An agent coordination method based on computing process conflict detection and consensus guarantee

Abstract

This invention discloses an agent coordination method based on conflict detection and consensus guarantee in the computation process, employing a three-stage coordination architecture: In the first processing stage, based on real-time monitoring of the computational capacity ratio, multiple rounds of conflict detection with decreasing time steps are performed on the confirmed trajectories to generate a set of conflict-free confirmed trajectories; in the second processing stage, using this set as a reference, conflict avoidance constraints are introduced into the optimization calculation of the trajectory to be optimized to generate an optimized trajectory; in the third processing stage, look-ahead conflict prediction is performed based on the state prediction model and the computational capacity ratio, and an advance coordination strategy is generated based on the prediction results; the final coordination scheme is integrated, and final conflict verification is performed. If the verification fails, a conflict elimination strategy is triggered; the scheme is executed, and the agent state is monitored in real time. This invention ensures the absolute conflict-free nature of the coordination result through multiple rounds of detection, final verification, and look-ahead prediction, maintaining the integrity of conflict detection even when computational resources are limited.

Description

Technical Field

[0001] This invention relates to the field of multi-agent system control technology, specifically to an agent coordination method based on conflict detection and consensus guarantee in the computation process. Background Technology

[0002] Multi-agent coordination systems are widely used in intelligent transportation, drone swarms, and other fields. Existing technologies typically employ a "planning-coordination-execution" model, generating conflict-free collaborative plans through conflict detection and resolution during the computation phase. However, the inventors have discovered the following significant drawbacks in existing methods: 1. Incomplete security verification chain, posing security risks: After reaching a consensus, existing coordination frameworks lack a final independent verification step for the consensus result. Due to factors such as computational errors, numerical precision, or communication delays, the agreed-upon plan may still contain implicit collision risks. This deficiency leads to potential physical conflicts during the actual execution phase, posing a serious threat in safety-critical scenarios. 2. Inadequate conflict detection mechanism with poor robustness: Existing methods typically perform conflict detection in a single round or simple iteration during the computation process, lacking a systematic multi-round detection and resolution mechanism. When the system scales up or the scenario becomes more complex, potential conflicts are easily missed. Especially when computational resources are limited (computational capacity ratio decreases), existing methods cannot guarantee the completeness of conflict detection, and system security decreases as resources decrease. 3. Lack of proactive conflict prevention capabilities: Current conflict detection mainly relies on static judgments based on the current state, failing to effectively integrate system dynamics models for forward prediction. Therefore, it cannot identify high-risk conflicts in advance, missing the optimal opportunity to take preventative measures and reducing the system's proactive risk avoidance capabilities in dynamic environments.

[0003] In summary, existing technologies suffer from broken security verification chains, incomplete collision detection, and a lack of foresight, making it difficult to provide reliable collision-free guarantees under complex real-world conditions. This limits the deployment and application of multi-agent systems in scenarios with high security requirements. Summary of the Invention

[0004] The purpose of this invention is to provide an agent coordination method based on conflict detection and consensus guarantee in the computation process, so as to overcome the shortcomings of existing technologies such as broken security verification chain, incomplete conflict detection and lack of foresight, which make it difficult to provide reliable collision-free guarantee under real complex conditions.

[0005] To achieve the above objectives, the present invention provides the following technical solutions: In a first aspect, the present invention provides an agent coordination method based on conflict detection and consensus guarantee in the computation process, comprising the following steps: S1 initializes the three-stage coordinated architecture system that executes the first, second, and third processing stages sequentially, and monitors the computing capacity ratio in real time. S2, In the first processing stage, based on the computing capacity ratio, perform multiple rounds of conflict detection with decreasing time steps on the confirmed agent trajectory to generate a set of conflict-free confirmed trajectories; S3, In the second processing stage, taking the set of confirmed trajectories without conflict as a reference, conflict avoidance constraints are introduced into the optimization calculation of the trajectory to be optimized to generate an optimized trajectory that does not conflict with the set of confirmed trajectories. S4, in the third processing stage, forward conflict prediction is performed based on the state prediction model and the computing capacity ratio, and an advance coordination strategy is generated based on the prediction results; S5, integrate the set of confirmed trajectories without conflict, the optimized trajectory, and the advance coordination strategy to form a final coordination scheme, and perform a final conflict verification on the final coordination scheme; if the verification fails, trigger the conflict elimination strategy. S6. Execute the final coordination scheme and monitor the agent's operating status in real time. When an abnormal status and / or distance abnormality is detected, trigger a security response.

[0006] Furthermore, the specific process of S2 is as follows: S2.1, coarse-grained detection performed at a first preset time step identifies agents as potential conflict pairs and forms a potential conflict set; S2.2, perform fine-grained detection on the potential conflict set with a second preset time step that is smaller than the first preset time step to form a real conflict set; S2.3, cross-validation detection is performed with a third preset time step that is smaller than the second preset time step to form a set of validation conflicts; S2.4, Based on the preset agent priority, adjust the trajectories of low-priority agents involved in the real conflict set; S2.5, Repeat step S2.2 on the adjusted trajectory until the verification conflict set is empty, and obtain the final conflict-free trajectory; S2.6, output the final conflict-free trajectory as a set of conflict-free confirmed trajectories to S3.

