Random game control method for unmanned aerial vehicle attitude control

Abstract

The invention relates to the technical field of unmanned aerial vehicle autonomous control, and discloses a random game control method for unmanned aerial vehicle attitude control. The method comprises the following steps: constructing a cooperative control system comprising an attitude sensing unit and a game decision-making unit, collecting real-time motion parameters of an unmanned aerial vehicle, and monitoring environment confrontation behaviors; dynamically evaluating the motion parameters and the confrontation behaviors to generate a situation result, and formulating a confrontation strategy according to the situation result; activating a virtual countermeasure entity according to the strategy, wherein the entity simulates a real countermeasure behavior pattern and generates a countermeasure signal; establishing a distributed communication network formed by a plurality of relay nodes, and dynamically adjusting the transmission parameters of the relay nodes according to the characteristics of the adversarial signals; finally, a control instruction is transmitted through the adjusted relay node, and closed-loop control over the attitude of the unmanned aerial vehicle is achieved. According to the invention, by introducing virtual confrontation simulation and communication network dynamic optimization, the intelligent decision-making capability of the unmanned aerial vehicle in a complex confrontation environment and the robustness of a control system are improved.

Description

Technical Field

[0001] This invention relates to the field of autonomous control technology for unmanned aerial vehicles (UAVs), specifically a stochastic game control method for attitude control of UAVs. Background Technology

[0002] Existing UAV attitude control technologies primarily rely on precise sensor feedback and pre-set control algorithms. These methods perform well in stable, predictable, and non-adversarial environments. However, when UAVs encounter intelligent adversarial behavior during missions, the effectiveness of their control systems drops sharply. Traditional control models lack the ability to perceive and make decisions about such proactive, intelligent adversarial factors online, failing to translate the adversarial situation into processable information within the control loop. This results in UAVs reacting slowly and employing limited strategies in adversarial environments, making them highly vulnerable to suppression or capture.

[0003] At the communication level, existing UAV systems mostly employ static or semi-static communication link configurations. Even with multi-relay communication networks, the transmission parameters of relay nodes are typically pre-set or only subject to limited adjustments based on conventional indicators such as channel quality. This communication architecture struggles to cope with the rapid, dynamic, and targeted communication interference induced by intelligent adversarial behavior. Adversaries can easily detect fixed communication patterns and implement precise jamming, thereby cutting off or severely degrading the transmission of control commands to the UAV, ultimately leading to loss of attitude control. The decision-making isolation of the control system and the static nature of the communication system are the main bottlenecks faced by existing technologies in adversarial environments. Summary of the Invention

[0004] The purpose of this invention is to provide a stochastic game control method for attitude control of unmanned aerial vehicles (UAVs) to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a stochastic game control method for attitude control of unmanned aerial vehicles (UAVs), the method comprising:

[0006] Construct a collaborative control system, which includes an attitude perception unit and a game decision unit. The attitude perception unit collects real-time motion parameters of the UAV, and the game decision unit monitors adversarial behavior in the environment.

[0007] The real-time motion parameters and the adversarial behavior are dynamically evaluated to generate a situation assessment result, and an adversarial strategy is formulated based on the situation assessment result.

[0008] The virtual adversarial entity is activated according to the adversarial strategy. The virtual adversarial entity simulates the behavior pattern of the real adversary and generates adversarial signals.

[0009] establishing a distributed communication network composed of a plurality of relay nodes, adjusting transmission parameters of the plurality of relay nodes according to characteristics of the confrontation signal;

[0010] delivering control instructions through the plurality of relay nodes after adjustment to complete closed-loop control of the attitude of the UAV.

[0011] Preferably, the real-time motion parameters and the confrontation behavior are dynamically evaluated to generate a situation assessment result, including:

[0012] Inertial measurement data streams are continuously acquired through the attitude perception unit, and confrontation behavior data streams are collected through the game decision unit;

[0013] The inertial measurement data streams and the confrontation behavior data streams are time-aligned to form a synchronized data sequence;

[0014] Statistical features and time-domain features of the synchronized data sequence are extracted to form a multi-dimensional feature vector;

[0015] A pattern recognition method is used to classify the multi-dimensional feature vector to output a current situation level;

[0016] A risk assessment coefficient is calculated based on the current situation level, and historical situation data is combined to generate the situation assessment result.

