate the Mahalanobis distance; S25: Compare the calculated Mahalanobis distance with the dynamic decision threshold. If the distance exceeds the threshold, a deception attack is determined and a security alert is generated.
[0012] Preferably, when calculating multiple consistency metrics, the following are included: A21: Location consistency residuals between the observed locations of each base station and the optimal estimated location after fusion: ; in, Let i be the target plane coordinates independently observed by the i-th base station. The optimal estimated coordinates of the target are calculated by fusing observation data from all base stations using a weighted least squares algorithm. A22: Velocity consistency residuals between the radial velocities observed at each base station and the theoretical radial velocities derived from the fused trajectories: ; in, Let be the radial velocity of the target observed by the i-th base station based on echo Doppler frequency shift. This is the theoretical value of the target's radial velocity relative to the corresponding base station, derived from the motion trajectory based on the optimal estimated coordinates. A23: Characteristic consistency residuals between the radar cross section values observed at each base station and their average values: ; in, The target RCS value observed and estimated by the i-th base station. This is the average of the RCS values observed at all base stations.
[0013] Preferably, the collaborative control and processing unit, when in operation, specifically includes: S31: Receive a perception service request from an upper-layer application or network management unit, and parse the perception service request to determine the perception target area, required accuracy, update frequency, and task priority; S32: Determine the collaboration mode based on task requirements and network status, and select macro base stations and micro base station clusters to participate in the collaboration; S33: Dynamically generate resource allocation schemes based on deep reinforcement learning models and send perception instructions to the base station cluster; S34: Receive sensing data from macro base station clusters and micro base station clusters, including preprocessed target observation information, raw channel state information and echo signals; S35: Perform time synchronization and spatial registration on the aggregated multi-source sensing data, and process it using a fusion algorithm to generate an environmental sensing map containing semantic information about the target's location, velocity, trajectory, and category. S36: Feed the generated perception information back to the upper layer application, and dynamically adjust the coordination strategy and resource allocation parameters based on the quality assessment of the perception results and changes in network status to optimize the performance of subsequent perception tasks.
[0014] As a preferred option, when determining the collaboration mode based on task requirements and network status, and selecting the macro base stations and micro base station clusters to participate in the collaboration, the decision-making collaboration mode includes: A31: If the perception task is a large-scale target search and coarse localization, the decision-making adopts a guided collaborative mode, designating a macro base station to perform wide-area scanning; A32: If the sensing task is to perform high-precision imaging or tracking of a specific area, the decision-making adopts a joint and collaborative mode, scheduling multiple micro base stations and macro base stations to perform multi-angle sensing simultaneously.
[0015] A method for collaborative sensing between macro and micro base stations includes the following steps: S41: Receive and parse the perception request from the upper layer application to clarify the specific requirements and objectives of the task; S42: Based on task requirements and the current network status, determine the optimal cooperative sensing mode and select participating base stations; S43: Dynamically generate resource allocation schemes using deep reinforcement learning models and send execution instructions to the base station; S44: Collect sensing data from macro base station clusters and micro base station clusters; S45: Perform fusion processing on the converged sensing data to generate an environmental sensing map containing target trajectory and semantic information; S46: Identify and defend against perceptual deception attacks through cross-validation of multi-source information; S47: Feed back the final environmental perception information to the upper-layer application; S48: Based on the quality assessment of the perception results and network changes, dynamically adjust the strategy and parameters to optimize subsequent performance.