[0007] Furthermore, the cross-validation detection process is specifically as follows: All potential conflict pairs in the potential conflict set are re-examined; Spatiotemporal location verification is performed on all real conflict pairs in the real conflict set; The potential conflict set and the actual conflict set are cross-compared. When an inconsistency is found, the results of the first two rounds of detection are confirmed using the third time step.

[0008] Furthermore, the specific process of S3 is as follows: S3.1, Receive a set of conflict-free confirmed trajectories; S3.2, Construct a trajectory optimization model. The objective function of the trajectory optimization model is to minimize the sum of control cost and terminal cost, and introduce conflict avoidance constraints into the trajectory optimization model. S3.3, Based on the trajectory optimization model, perform iterative calculation on the trajectory of the agent to be optimized. After each iteration, detect the conflict between the current optimized trajectory and all trajectories in the set of confirmed trajectories without conflict in real time. S3.4 If a conflict is detected in S3.3, the corresponding conflict avoidance constraints are strengthened in the trajectory optimization model, and iterative optimization is re-executed. S3.5, Real-time monitoring of computing capacity ratio. When the real-time monitoring computing capacity ratio exceeds a preset threshold, a conflict detection enhancement mechanism is triggered. The conflict detection enhancement mechanism includes at least one of the following: reducing the sampling time step of conflict detection and increasing the number of trajectory optimization iterations. S3.6 If, after triggering the conflict detection enhancement mechanism, it is still impossible to generate an optimized trajectory that does not conflict with the set of confirmed trajectories that does not conflict through iterative optimization, then a system degradation strategy is triggered. The system degradation strategy includes reducing the number of agents or reducing the speed of agent movement. S3.7, an optimized trajectory that satisfies all constraints and is not in conflict with the confirmed trajectory set without conflict is obtained through iterative optimization.

[0009] 5. The agent coordination method based on computational process conflict detection and consensus guarantee according to claim 1, characterized in that the specific process of S4 is as follows: S4.1, use a state prediction model to predict the state of each agent within a future look-ahead time window to obtain the predicted state; S4.2, detect potential conflicts between any agent pair within the look-ahead time window based on the predicted state, and calculate the probability of future conflicts based on the real-time monitored computing capacity ratio; S4.3 When S4.2 detects a potential conflict and / or the calculated probability of a future conflict exceeds a preset threshold, calculate an advance coordination strategy to avoid the potential conflict; S4.4, The advance coordination strategy generated in S4.3 is fed back as input to the trajectory optimization calculation in S3.

[0010] Furthermore, the specific process of performing final conflict verification on the final coordination scheme in S5 is as follows: The trajectories of all agents in the final coordination scheme are subjected to full-match and full-conflict detection throughout the entire execution cycle using a time step. The all-pair all-collision detection traverses all agent pairs and calculates the distance between the predicted positions of agents at each discrete time point; If at any point in time the distance is less than the preset minimum safe distance, the verification is deemed to have failed. If the verification passes, the final coordination scheme is mathematically proven to be free of physical conflicts during execution. If the verification of S5.2 fails, the conflict elimination strategy is triggered.

[0011] Furthermore, the conflict resolution strategy includes at least one of the following: Strategy 1: Revert to the conflict-free coordination result of the previous coordination cycle; Strategy 2: Trigger emergency coordination and re-execute S2 to S4; Strategy 3: Trigger system downgrade, reduce the number of agents and / or reduce the maximum speed of agents, and then re-execute S2 to S4.

[0012] Furthermore, the specific process of S6 is as follows: S6.1, Control all agents to execute according to the trajectory determined in the final coordination scheme; S6.2, calculates in real time the deviation between the actual position and the desired position of each agent, as well as the actual distance between any two agents; S6.3, trigger a security response based on the monitoring results of S6.2: If a state deviation is detected to exceed the first preset threshold but not the higher second preset threshold, then the corresponding agent is fine-tuned within the current execution cycle. If the detected state deviation exceeds the second preset threshold, emergency coordination is triggered, and S2 to S5 are re-executed; If the actual distance is detected to be less than the minimum safe distance minus the safety margin, an emergency braking action is triggered for the corresponding agent pair.

[0013] Furthermore, this also includes S7: The critical phase transition of a three-stage coordination architecture system is detected by finite size scaling analysis, and the critical point of the three-stage coordination architecture system is determined by calculating the cumulative amount of Binder under different agent numbers. When the computing capacity ratio monitored in real time in S1 exceeds the critical point, it is predicted that the three-stage coordination architecture system will fail. The corresponding self-healing strategy is triggered based on the predicted fault type.