[0017] Preferably, the confrontation strategy is formulated according to the situation assessment result, including:

[0018] The threat level index and the uncertainty index in the situation assessment result are analyzed;

[0019] A strategy decision matrix is established, which contains a plurality of preset coping schemes;

[0020] The threat level index and the uncertainty index are matched and selected in the strategy decision matrix to determine a basic coping scheme;

[0021] The basic coping scheme is adaptively corrected in combination with real-time environmental parameters to generate the confrontation strategy.

[0022] Preferably, the virtual confrontation entity is activated according to the confrontation strategy, including:

[0023] A virtual entity database is configured, which stores a plurality of confrontation entity models;

[0024] According to the confrontation intensity requirement in the confrontation strategy, a corresponding confrontation entity model is selected from the virtual entity database;

[0025] Initialize the physical parameters and behavioral logic of the adversarial entity model to give it characteristics similar to those of a real adversary.

[0026] Activate the signal generation module of the adversarial entity model to prepare for generating the adversarial signal.

[0027] Preferably, the virtual adversarial entity simulates the behavior patterns of a real adversary and generates adversarial signals, including:

[0028] Receive behavioral pattern instructions from the adversarial strategy and parse the action sequence in the behavioral pattern instructions;

[0029] Simulate the decision-making process of a real adversary and dynamically adjust the sequence of actions based on environmental feedback;

[0030] Multiple types of interference signals are generated by a multi-channel signal generator according to the adjusted action sequence;

[0031] The various types of interference signals are combined and encoded to form the complete countermeasure signal.

[0032] Preferably, establishing a distributed communication network includes:

[0033] Survey the terrain features and communication environment of the drone operation area, and plan the network coverage area;

[0034] Calculate the minimum number of relay nodes required and their optimal distribution locations based on the network coverage area;

[0035] Relay devices with adaptive frequency modulation capabilities are deployed at the optimal distribution locations;

[0036] Configure the communication protocol and data forwarding rules between the relay devices to form the distributed communication network.

[0037] Preferably, adjusting the transmission parameters of the plurality of relay nodes according to the characteristics of the countermeasure signal includes:

[0038] Monitor the spectral characteristics and power fluctuation characteristics of the countermeasure signal;

[0039] Analyze the impact of the countermeasures on the communication channel and evaluate the channel quality indicators;

[0040] The transmit power and modulation scheme are dynamically adjusted based on the aforementioned channel quality indicators;

[0041] The optimal communication frequency band is adaptively selected, and the relevant parameters of the data packet retransmission mechanism are updated.

[0042] Preferably, the transmission of control commands through the adjusted plurality of relay nodes includes:

[0043] The unmanned aerial vehicle attitude control instruction is divided into multiple data blocks, and a sequence identifier is added to each data block;

[0044] According to the real-time load conditions of the multiple relay nodes, the transmission paths of the data blocks are allocated;

[0045] Redundant transmission channels are established among the multiple relay nodes to ensure reliable transmission of the data blocks;

[0046] The data blocks are recombined and checked at the receiving end to restore the complete control instruction.

[0047] Preferably, the closed-loop control of the unmanned aerial vehicle attitude is completed, comprising:

[0048] The recombined control instruction is sent to the unmanned aerial vehicle flight control system;

[0049] The unmanned aerial vehicle attitude response data are collected in real time through a sensor array;

[0050] The deviation value of the attitude response data from the expected control target is compared;

[0051] According to the deviation value, the parameter settings of the subsequent control instruction are adjusted to realize closed-loop feedback control.

[0052] Preferably, the virtual entity database is configured, comprising:

[0053] The historical confrontation data are collected, and the historical confrontation data include the action trajectory of the confrontation party, the communication interaction record and the environmental parameter;

[0054] The collected historical confrontation data are subjected to integrity check, invalid data segments are removed, and missing data points are supplemented;

[0055] The processed historical confrontation data are subjected to pattern mining by using a clustering analysis method to identify typical confrontation behavior categories;

[0056] For each typical confrontation behavior category, a confrontation entity model is constructed, and each confrontation entity model includes a behavior decision tree, a state transition probability and an action response function;

[0057] The constructed confrontation entity model is imported into a database management system, and a multi-level index structure is established according to the behavior category and the model complexity;

[0058] A database update trigger is set, and when the newly added confrontation data reach a threshold value, a model optimization process is automatically started.