[0016] The beneficial effects of this invention are: 1. Compared with existing technologies that use fixed or simple rule-driven resource allocation schemes, which are difficult to cope with dynamically changing network environments and diverse perception task requirements, resulting in low resource utilization efficiency and limited system performance, this solution adopts an intelligent dynamic resource allocation mechanism based on deep reinforcement learning. It can autonomously perceive changes in network status and task requirements, and optimize the global configuration of power, spectrum, beam and time resources in real time, significantly improving resource utilization efficiency and system adaptability. 2. Compared with the lack of effective perception security protection mechanisms in existing technologies, the system is susceptible to deception attacks, which can lead to distortion of perception information and seriously affect the reliability of the system. This solution innovatively introduces a security module that cross-verifies multi-source information. By comprehensively evaluating the consistency indicators of multiple dimensions such as position, speed and physical characteristics, it can accurately identify and resist false targets and deception signals, which greatly enhances the credibility and security of the perception data in adversarial environments. 3. Compared with the existing technology where the macro and micro base station collaboration mechanism is singular and cannot flexibly adapt to the optimal collaboration mode according to the characteristics of the task, resulting in difficulty in achieving both sensing coverage and accuracy, this solution proposes an intelligent collaboration strategy decision mechanism. It dynamically selects guided collaboration or joint collaboration mode according to the scope and accuracy requirements of the sensing task, giving full play to the complementary advantages of the wide-area coverage of macro base stations and the high-precision sensing of micro base stations, and achieving the optimization of sensing performance in different application scenarios. 4. Compared with the existing technology that adopts isolated communication and sensing functions, which leads to resource competition and mutual interference between the two and makes it difficult to achieve optimal overall performance, this solution adopts an integrated communication and sensing design, constructs a unified resource optimization framework and collaborative processing mechanism, and uses deep reinforcement learning to achieve joint dynamic allocation of communication and sensing resources, effectively suppressing mutual interference, and ultimately maximizing the overall benefits of synergistic improvement in communication capacity and sensing accuracy. Attached Figure Description
[0017] Figure 1 The diagram shown is a structural schematic of the macro-micro base station collaborative sensing system of the present invention. Figure 2 The diagram shown is a flowchart illustrating the macro-micro base station collaborative sensing method of the present invention. Detailed Implementation
[0018] The present invention will be further described below with reference to the accompanying drawings and embodiments.
[0019] Please see Figures 1-2 The present invention provides an embodiment: a system for collaborative sensing between macro and micro base stations, comprising: Macro base station clusters are responsible for wide-area communication coverage and wide-area coarse-grained environmental perception; Micro base station clusters are deployed in hotspot areas within the coverage area of macro base station clusters, responsible for providing high-capacity data services and high-precision local environment perception; The collaborative control and processing unit communicates with the macro base station cluster and the micro base station cluster. It is used to schedule the sensing tasks of macro base stations and micro base stations, and to fuse the sensing data from macro base stations and micro base stations to generate environmental sensing information.
[0020] In this embodiment, the present invention constructs a heterogeneous collaborative network architecture of macro base station clusters and micro base station clusters. Macro base stations are used to achieve wide-area communication coverage and coarse-grained environmental perception, while micro base stations provide high-capacity data services and high-precision local perception. This achieves deep integration of communication and perception functions at the network architecture level, significantly improving the system's coverage, service capacity, and multi-dimensional perception capabilities.
[0021] Preferably, the collaborative control and processing unit includes: A11: Dynamic resource allocation module, used to dynamically allocate sensing resources based on sensing task requirements and network status; A12: Perception security module, used to identify and defend against deception attacks on the perception system through cross-verification of multi-source information.
[0022] In this embodiment, the present invention integrates a dynamic resource allocation module and a perception security module into the collaborative control and processing unit, thereby realizing intelligent resource scheduling for perception tasks and secure verification of multi-source information. This effectively improves the system's resource utilization efficiency while ensuring the reliability and security of perception data.
[0023] Preferably, the dynamic resource allocation module employs a deep reinforcement learning model. Specifically, the dynamic resource allocation module, during its operation, includes: S11: State observation step, real-time observation and acquisition of network state information and perception task requirements, which together constitute the state vector of the deep reinforcement learning model at time t. S12: Decision generation step, input the state vector into the trained deep reinforcement learning policy network, and the policy network outputs the action to allocate communication and sensing resources to macro base stations and micro base stations in the current state. The action includes the allocation scheme of power, spectrum, beam and time resources. S13: Action execution steps, execute the actions output by the policy network, and send the resource allocation instructions to the corresponding macro base stations and micro base stations; S14: Benefit evaluation step: After the action is performed, observe the environmental feedback, calculate the immediate reward, and enter a new state; S15: Model update step, using the obtained empirical tuples to update the parameters of the deep reinforcement learning model and optimize the strategy.
[0024] In this embodiment, the present invention uses a deep reinforcement learning model to achieve dynamic resource allocation. Through a closed-loop optimization process of state observation, decision generation, action execution, benefit evaluation and model update, the system can autonomously adapt to changes in network state and fluctuations in task requirements, significantly improving the intelligence level of resource allocation and the overall performance of the system.