[0014] Secondly, the present invention also provides an agent coordination system based on computational process conflict detection and consensus guarantee, for executing an agent coordination method based on computational process conflict detection and consensus guarantee, comprising: The system initialization and status monitoring module is used to initialize the three-stage coordination architecture and monitor the computing capacity ratio of the three-stage coordination architecture system in real time. The first processing stage module is used to perform multiple rounds of conflict detection with decreasing time steps on the confirmed agent trajectory based on the computing capacity ratio in the first processing stage, and generate a set of conflict-free confirmed trajectories. The third processing stage module is used to introduce conflict avoidance constraints in the optimization calculation of the trajectory to be optimized, with reference to the set of confirmed trajectories without conflict, and to generate an optimized trajectory that does not conflict with the set of confirmed trajectories. The consensus reaching and verification module is used to integrate the set of conflict-free confirmed trajectories, the optimized trajectories, and the pre-coordination strategy to form a final coordination scheme, and to perform a final conflict verification on the final coordination scheme; if the verification fails, a conflict elimination strategy is triggered. The execution and monitoring module is used to execute the final coordination scheme and monitor the operation status of the intelligent agent in real time. When an abnormal status or distance is detected, a safety response is triggered.

[0015] Compared with the prior art, the present invention has the following beneficial technical effects: This invention provides an agent coordination method based on conflict detection and consensus guarantee in the computational process. It uses the computational capacity ratio as a core parameter and dynamically adjusts the intensity of multi-round conflict detection in the first processing stage and the look-ahead prediction and execution monitoring in the third processing stage to ensure that the coordination result after consensus is reached is absolutely free of physical conflicts. Through multi-round conflict detection and computational resource management, the integrity of conflict detection can still be guaranteed even under limited computational resources. Before outputting the final coordination scheme, this invention sets up an independent final conflict verification stage to perform a final security check on all integrated trajectories. If the verification fails, a conflict elimination strategy is immediately triggered for correction. This invention, through multi-round detection, final verification, and look-ahead prediction, ensures the absolute conflict-free nature of the coordination result and maintains the integrity of conflict detection even under limited computational resources.

[0016] Specifically, unlike existing technologies that rely on passive conflict detection based on the current state, this invention introduces look-ahead conflict prediction based on a state prediction model in the third processing stage. By combining critical phase transition theory with the assessment of system conflict probability, high-risk situations can be identified before physical conflict occurs, and early coordination strategies can be automatically generated and triggered. When the computational capacity ratio decreases, the safety level is maintained by optimizing the detection strategy and activating preventive mechanisms in advance. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of an agent coordination method based on conflict detection and consensus guarantee in the computation process, as described in an embodiment of the present invention.

[0018] Figure 2 This is a schematic diagram of a three-segment coordination framework system in an embodiment of the present invention.

[0019] Figure 3 This is a flowchart of the multi-round collision detection process in an embodiment of the present invention.

[0020] Figure 4 This is a flowchart illustrating the final conflict verification process in an embodiment of the present invention.

[0021] Figure 5 This is a timing diagram of conflict detection and elimination in Embodiment 1 of the present invention.

[0022] Figure 6 This is a comparison chart of conflict detection integrity under different computing capacity ratios in Example 2 of this invention. Detailed Implementation

[0023] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of the invention. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0024] This invention proposes an agent coordination method based on a three-stage coordination architecture system (the first processing stage corresponds to a freeze window, the second processing stage corresponds to a planning window, and the third processing stage corresponds to a look-ahead window). This invention does not protect the three-stage coordination framework system itself, but rather the further innovations based on it. (See reference...) Figure 2 This is a schematic diagram of the three-segment coordination framework system.

[0025] See Figure 1 The present invention provides an agent coordination method based on computational process conflict detection and consensus guarantee, comprising the following steps: S1 initializes the three-stage coordinated architecture system that executes the first, second, and third processing stages sequentially, and monitors the computing capacity ratio in real time. In a more specific embodiment of the present invention, the expression for calculating the capacity ratio is: ,in For the actual entropy calculation of the system, The theoretical minimum computational entropy is given by N, where N is the number of agents. For maximum relative velocity, For minimum safe distance, To control the update cycle, This represents the minimum computation time.

[0026] S2, In the first processing stage, based on the computing capacity ratio, perform multiple rounds of conflict detection with decreasing time steps on the confirmed agent trajectory to generate a set of conflict-free confirmed trajectories; In a more specific embodiment of the present invention, the specific process of S2 is as follows: First round of detection: S2.1, with the first preset time step. (The coarse-grained detection performed by the control update cycle quickly identifies obvious conflict pairs, identifies agents as potential conflict pairs, and forms a potential conflict set. Second round of detection: S2.2, with a second preset time step that is smaller than the first preset time step. Fine-grained detection is performed on the potential conflict set to form the real conflict set; Third round of detection: S2.3, with a third preset time step that is smaller than the second preset time step. ,and Cross-validation of the results from the first two rounds of testing was performed, using a decreasing time step ( > > Cross-validation and cross-validation ensure that all potential conflicts are detected and eliminated during the computation phase.

[0027] In this embodiment, the present invention performs multiple rounds of conflict detection and resolution during the calculation process to ensure that all potential conflicts are eliminated during the calculation stage. This is an improvement over the single-round conflict detection in the three-stage coordination framework system.