[0059] Compared with the prior art, the present application has the following advantages:

[0060] The virtual counter entity is activated by the game decision unit. The virtual counter entity is not a preset static interference source, but can simulate the behavior mode of a real counter intelligent entity and generate a corresponding counter signal according to a real-time situation evaluation result. This mechanism converts the counter environment from an uncontrollable external factor into a manageable and interactive simulated element inside the control system. The control algorithm learns and adapts in the continuous game process with the virtual entity, so that the unmanned aerial vehicle attitude control strategy has foresight and adaptability. The control system is no longer a passive response to the deviation of the historical state, but can generate and practice coping strategies for potential counter threats, thereby showing stronger prediction ability and strategy flexibility when facing real counter.

[0061] A distributed communication network composed of multiple relay nodes is established, and its transmission parameters can be dynamically adjusted according to the characteristics of the signals generated by the virtual counter entity. The characteristics of the counter signal are regarded as an indirect reflection of the counter behavior mode, and the communication network optimizes the transmission strategy in real time accordingly. When the virtual entity simulates a certain frequency band interference, the communication network can guide the data flow to switch to a frequency band or node that is not interfered; when simulating high-power suppression, the network can adaptively adjust the transmission power or use frequency hopping, spread spectrum and other anti-interference technologies. This deep coupling makes the communication link no longer a fragile and relatively independent link in the control loop, but evolves into a dynamic defense layer that actively participates in the counter, improving the transmission resilience and reliability of control instructions in complex electromagnetic environments. BRIEF DESCRIPTION OF DRAWINGS

[0062] Figure 1 A working principle diagram of the random game control method for unmanned aerial vehicle attitude control described in the present application;

[0063] Figure 2 A flowchart for dynamically evaluating real-time motion parameters and generating situation evaluation results for counter behaviors;

[0064] Figure 3 A flowchart for activating a virtual counter entity according to a counter strategy;

[0065] Figure 4 A scatter plot for counter behavior clustering analysis;

[0066] Figure 5 A biaxial column chart for counter signal characteristics. DETAILED DESCRIPTION

[0067] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0068] Referring to Figure 1 The present application provides a random game control method for unmanned aerial vehicle attitude control, which comprises: constructing a cooperative control system to realize intelligent attitude control of the unmanned aerial vehicle. The cooperative control system is composed of an attitude perception unit and a game decision unit. The attitude perception unit is responsible for collecting real-time motion parameters of the unmanned aerial vehicle, such as angular velocity and acceleration. The game decision unit is used to monitor the countermeasures in the environment, such as interference signals or actions of hostile unmanned aerial vehicles. Then, the system dynamically evaluates the real-time motion parameters and countermeasures, generates a situation evaluation result, and formulates a countermeasure strategy based on the result to cope with the uncertain environment. According to the countermeasure strategy, the system activates a virtual countermeasure entity, which simulates the behavior pattern of the real countermeasure party and generates corresponding countermeasure signals, such as electromagnetic interference or false data. At the same time, a distributed communication network is established, which is composed of multiple relay nodes. According to the characteristics of the countermeasure signals, the transmission parameters of the relay nodes, such as frequency and power, are dynamically adjusted. Finally, through the adjusted relay nodes, control instructions are transmitted to realize closed-loop control of the unmanned aerial vehicle attitude, ensuring stability and reliability.

[0069] Embodiment 1: Referring to Figure 2 In specific implementation, the attitude perception unit continuously acquires an inertial measurement data stream, which contains angular velocity, acceleration and azimuth information of the unmanned aerial vehicle. The game decision unit collects a countermeasure behavior data stream, which includes the intensity of interference signals detected in the environment and the motion trajectory of hostile targets. In specific implementation, the inertial measurement data stream and the countermeasure behavior data stream are processed by time sequence alignment, and a timestamp synchronization mechanism is used to integrate the two types of data streams into a synchronous data sequence, ensuring the time consistency of the data points. The statistical features and time domain features of the synchronous data sequence are extracted, including mean, variance and peak value, and the autocorrelation function and power spectral density, which constitute a multi-dimensional feature vector. A pattern recognition method is used to classify the multi-dimensional feature vector, which is based on a support vector machine algorithm, and outputs the current situation level, which is divided into low, medium and high levels. Based on the current situation level, a risk assessment coefficient is calculated, and a situation evaluation result is generated in combination with historical situation data; the calculation of the risk assessment coefficient uses the following formula:

[0070]

[0071] Wherein: represents the risk assessment coefficient, represents the numerical mapping of the current situation level, represents the numerical value of the i-th data point in the historical situation data, represents the total number of historical data points, and represents a weight coefficient, which is determined by pre-calibration. In some embodiments, a threat level indicator and an uncertainty indicator in the situational assessment result are analyzed, the threat level indicator is derived from the quantitative output of the current situation level, and the uncertainty indicator is calculated based on the noise level of the data stream. A strategy decision matrix is established, which contains multiple preset response schemes designed for different combinations of threat levels and uncertainties. According to the threat level indicator and the uncertainty indicator, a basic response scheme is determined by matching and selecting in the strategy decision matrix, and the matching process uses the nearest neighbor algorithm. The basic response scheme is adaptively corrected in combination with real-time environmental parameters, including wind speed and visibility, to generate a countermeasure strategy. Optionally, the training of the pattern recognition method uses a labeled historical data set to improve classification accuracy. It can be understood that the time alignment process can be realized by a hardware timer to ensure data synchronization accuracy. In some embodiments, historical situation data is stored in a ring buffer to support rolling updates.

[0072] Embodiment 2: see Figure 3 In specific implementation, a virtual entity database is configured, which stores a plurality of countermeasure entity models. According to the countermeasure strength requirement in the countermeasure strategy, a corresponding countermeasure entity model is selected from the virtual entity database. The physical parameters and behavior logic of the countermeasure entity model are initialized, including virtual mass and virtual kinematic constraints, and decision rule library, so as to have similar characteristics to the real countermeasure party. The signal generation module of the countermeasure entity model is activated to prepare to generate countermeasure signals. Historical countermeasure data is collected, which includes the action trajectory, communication interaction record and environmental parameter of the countermeasure party. The collected historical countermeasure data is subjected to integrity check, and the integrity score of the data segment is calculated by using the following formula:

[0073]

[0074] Wherein: represents the integrity score, represents the number of data points verified, The total number of data points is represented, data segments with integrity scores below a preset threshold are excluded, and missing data points are supplemented using an interpolation method. A clustering analysis method is used to mine patterns from the processed historical adversarial data. The clustering analysis method uses the K-means algorithm to identify typical adversarial behavior categories, such as active interference type, silent observation type, and decoy type. For each typical adversarial behavior category, an adversarial entity model is constructed, each adversarial entity model includes a behavior decision tree, a state transition probability, and an action response function. The behavior decision tree is used to describe the conditional branching logic, the state transition probability defines the possibility of state transition, and the action response function maps the input stimulus to the output action. The constructed adversarial entity model is imported into a database management system, which uses a relational database and establishes a multi-level index structure according to the behavior category and model complexity. The behavior category is used as the first-level index, and the model complexity is used as the second-level index. A database update trigger is set to monitor new data write operations. When the newly added adversarial data reaches a threshold, the model optimization process is automatically started. The model optimization process involves recalculating the state transition probability and adjusting the behavior decision tree nodes.

[0075] In some embodiments, the virtual entity database is deployed on distributed storage nodes to support high concurrency access. Optionally, the integrity verification process includes data format verification and logical consistency check. It can be understood that the number of clusters required for clustering analysis is determined by the silhouette coefficient method. In specific implementation, the specific implementation of the silhouette coefficient method is as follows: based on the processed historical adversarial data, the silhouette coefficients of all data points under different preset cluster numbers are calculated. The silhouette coefficient reflects the tightness of the data points to their own cluster and the separation degree from the adjacent cluster. By comparing the average silhouette coefficients corresponding to different cluster numbers, the cluster number corresponding to the maximum average silhouette coefficient is determined as the optimal cluster number, thereby completing the division of typical adversarial behavior categories in historical adversarial data. In some embodiments, the construction of the behavior decision tree uses the C4.5 algorithm to learn and generate from labeled data. Optionally, the threshold set by the database update trigger is 1GB of data.