[0025] Preferably, after the action is performed, when observing environmental feedback and calculating the immediate reward, the immediate reward is calculated using the following formula: ; in, This indicates that at time step t, the action was performed. Then, the immediate rewards obtained from the environment, , , and The weighting coefficients are used to balance different optimization objectives. Indicates the system is in action The overall communication throughput benefit, Indicates the system is in action The overall perception accuracy benefits. A penalty term characterizing the communication-sensing or sensing-sensing interference that may occur due to resource allocation. A penalty term characterizing the total power consumption cost of the system. The communication service load status at time step t. For the resource allocation action selected by the agent at time step t, To perceive the task requirements and priority status at time step t.
[0026] In this embodiment, the present invention designs an instant reward function based on multi-objective optimization. By weighted combination of communication throughput benefits, perception accuracy benefits, interference penalties and power consumption cost penalties, it achieves dynamic balance and joint optimization of communication performance and perception performance, ensuring that the system can maximize global benefits in different scenarios.
[0027] As a preferred approach, the deep reinforcement learning model employs an actor-critic architecture, comprising an actor network that generates decision actions and a critic network that evaluates state values. During the model update step, the parameters of both networks are updated simultaneously. When updating the parameters of the actor network, the objective is to maximize the expected cumulative reward in the future; when updating the parameters of the critic network, the objective is to minimize the error in its prediction of state values.
[0028] In this embodiment, the present invention employs a deep reinforcement learning model with an actor-critic architecture. Through a collaborative mechanism where the actor network generates decision actions and the critic network evaluates state values, the parallel execution of strategy optimization and value evaluation is achieved, significantly improving the model's convergence speed and decision stability.
[0029] As a preferred embodiment, when the perception security module identifies and defends against deception attacks on the perception system through cross-verification of multi-source information, it specifically includes: S21: Receive sensing observation data of the same potential target from at least one macro base station and multiple micro base stations, including the target position, radial velocity and radar cross section characteristics estimated by each base station; S22: Correlate and register the observation data in a unified spatiotemporal coordinate system; S23: Calculate multiple consistency metrics; S24: Construct a residual vector containing multiple consistent residuals and calculate the Mahalanobis distance; S25: Compare the calculated Mahalanobis distance with the dynamic decision threshold. If the distance exceeds the threshold, a deception attack is determined and a security alert is generated.
[0030] In this embodiment, the present invention utilizes the multi-source information cross-verification mechanism of the perception security module to perform consistency verification using observation data from multiple base stations, effectively identifying and resisting deception attacks, and significantly improving the reliability and security of the system in adversarial environments.
[0031] Preferably, when calculating multiple consistency metrics, the following are included: A21: Location consistency residuals between the observed locations of each base station and the optimal estimated location after fusion: ; in, Let i be the target plane coordinates independently observed by the i-th base station. The optimal estimated coordinates of the target are calculated by fusing observation data from all base stations using a weighted least squares algorithm. A22: Velocity consistency residuals between the radial velocities observed at each base station and the theoretical radial velocities derived from the fused trajectories: ; in, Let be the radial velocity of the target observed by the i-th base station based on echo Doppler frequency shift. This is the theoretical value of the target's radial velocity relative to the corresponding base station, derived from the motion trajectory based on the optimal estimated coordinates. A23: Characteristic consistency residuals between the radar cross section values observed at each base station and their average values: ; in, The target RCS value observed and estimated by the i-th base station. This is the average of the RCS values observed at all base stations.
[0032] In this embodiment, the present invention constructs a comprehensive perception data verification system by calculating multi-dimensional metrics such as position consistency residual, velocity consistency residual, and feature consistency residual. This system can identify abnormal data from multiple levels, including spatial location, motion characteristics, and physical properties, thereby improving the accuracy and robustness of deception attack detection.
[0033] Preferably, the collaborative control and processing unit, when in operation, specifically includes: S31: Receive a perception service request from an upper-layer application or network management unit, and parse the perception service request to determine the perception target area, required accuracy, update frequency, and task priority; S32: Determine the collaboration mode based on task requirements and network status, and select macro base stations and micro base station clusters to participate in the collaboration; S33: Dynamically generate resource allocation schemes based on deep reinforcement learning models and send perception instructions to the base station cluster; S34: Receive sensing data from macro base station clusters and micro base station clusters, including preprocessed target observation information, raw channel state information and echo signals; S35: Perform time synchronization and spatial registration on the aggregated multi-source sensing data, and process it using a fusion algorithm to generate an environmental sensing map containing semantic information about the target's location, velocity, trajectory, and category. S36: Feed the generated perception information back to the upper layer application, and dynamically adjust the coordination strategy and resource allocation parameters based on the quality assessment of the perception results and changes in network status to optimize the performance of subsequent perception tasks.