[0028] In a more specific embodiment provided by the present invention, the cross-validation detection process is as follows: All potential conflict pairs in the potential conflict set are re-detected using a third preset time step to confirm whether any real conflicts have been missed. For all real conflict pairs in the real conflict set, perform spatiotemporal location verification, and re-detect using a third preset time step to determine the precise location and time of the real conflict pairs. The potential conflict set is cross-compared with the actual conflict set. When an inconsistency is found, the detection results of the first two rounds are confirmed at the third time step to ensure that all potential conflicts are detected and eliminated during the calculation phase.

[0029] S2.4, Based on the preset agent priority, adjust the trajectories of low-priority agents involved in the real conflict set; S2.5, Repeat step S2.2 on the adjusted trajectory until the conflict set is verified to be empty, and the final conflict-free trajectory is obtained; S2.6, output the final conflict-free trajectory as a set of conflict-free confirmed trajectories to S3.

[0030] In a more specific embodiment provided by the present invention, the specific algorithm for multi-round collision detection is as follows: The first round of detection algorithm is as follows: Input: All confirmed tracks Time window First preset time step ; Initialization: Potential conflict set ; For each time point , , ,..., For each pair of agents , Calculate distance ;if ,Will Add to potential conflict set

[0031] Output potential conflict set .

[0032] The second-round detection algorithm is as follows: Input: Set of potential conflicts Second preset time step ; Initialization: Set of real conflicts ; For each potential conflict : In the time window Within, use the second preset time step. Perform fine particle size detection; For each time point , ,..., ; Calculate distance ; if ,Will Add to real conflict set

[0033] Output: Set of real conflicts .

[0034] See Figure 3 The third round of detection algorithm is as follows: Input: Set of potential conflicts A collection of real conflicts Third preset time step ,and ; Initialization: Validate conflict set ; For each potential conflict ; within the time window Within, use the third preset time step. Perform verification and testing for each time point. , ,..., ; Calculate distance ; if ,Will join in ; For each real conflict ; In the time window Within, use the third preset time step. Verification and testing were conducted to confirm the precise location and timing of the conflict.

[0035] Cross-comparison: If the potential conflict set Collection of real conflicts There are inconsistencies, so a third preset time step is used. Please confirm.

[0036] Output: Set of verification conflicts (if If so, the verification is successful.

[0037] A spatiotemporal tube collision detection algorithm is employed, specifically applied to multi-round collision detection in the first processing stage of multi-agent coordination, and organically combined with a three-round detection method; the expression of the spatiotemporal tube collision detection algorithm is: (1) in, Let i be the predicted position of agent i at time t (based on the confirmed trajectory), and let i be the predicted position of agent i at time t. Energy body physical radius, For minimum safe distance, the detection time window It covers the entire timeframe of the first processing stage. The spatiotemporal tube algorithm uses different time steps in each of the three rounds of detection. , , By combining the spatiotemporal management algorithm with a multi-round detection method, the integrity and accuracy of conflict detection are ensured. This is a new application of the spatiotemporal management algorithm in the field of multi-agent coordination.

[0038] S2.4, Based on the preset agent priority, adjust the trajectories of low-priority agents involved in the real conflict set; In a more specific embodiment of the present invention, when a conflict is detected, a priority adjustment strategy is employed: The adjustment targets are determined based on the priority of the agents (such as urgency, task importance, etc.); Adjust the trajectory of low-priority agents (e.g., decelerate, change path, etc.) to eliminate conflicts; After reorganization, conflict detection is performed again until all conflicts are eliminated.

[0039] A priority adjustment strategy is used to resolve conflicts, ensuring that all spatiotemporal trajectories within the first processing phase meet the safe distance constraints. This ensures that there is no physical conflict.

[0040] S2.5, Repeat step S2.2 on the adjusted trajectory until the verification conflict set is empty, and obtain the final conflict-free trajectory; S2.6, output the final conflict-free trajectory as a set of conflict-free confirmed trajectories to S3.

[0041] Mathematical proof methods are used to ensure conflict elimination: Prerequisites: System assumptions: (1) The agent dynamics model is deterministic, the control input is precise, and there is no external interference; (2) Detection assumptions: The conflict detection algorithm is complete, and the time step satisfies Sampling Theorem ( Satisfying the sampling frequency The numerical calculation error is within the allowable range (relative error). (3) Scope of application: Number and scale of intelligent agents Control the update cycle Minimum safe distance .

[0042] Theorem: If all trajectories in the first processing stage satisfy the following conditions after multiple rounds of conflict detection and resolution: Therefore, the trajectory within the first processing stage is mathematically proven to be free of physical conflicts.