[0076] Referring to Figure 4 The figure is the core visualization result of adversarial behavior pattern mining: the horizontal axis is the adversarial intensity feature, and the vertical axis is the behavior frequency feature. The three clusters correspond to typical adversarial behavior categories, active interference type, silent observation type, and decoy type. The star mark is the center of each cluster. The figure is the key basis for constructing the virtual entity database: by clearly defining the feature boundaries and cluster centers of the three typical adversarial behaviors, quantitative feature support can be provided for subsequent construction of adversarial entity models, ensuring that the virtual adversarial entity can accurately simulate the behavior patterns of the real adversary.

[0077] Example 3: In specific implementation, behavioral pattern instructions from the adversarial strategy are received. These instructions are control commands issued in a structured data format. The action sequences within these instructions are parsed, and each sequence contains a series of predefined basic operational units and their timestamps. The decision-making process of a real adversary is simulated. This process is based on a feedback loop that includes environmental state assessment and action utility prediction. The action sequences are dynamically adjusted based on environmental feedback, which includes state change data generated by the interaction between the virtual adversarial entity and the simulated environment. A multi-channel signal generator generates multiple types of interference signals according to the adjusted action sequences. This generator can output radio frequency noise, pulse interference, and waveform deception signals in parallel. These multiple types of interference signals are combined and encoded to form a complete adversarial signal. The combination and encoding process employs time-division multiplexing and frequency-division multiplexing techniques.

[0078] In some embodiments, parsing behavioral pattern instructions involves syntactic analysis and semantic understanding to extract executable action parameters. Optionally, the simulated decision-making process evaluates action selection using the following utility function formula:

[0079]

[0080] in: This represents the utility value of choosing action a in state s. Indicates the number of evaluation features. Represents the i-th feature The weighting coefficients, This represents the quantized value of the i-th feature of action a in state s. It can be understood that the weight assignment is based on the priority setting of the adversarial strategy. The logic of dynamically adjusting the action sequence is to trigger a predefined alternative action sequence for replacement when the utility value falls below a threshold.

[0081] In some embodiments, each channel of the multi-channel signal generator has an independent digital signal processor for generating baseband signals. Optionally, the frame structure of the signal combination encoding includes a synchronization header, payload data, and a checksum.

[0082] See Figure 5 This chart visualizes the generated adversarial signals from virtual adversarial entities. It uses a dual-axis bar chart to present the core characteristics of five types of adversarial signals: the horizontal axis represents signal type, the left vertical axis represents signal strength, and the right vertical axis represents duration. Purple bars represent signal strength, and orange bars represent duration. The technical value of this chart lies in: clearly defining the characteristic differences of different adversarial signals, providing a quantitative basis for adjusting transmission parameters in distributed communication networks, and serving as crucial data support for connecting the generation of virtual adversarial signals with the dynamic adjustment of communication parameters.

[0083] In some embodiments, the surveying of the topographical features of the operation area of the UAV and the communication environment, including the elevation data and the obstacle profile, the spectrum occupancy and the interference source distribution, is performed to plan the network coverage to determine the required service area. The minimum number of relay nodes and their optimal distribution locations are calculated based on the network coverage, using a network coverage optimization algorithm with the optimization goal of minimizing the number of nodes while satisfying the communication quality. The relay devices with adaptive frequency modulation are deployed at the optimal distribution locations, supporting multi-band operation and dynamic beamforming. The communication protocol between the relay devices is configured using a TDMA-based medium access control mechanism, and the data forwarding rules are defined to form a distributed communication network. In some embodiments, the spectrum features, including the center frequency and the bandwidth, and the power fluctuation features, including the peak power and the average power, of the counter signal are monitored to analyze the impact of the counter signal on the communication channel, which is evaluated by measuring the signal-to-noise ratio degradation value. The channel quality indicator is evaluated to quantify the link reliability. The transmission power and the modulation mode are dynamically adjusted based on the channel quality indicator, and the optimal communication frequency band is adaptively selected based on the spectrum sensing results. The related parameters of the data packet retransmission mechanism, such as the number of retransmissions and the timeout threshold, are updated. The channel quality indicator is calculated using the following formula:

[0084]

[0085] wherein: CQI represents the channel quality indicator, SNR represents the signal-to-noise ratio, BER represents the bit error rate, and a represents the calibration coefficient, which is determined by link budget analysis.