[0034] In this embodiment, the present invention achieves full lifecycle management of sensing tasks through the complete workflow of the collaborative control and processing unit, from task reception and parsing, collaborative decision-making, resource scheduling, data aggregation to fusion processing, ensuring the efficiency of system collaborative work and the accuracy of sensing information generation.
[0035] As a preferred option, when determining the collaboration mode based on task requirements and network status, and selecting the macro base stations and micro base station clusters to participate in the collaboration, the decision-making collaboration mode includes: A31: If the perception task is a large-scale target search and coarse localization, the decision-making adopts a guided collaborative mode, designating a macro base station to perform wide-area scanning; A32: If the sensing task is to perform high-precision imaging or tracking of a specific area, the decision-making adopts a joint and collaborative mode, scheduling multiple micro base stations and macro base stations to perform multi-angle sensing simultaneously.
[0036] In this embodiment, the present invention intelligently selects guided or joint collaborative modes according to the different characteristics of the sensing task, giving full play to the wide-area coverage advantage of macro base stations and the high-precision sensing capability of micro base stations, thereby achieving optimal configuration of sensing resources and maximization of collaborative efficiency in different scenarios.
[0037] A method for collaborative sensing between macro and micro base stations includes the following steps: S41: Receive and parse the perception request from the upper layer application to clarify the specific requirements and objectives of the task; S42: Based on task requirements and the current network status, determine the optimal cooperative sensing mode and select participating base stations; S43: Dynamically generate resource allocation schemes using deep reinforcement learning models and send execution instructions to the base station; S44: Collect sensing data from macro base station clusters and micro base station clusters; S45: Perform fusion processing on the converged sensing data to generate an environmental sensing map containing target trajectory and semantic information; S46: Identify and defend against perceptual deception attacks through cross-validation of multi-source information; S47: Feed back the final environmental perception information to the upper-layer application; S48: Based on the quality assessment of the perception results and network changes, dynamically adjust the strategy and parameters to optimize subsequent performance.
[0038] In this embodiment, the present invention achieves intelligent processing of the entire process from task reception to parameter optimization through a complete macro-micro base station collaborative sensing method, which significantly improves the accuracy, real-time performance and reliability of environmental sensing, and provides an efficient and feasible technical solution for future integrated communication and sensing networks.
[0039] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.
Claims
1. A system for collaborative sensing between macro and micro base stations, characterized in that: include: Macro base station clusters are responsible for wide-area communication coverage and wide-area coarse-grained environmental perception; Micro base station clusters are deployed in hotspot areas within the coverage area of macro base station clusters, responsible for providing high-capacity data services and high-precision local environment perception; The collaborative control and processing unit communicates with the macro base station cluster and the micro base station cluster. It is used to schedule the sensing tasks of macro base stations and micro base stations, and to fuse the sensing data from macro base stations and micro base stations to generate environmental sensing information.
2. The system for collaborative sensing between macro and micro base stations according to claim 1, characterized in that: The collaborative control and processing unit includes: A11: Dynamic resource allocation module, used to dynamically allocate sensing resources based on sensing task requirements and network status; A12: Perception security module, used to identify and defend against deception attacks on the perception system through cross-verification of multi-source information.
3. The macro-micro base station collaborative sensing system according to claim 2, characterized in that: The dynamic resource allocation module employs a deep reinforcement learning model. Specifically, when the dynamic resource allocation module is working, it includes: S11: State observation step, observe and acquire network state information and perception task requirements in real time, and form the state vector of the deep reinforcement learning model at time t. S12: Decision generation step, input the state vector into the trained deep reinforcement learning policy network, and the policy network outputs the action to allocate communication and sensing resources to macro base stations and micro base stations in the current state. The action includes the allocation scheme of power, spectrum, beam and time resources. S13: Action execution steps, execute the actions output by the policy network, and send the resource allocation instructions to the corresponding macro base stations and micro base stations; S14: Benefit evaluation step: After the action is performed, observe the environmental feedback, calculate the immediate reward, and enter a new state; S15: Model update step, using the obtained empirical tuples to update the parameters of the deep reinforcement learning model and optimize the strategy.