[0043] In this embodiment, the collision detection algorithm is complete: three rounds of detection cover all trajectory pairs and time points. The first round is a coarse-grained detection (first preset time step). Covering all trajectory pairs to ensure no omissions; second round of fine-grained detection (second preset time step). The first round of detected conflicts is used for precise detection; the third round is used for verification detection (third preset time step). ,and Cross-validation is used to ensure no omissions. Due to the decreasing time step... ), and the third preset time step Much smaller than the control update cycle ,according to The sampling theorem can detect all potential collisions (sampling frequency). (2) Convergence of the conflict resolution algorithm: The priority adjustment strategy eliminates all conflicts in a finite number of steps. Since the number of conflicts is finite (at most...), (Yes), and each adjustment eliminates at least one conflict; the algorithm can achieve this in at most... The trajectory converges within a step. Therefore, if all trajectories satisfy the safe distance constraint, it can be mathematically proven that there are no physical conflicts.

[0044] Parameter selection basis: Third preset time step Selection criteria: based on According to the sampling theorem, to accurately detect collisions, the sampling frequency should be at least twice the system's highest frequency. For a system controlling the update cycle, the highest frequency is approximately 10 Hz, therefore the sampling time step should be less than 0.05 s. Considering safety... Margin, selection t3≤0.01s, ensuring detection accuracy. (2) and Selection criteria: Experimental verification shows that this decreasing time step relationship can balance computational efficiency and detection accuracy while ensuring detection integrity.

[0045] S3, in the second processing stage, with the set of confirmed trajectories without conflict as a reference, conflict avoidance constraints are introduced in the optimization calculation of the trajectory to be optimized to generate an optimized trajectory that does not conflict with the set of confirmed trajectories, which can prevent conflicts during the optimization process. In a more specific embodiment of the present invention, the specific process of S3 is as follows: S3.1, Receive a set of conflict-free confirmed trajectories; S3.2, Construct a trajectory optimization model. The objective function of the trajectory optimization model is to minimize the sum of control cost and terminal cost, and introduce conflict avoidance constraints into the trajectory optimization model. The expression for the trajectory optimization model is: (2) (3) (4) (5) Where H represents optimization. Q and R are weight matrices. This is the cost to the end user.

[0046] S3.3, Based on the trajectory optimization model, perform iterative calculation on the current agent trajectory to be optimized, and after each iteration, detect in real time the conflict between the current optimized trajectory and all trajectories in the set of confirmed trajectories without conflict; S3.4 If a conflict is detected in S3.3, the corresponding conflict avoidance constraints are strengthened in the trajectory optimization model, and iterative optimization is re-executed. S3.5, Real-time monitoring of computing capacity ratio. When the real-time monitoring computing capacity ratio exceeds a preset threshold, a conflict detection enhancement mechanism is triggered. The conflict detection enhancement mechanism includes at least one of the following: reducing the sampling time step of conflict detection and increasing the number of trajectory optimization iterations. Specifically, when the computational capacity ratio hour, Enhanced conflict detection triggered by preset thresholds: Increase the sampling density of collision detection (time step from) Reduce to Increase the number of optimization iterations to ensure that conflict constraints are fully satisfied. S3.6 If, after triggering the conflict detection enhancement mechanism, it is still impossible to generate an optimized trajectory that does not conflict with the set of confirmed trajectories that does not conflict through iterative optimization, then a system degradation strategy is triggered. The system degradation strategy includes reducing the number of agents or reducing the speed of agent movement. S3.7, an optimized trajectory that satisfies all constraints and is not in conflict with the confirmed trajectory set without conflict is obtained through iterative optimization.

[0047] S4. In the third processing stage, forward conflict prediction is performed based on the state prediction model and the computing capacity ratio, and an advance coordination strategy is generated based on the prediction results. This strategy can predict possible future conflicts and coordinate them in advance. See Figure 4 In a more specific embodiment provided by the present invention, the specific process of S4 is as follows: S4.1, use a state prediction model to predict the state of each agent within a future look-ahead time window to obtain the predicted state; The expression for the state prediction model to predict the state of each agent within a future look-ahead time window is as follows: (6) S4.2, detect potential conflicts between any agent pair within the look-ahead time window based on the predicted state, and calculate the probability of future conflicts based on the real-time monitored computing capacity ratio; The expression for calculating the probability of future conflicts based on the computing capacity ratio obtained from real-time monitoring is as follows: (7) in This is a forward-looking time window.

[0048] S4.3 When S4.2 detects a potential conflict and / or the calculated probability of a future conflict exceeds a preset threshold, calculate an advance coordination strategy to avoid the potential conflict; S4.4, The advance coordination strategy generated in S4.3 is fed back as input to the trajectory optimization calculation in S3.

[0049] In this embodiment, the expression for predicting the conflict probability based on the calculated capacity ratio ρ is: (8) Where α, β, and γ are empirical parameters, determined experimentally: (Typical value 1.0) (Typical value 0.5) (Typical value 1.0) This is the critical point (determined through finite-size scaling analysis). When... (in When this occurs, the conflict prevention mechanism is triggered.

[0050] Methods for selecting parameters α, β, and γ: (a) Initial parameter settings: Based on experience, set the initial parameter values ​​as follows: α=1.0, β=0.5, γ=1.0.