[0086] Table 1: Relationship table for adjusting transmission parameters of relay nodes

[0087] Channel quality index Q range Transmit power adjustment amount (dBm) Modulation mode switching target Q>80 +0 64-QAM 50≤Q≤80 -3 16-QAM Q<50 -6 QPSK

[0088] In some embodiments, referring to Table 1, the network coverage planning is simulated and verified using digital maps and propagation models. Optionally, the deployment of relay devices uses a UAV delivery method to quickly set up the network. It can be understood that the channel quality indicator evaluation period is set to 100 milliseconds to achieve real-time adjustment. In some embodiments, the related parameters of the data packet retransmission mechanism include adaptive retransmission timeout time calculation. Optionally, the spectrum sensing process uses an energy detection algorithm to identify idle frequency bands

[0089] In some embodiments, the UAV attitude control instruction is segmented into multiple data blocks, each data block has a fixed size of 128 bytes, a sequence identifier is added to each data block, the sequence identifier includes a data block serial number and timestamp information. According to the real-time load condition of the multiple relay nodes, the transmission path of the data block is allocated, the real-time load condition is obtained by querying the current buffer usage rate and processor utilization rate of the relay node, a redundant transmission channel is established between the multiple relay nodes, the redundant transmission channel uses a multi-path transmission protocol to ensure reliable transmission of the data block, and the data block is reassembled and checked at the receiving end, the reassembly and checking process includes checking the continuity of the sequence identifier and calculating the cyclic redundancy check code, and the complete control instruction is restored. In some embodiments, the reassembled control instruction is sent to the UAV flight control system, the UAV attitude response data is collected in real time through a sensor array, the sensor array includes a gyroscope, an accelerometer, and a magnetometer, the deviation value between the attitude response data and the expected control target is compared, and the deviation value is calculated using the following formula:

[0090]

[0091] wherein: represents the attitude deviation value, , , represents the roll rate, pitch rate, and yaw rate of the expected control target, , , represents the roll rate, pitch rate, and yaw rate collected in real time by the sensor array. According to the deviation value, the parameter settings of the subsequent control instruction are adjusted, the parameter settings include proportional gain, integral gain, and derivative gain, and closed-loop feedback control is realized.

[0092] In some embodiments, the data block segmentation uses a fixed-length grouping method, and the sequence identifier uses a 32-bit incremental counter. Optionally, the transmission path allocation is based on the Dijkstra shortest path algorithm to calculate the optimal path. It can be understood that the redundant transmission channel is realized by sending copies of the data block on different physical links. In some embodiments, the data acquisition frequency of the sensor array is set to 100 Hz to ensure control real-time performance.

[0093] It is to be understood that the terminology used herein such as first and second, and the like, is only used to distinguish one entity or action from another entity or action, and does not necessarily require or imply any such actual relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0094] While embodiments of the present application have been shown and described with reference to particular embodiments thereof, it will be understood by those skilled in the art that various changes in form and details can be made therein without departing from the spirit and scope of the application. The scope of the application is defined by the appended claims and their equivalents.

Claims

1. A stochastic game control method for attitude control of a UAV, characterized in that, The method comprises: A cooperative control system is constructed, which comprises a posture sensing unit and a game decision unit. Real-time motion parameters of the UAV are collected by the posture sensing unit, and the antagonistic behavior in the environment is monitored by the game decision unit; The real-time motion parameters and the antagonistic behavior are dynamically evaluated to generate a situation evaluation result, and an antagonistic strategy is formulated according to the situation evaluation result; A virtual antagonistic entity is activated according to the antagonistic strategy, which simulates the behavior pattern of the real antagonist and generates an antagonistic signal; A distributed communication network is established, which is composed of multiple relay nodes. The transmission parameters of the multiple relay nodes are adjusted according to the characteristics of the antagonistic signal; The closed-loop control of the UAV posture is completed by delivering control instructions through the adjusted multiple relay nodes.

2. The stochastic game control method for attitude control of UAV according to claim 1, wherein, The dynamic evaluation of the real-time motion parameters and the antagonistic behavior to generate a situation evaluation result comprises: Inertial measurement data stream is continuously obtained by the posture sensing unit, and antagonistic behavior data stream is collected by the game decision unit; The inertial measurement data stream and the antagonistic behavior data stream are time-aligned to form a synchronous data sequence; The statistical features and time-domain features of the synchronous data sequence are extracted to form a multi-dimensional feature vector; The multi-dimensional feature vector is classified by using a pattern recognition method, and the current situation level is output; Based on the current situation level, a risk assessment coefficient is calculated, and the situation evaluation result is generated in combination with historical situation data.