4. A macro-micro base station collaborative sensing system according to claim 3, characterized in that: After the action is executed, environmental feedback is observed, and the immediate reward is calculated using the following formula: ； in, This indicates that at time step t, the action was performed. Then, the immediate rewards obtained from the environment, , , and The weighting coefficients are used to balance different optimization objectives. Indicates the system is in action The overall communication throughput benefit, Indicates the system is in action The overall perception accuracy benefits. A penalty term characterizing the communication-sensing or sensing-sensing interference that may occur due to resource allocation. A penalty term characterizing the total power consumption cost of the system. The communication service load status at time step t. For the resource allocation action selected by the agent at time step t, To perceive the task requirements and priority status at time step t.
5. A macro-micro base station collaborative sensing system according to claim 4, characterized in that: The deep reinforcement learning model adopts an actor-critic architecture, which includes an actor network that generates decision actions and a critic network that evaluates the value of the state. In the model update step, the parameters of both networks are updated simultaneously. When updating the parameters of the actor network, the goal is to maximize the expected cumulative reward in the future. When updating the parameters of the critic network, the goal is to minimize its error in predicting the state value.
6. A macro-micro base station collaborative sensing system according to claim 5, characterized in that: When the perception security module identifies and defends against deception attacks on the perception system through cross-verification of multi-source information, it specifically includes: S21: Receive sensing and observation data of the same potential target from at least one macro base station and multiple micro base stations; S22: Correlate and register the observation data in a unified spatiotemporal coordinate system; S23: Calculate multiple consistency metrics; S24: Construct a residual vector containing multiple consistent residuals and calculate the Mahalanobis distance; S25: Compare the calculated Mahalanobis distance with the dynamic decision threshold. If the distance exceeds the threshold, a deception attack is determined and a security alert is generated.
7. A macro-micro base station collaborative sensing system according to claim 6, characterized in that: When calculating multiple consistency metrics, including: A21: Location consistency residuals between the observed locations of each base station and the optimal estimated locations after fusion; A22: Velocity consistency residuals between the radial velocities observed at each base station and the theoretical radial velocities derived from the fused trajectories; A23: Characteristic consistency residuals between the radar cross section values observed at each base station and their average values.
8. A macro-micro base station collaborative sensing system according to claim 7, characterized in that: When the collaborative control and processing unit is in operation, it specifically includes: S31: Receive a perception service request from an upper-layer application or network management unit, and parse the perception service request to determine the perception target area, required accuracy, update frequency, and task priority; S32: Determine the collaboration mode based on task requirements and network status, and select macro base stations and micro base station clusters to participate in the collaboration; S33: Dynamically generate resource allocation schemes based on deep reinforcement learning models and send perception instructions to the base station cluster; S34: Receive sensing data from macro base station clusters and micro base station clusters, including preprocessed target observation information, raw channel state information and echo signals; S35: Perform time synchronization and spatial registration on the aggregated multi-source sensing data, and process it using a fusion algorithm; S36: Feed the generated perception information back to the upper layer application, and dynamically adjust the coordination strategy and resource allocation parameters based on the quality assessment of the perception results and changes in network status.
9. A macro-micro base station collaborative sensing system according to claim 8, characterized in that: When determining the collaboration mode based on task requirements and network status, and selecting macro base stations and micro base station clusters to participate in the collaboration, the decision-making collaboration modes include: A31: If the perception task is a large-scale target search and coarse localization, the decision-making adopts a guided collaborative mode, designating a macro base station to perform wide-area scanning; A32: If the sensing task is to perform high-precision imaging or tracking of a specific area, the decision-making adopts a joint and collaborative mode, scheduling multiple micro base stations and macro base stations to perform multi-angle sensing simultaneously.
10. A method for collaborative sensing between macro and micro base stations, characterized in that: Includes the following steps: S41: Receive and parse the perception request from the upper layer application to clarify the specific requirements and objectives of the task; S42: Based on task requirements and the current network status, determine the optimal cooperative sensing mode and select participating base stations; S43: Dynamically generate resource allocation schemes using deep reinforcement learning models and send execution instructions to the base station; S44: Collect sensing data from macro base station clusters and micro base station clusters; S45: Perform fusion processing on the converged sensing data to generate an environmental sensing map containing target trajectory and semantic information; S46: Identify and defend against perceptual deception attacks through cross-validation of multi-source information; S47: Feed back the final environmental perception information to the upper-layer application; S48: Based on the quality assessment of the perception results and network changes, dynamically adjust the strategy and parameters to optimize subsequent performance.