[0051] (b) Data Acquisition: Measuring the actual collision probability under different computational capacity ratios ρ. .

[0052] (c) Parameter optimization: The parameters are optimized using the least squares method to improve the predicted conflict probability. Conflict probability with reality The mean square error is minimized: (9) (d) Parameter validation: Validate the optimized parameters on an independent test set to ensure that the prediction accuracy meets the requirements (prediction error <5%).

[0053] (e) Parameter range determination: Through multiple experiments, the effective range of the parameters was determined as follows: α∈[0.5, 2.0], β∈[0.1, 1.0], γ∈[0.5, 2.0]. The parameter values ​​may vary slightly for different application scenarios, but they are generally within the above range.

[0054] S5. Integrate the set of confirmed non - conflicting trajectories, the optimized trajectories, and the advance coordination strategy to form a final coordination plan, and perform a final conflict verification on the final coordination plan; if the verification fails, trigger a conflict resolution strategy; In a more specific embodiment provided by the present invention, the specific process of performing the final conflict verification on the final coordination plan in S5 is as follows: Use a time step to perform all - against - all conflict detection on the trajectories of all agents in the final coordination plan over the entire execution cycle; this step uses a high - precision conflict detection algorithm (time step , much smaller than the control update period ); the detection time window covers the entire execution cycle (from the current time to the future time , where or determined according to system requirements, but ); The all - against - all conflict detection traverses all agent pairs, and at each discrete time point, calculates the distance between the predicted positions of agents; If at any time point, the distance is less than the preset minimum safety distance, it is determined that the verification fails; If the verification passes, the final coordination plan is guaranteed by mathematical proof to have no physical conflicts during execution. If the verification in S5.2 fails, trigger a conflict resolution strategy.

[0055] Specifically, the specific algorithm for the final conflict verification is: Input: The final trajectories of all agents , the execution cycle , the verification time step ; Initialization: Conflict set ; All - against - all conflict detection: For each pair of agents (i, j), i < j (a total of N(N 1) / 2 pairs); For each time point

[0056] Calculate the distance between the predicted positions of agents ; If Add (i, j, t) to ; Verification result: If , the verification passes, and it is guaranteed by mathematical proof that there are no physical conflicts during execution; if Verification failed, triggering the conflict resolution strategy; Output: Validation result (pass / fail). If it fails, output the set of conflicts. This triggers the conflict resolution strategy.

[0057] Conflict resolution strategies include at least one of the following: Strategy 1: Revert to the conflict-free coordination result of the previous coordination cycle; Strategy 2: Trigger emergency coordination and re-execute S2 to S4; Strategy 3: Trigger system downgrade, reduce the number of agents and / or reduce the maximum speed of agents, and then re-execute S2 to S4.

[0058] No-conflict guarantee theorem: Prerequisites: (1) System assumptions: The agent dynamics model is deterministic, the control input is precise, and there is no external interference; (2) Detection assumptions: The final conflict verification algorithm is complete, and the time step satisfies the Nyquist sampling theorem. (2) The sampling frequency is greater than 2 × 10 Hz, and the numerical calculation error is within the allowable range (relative error < 10-6); (3) Scope of application: the number of agents N ≤ 200, and the control update cycle Execution cycle Minimum safe distance .

[0059] Theorem: If the final conflict verification passes (all trajectory pairs (i, j) at all time points) Meet safety distance constraints Then, through mathematical proof, it is guaranteed that there will be no physical conflict during execution.

[0060] In this embodiment, the completeness of the final collision verification algorithm is as follows: all-pair collision detection covers all trajectory pairs (i,j)(i j, total N (N 1) / 2 pairs) and all time points The number of detection time points is (2) High precision in final verification: time step (in (To control the update cycle), much smaller than the control update cycle. For The system sampling frequency According to the Nyquist sampling theorem, all potential conflicts can be detected. (3) Determinism of the physical system: Under known control input, the state evolution of the physical system is deterministic (assuming the preconditions are met). If the high-precision verification passes (all trajectory pairs satisfy the safe distance constraint at all time points), then there will necessarily be no conflict during execution. This is the key improvement of this invention compared to the Prollect framework: Although the Prollect framework proves the feasibility of recursion, it does not provide a final conflict verification mechanism before execution, and cannot guarantee the absence of physical conflicts during execution through mathematical proof.

[0061] Parameter selection criteria: Selection criteria: Based on the Nyquist sampling theorem and experimental verification, to accurately detect collisions, the sampling time step should be at least less than 1 / 5 of the control update cycle. This ensures high accuracy in the final verification while maintaining verification efficiency.

[0062] S6. Execute the final coordination scheme and monitor the agent's operating status in real time to ensure conflict-free execution. When an abnormal status and / or distance abnormality is detected, trigger a security response.

[0063] In a more specific embodiment provided by the present invention, the specific process of S6 is as follows: S6.1, Control all agents to execute according to the trajectory determined in the final coordination scheme. All agents shall strictly follow the trajectory confirmed in the first processing stage and shall not deviate from it. S6.2, calculate in real time the deviation between the actual position and the desired position of each agent, as well as the actual distance between any two agents, expressed as: (10) in, The deviation between the actual position and the desired position for each agent. For the deviation of the desired position, if (e.g., 0.5 m) triggers exception handling.