3. The stochastic game control method for unmanned vehicle attitude control of claim 2, wherein, The formulation of the antagonistic strategy according to the situation evaluation result comprises: The threat level index and the uncertainty index in the situation evaluation result are analyzed; A strategy decision matrix is established, which contains multiple preset coping schemes; The threat level index and the uncertainty index are matched and selected in the strategy decision matrix to determine a basic coping scheme; The basic coping scheme is adaptively modified in combination with real-time environmental parameters to generate the antagonistic strategy.

4. The stochastic game control method for attitude control of UAV according to claim 1, wherein, The activation of the virtual antagonistic entity according to the antagonistic strategy comprises: A virtual entity database is configured, which stores multiple antagonistic entity models; According to the antagonistic intensity requirement in the antagonistic strategy, the corresponding antagonistic entity model is selected from the virtual entity database; The physical parameters and behavior logic of the antagonistic entity model are initialized to have similar characteristics to the real antagonist; The signal generation module of the antagonistic entity model is activated to prepare to generate the antagonistic signal.

5. The stochastic game control method for UAV attitude control of claim 4, wherein, The virtual antagonistic entity simulates the behavior pattern of the real antagonist and generates the antagonistic signal, which comprises: The behavior pattern instruction in the antagonistic strategy is received, and the action sequence in the behavior pattern instruction is analyzed; The decision-making process of the real antagonist is simulated, and the action sequence is dynamically adjusted according to the environmental feedback; Multiple types of interference signals are generated by a multi-channel signal generator according to the adjusted action sequence; The multiple types of interference signals are combined and encoded to form a complete antagonistic signal.

6. The stochastic game control method for unmanned vehicle attitude control of claim 1, wherein, The establishment of the distributed communication network comprises: The topographic features and communication environment of the UAV operation area are surveyed, and the network coverage range is planned; According to the network coverage, the minimum number of required relay nodes and their optimal distribution positions are calculated; Relay devices with adaptive frequency modulation function are deployed at the optimal distribution positions; The communication protocol and data forwarding rules between the relay devices are configured to form the distributed communication network.

7. The stochastic game control method for UAV attitude control of claim 6, wherein, The adjustment of the transmission parameters of the multiple relay nodes according to the characteristics of the countermeasure signal includes: Monitoring the spectral characteristics and power fluctuation characteristics of the countermeasure signal; Analyzing the influence degree of the countermeasure signal on the communication channel and evaluating the channel quality index; According to the channel quality index, the transmission power and modulation mode are dynamically adjusted; The optimal communication frequency band is adaptively selected, and the related parameters of the data packet retransmission mechanism are updated.

8. The stochastic game control method for unmanned vehicle attitude control of claim 1, wherein, The control instructions are transmitted by the multiple relay nodes after adjustment, including: The UAV attitude control instructions are divided into multiple data blocks, and a sequence identifier is added to each data block; According to the real-time load condition of the multiple relay nodes, the transmission path of the data block is allocated; Redundant transmission channels are established between the multiple relay nodes to ensure reliable transmission of data blocks; The data blocks are recombined and checked at the receiving end to restore the complete control instructions.

9. The stochastic game control method for UAV attitude control of claim 8, wherein, The closed-loop control of the UAV attitude is completed, including: The recombined control instructions are sent to the UAV flight control system; The UAV attitude response data are collected in real time through a sensor array; The deviation value of the attitude response data from the expected control target is compared; According to the deviation value, the parameter settings of the subsequent control instructions are adjusted to realize closed-loop feedback control.

10. The stochastic game control method for attitude control of UAV according to claim 4, wherein, The virtual entity database is configured, including: Collecting historical countermeasure data, which includes the action trajectory, communication interaction record and environmental parameters of the countermeasure party; Performing integrity check on the collected historical countermeasure data, eliminating invalid data segments and supplementing missing data points; Using clustering analysis method to mine the processed historical countermeasure data, and identifying typical countermeasure behavior categories; For each typical countermeasure behavior category, an countermeasure entity model is constructed, which includes behavior decision tree, state transition probability and action response function; The constructed countermeasure entity model is imported into the database management system, and a multi-level index structure is established according to the behavior category and model complexity; Setting up a database update trigger, which automatically starts the model optimization process when the newly added countermeasure data reaches a threshold.

Images