[0064] S6.3, trigger a security response based on the monitoring results of S6.2: If a state deviation is detected to exceed the first preset threshold but not the higher second preset threshold, then the corresponding agent is fine-tuned within the current execution cycle. Specifically, if the abnormality is small Fine-tuning can be performed within the current cycle without affecting the confirmed trajectory; If the detected state deviation exceeds the second preset threshold, emergency coordination is triggered, and S2 to S5 are re-executed; Specifically, if the abnormality is large This triggers emergency coordination, requiring a recalculation of the coordination across the three processing stages.

[0065] If the actual distance is detected to be less than the minimum safe distance minus the safety margin, an emergency braking action is triggered for the corresponding agent pair.

[0066] The expression for the real-time monitoring of the actual distance between intelligent agents is: (11) if ( (For safety margin), emergency braking is triggered.

[0067] In a more specific embodiment provided by the present invention, S7 is also included: The critical phase transition of a three-stage coordination architecture system is detected by finite size scaling analysis, and the critical point of the three-stage coordination architecture system is determined by calculating the cumulative amount of Binder under different agent numbers. When the computing capacity ratio ρ monitored in real time in S1 exceeds the critical point, it is predicted that the three-stage coordination architecture system will fail. The calculation formula for the critical phase transition of the detection system based on finite-size scaling analysis is as follows: (12) Where m is the order parameter (conflict probability), and the value of ρ corresponding to the convergence of U4 to the same value under different system sizes N is the critical point. .

[0068] In this embodiment, the method for determining the critical point is as follows: Data acquisition: Under different agent number scales N (e.g., N=50, 100, 150, 200), measure the conflict probability m(ρ,N) corresponding to different computational capacity ratios ρ (e.g., ρ=0.8, 0.9, 1.0, 1.1, 1.2, 1.3, 1.4).

[0069] Binder cumulative amount calculation: For each (ρ, N) combination, calculate the Binder cumulative amount:

[0070] Where mk represents the k-th moment of m, obtained by statistical averaging of multiple independent experiments.

[0071] Finite-size scaling analysis: Plot the curve of U4(ρ,N) as a function of ρ for different values ​​of N. When the curves of U4 for different N values ​​intersect or converge at a certain ρ value, that ρ value is the critical point. .

[0072] Critical point determination: Through experimental verification, when At the same time, under different system sizes N The value converges to approximately 0.623, indicating that the three-stage coordination architecture system undergoes a critical phase transition at this point.

[0073] The corresponding self-healing strategy is triggered based on the predicted fault type.

[0074] when At that time, predict system failures: when At that time, the predictive information theory conflict failure (insufficient conflict detection capability) occurred.

[0075] when At that time, predict metastable deadlock failure (decreased coordination efficiency).

[0076] Trigger the corresponding self-healing strategy based on the fault type: Information theory conflict failure: increasing computational resources makes... If it cannot be increased, trigger system downgrade; Metastable deadlock fault: Introducing optimal stochastic time quantization This will improve coordination efficiency.

[0077] Example 1 See Figure 5 Intelligent traffic intersection coordination The scenario describes a four-way intersection with no traffic lights, where 100 autonomous vehicles need to safely cross without any physical conflicts.

[0078] Three-stage coordination architecture system parameters: Number of agents: ; Maximum relative speed: ; Minimum safe distance: ; Control the update cycle: ; Minimum computation time: ; Computational minimum computational entropy: (13) Coordination process: Step 1: Initialization of the three-stage coordination architecture system Three-stage coordination architecture system status: The three-stage coordination architecture system is in a supercritical state with sufficient conflict detection capabilities.

[0079] Step 2: First Processing Stage – Confirmed Trajectory Conflict Detection Multi-round collision detection: First round (coarse-grained): First preset time step 15 potential conflicts were detected.

[0080] Second round (fine-grained): Second preset time step size 15 potential conflicts were precisely detected, and 8 real conflicts were confirmed.

[0081] Third round (verification): Third preset time step Eight pairs of real conflicts were verified to ensure that none were missed.

[0082] Conflict resolution: For the 8 pairs of real conflicts, a priority adjustment strategy is adopted: Priority is determined based on the urgency of the vehicle (emergency vehicles have higher priority); Adjust the trajectory of low-priority vehicles (by slowing them down or changing their path); After adjustments, re-run the conflict detection until all conflicts are eliminated.

[0083] Conflict resolution guarantee: After three rounds of conflict detection and resolution, all trajectories in the first processing phase met the requirements. This guarantees absolutely no conflicts.

[0084] Step 3: Second Processing Stage – Trajectory Optimization Calculation Constraint optimization: For the trajectory that needs to be optimized, conflict avoidance constraints are added to the optimization problem to ensure that the optimized trajectory does not conflict with the confirmed trajectory.

[0085] Real-time conflict detection: During the optimization iteration process, conflicts between the optimized trajectory and the confirmed trajectory are detected after each iteration. If a conflict is detected, conflict avoidance constraints are added to the constraints, and the optimization is restarted.

[0086] Optimization result: After 10 iterations, all optimized trajectories are free from conflict with the confirmed trajectories, and the optimization is complete.

[0087] Step 4: Third Processing Stage – Forward-Looking Conflict Prediction State prediction: Predict the state within the next 2 seconds ( The status of all vehicles; Conflict prediction: Based on the predicted state, 3 pairs of potential conflicts were detected; Early coordination: Calculate the optimal coordination strategy to make the three low-priority vehicles slow down in advance to avoid potential conflicts; Coordination Application: The coordination strategy is applied to the trajectory optimization in the second processing stage for re-optimization.

[0088] Step 5: Consensus Reached and Final Conflict Verification Consensus reached: Integrate the results of the three processing phases to form a final coordination plan; Final conflict verification: Perform full-match, full-collision detection on the final trajectories of all 100 vehicles; Detection time window: Second).

[0089] Detection time step: (high precision).

[0090] Test results: All trajectory pairs met the safe distance constraints at all time points, and the verification was successful.

[0091] Conflict-free guarantee: The completeness and high precision verified by the final conflict guarantee that there will be absolutely no physical conflicts during execution.

[0092] Step 6: Execution and Real-time Monitoring Execute confirmed trajectories: All vehicles strictly follow the trajectories confirmed in the first processing stage; Conflict monitoring: Real-time monitoring of the actual distance between vehicles. Monitoring results: The actual distance between all vehicle pairs is greater than the safe distance, and there is no risk of conflict.

[0093] Performance metrics: Physical conflict rate: 0% (0 conflicts in 10,000 tests, 95% confidence interval: [0%, 0.03%]), which is a significant improvement over existing methods (0.1-0.5%, 95% confidence interval: [0.08%, 0.52%]) (p<0.001, two-tailed t-test, Bonferroni correction).

[0094] The collision detection completeness rate was 99.9% (95% confidence interval: [99.7%, 100%] when ρ≥ρc), which is a significant improvement compared to existing methods (85%, 95% confidence interval: [83%, 87%]) (p<0.001).

[0095] Coordination delay: <50ms (from vehicle entering the area to obtaining the coordinated trajectory, 95% confidence interval: [35ms, 48ms]), which is significantly reduced compared to existing methods (100ms, 95% confidence interval: [85ms, 115ms]) (p<0.001).

[0096] Three-stage coordination architecture system capacity: Simultaneous coordination of >100 vehicles (test range: 50-150 vehicles).

[0097] To verify the technical effectiveness of the present invention, a comparative experiment with existing methods was conducted, see reference [link to relevant documentation]. Figure 6 : Comparison with ORCA method: Physical collision rate: 0% for this invention vs. 0.2% for DMPC (10,000 tests, p<0.001); Coordination delay: 35-48ms for this invention vs. 80-120ms for ORCA (p<0.001); Three-stage coordination architecture system capacity: This invention > 100 vehicles vs. ORCA < 50 vehicles; Collision detection completeness: This invention 99.9% vs. ORCA 78% Comparison with DMPC method Physical collision rate: 0% vs. DMPC 0.2% (10,000 tests, p<0.001); Computational complexity: This invention vs. DMPC ; Three-stage coordination architecture system reliability: This invention 99.7% vs. DMPC 94%; Collision detection completeness: 99.9% for this invention vs. 88% for DMPC.

[0098] Comparison with the Prollect framework: Physical collision rate: 0% for this invention vs. 0.1% for Protect (10,000 tests, p<0.001); Collision detection completeness: This invention 99.9% vs. Prollect 92%; Three-stage coordination architecture system reliability: 99.7% for this invention vs. 96% for Protect; Final Conflict Verification: This invention provides a final conflict verification mechanism, which is not provided by the Prollect framework.

[0099] Example 2 See Figure 6 Scenarios with limited computing resources Scenario Description: Under conditions of limited computing resources The system can still guarantee the integrity of conflict detection.

[0100] Three-stage coordination architecture system parameters: Number of agents: N=100; Calculate the capacity ratio: (Computational resources are limited); Conflict detection enhancement mechanism: Insufficient computing resources detected: The system detected that This triggers the conflict detection enhancement mechanism.

[0101] Enhancement measures: Increase the sampling density of collision detection: time step from Reduce to ; Increase the number of optimization iterations from 10 to 20 to ensure that conflict constraints are fully satisfied.

[0102] Prioritize high-priority agents: use simplified trajectories for low-priority agents to reduce computation.

[0103] Conflict detection results: After the enhancement mechanism, the conflict detection integrity rate reached 98.5%, which is slightly lower than the normal level (99.9%), but still ensures the integrity of conflict detection.

[0104] Performance Comparison: Compared to existing methods (which do not employ collision detection enhancement mechanisms): The collision detection completeness rate was 98.5% (this invention, 95% confidence interval: [97.8%, 99.2%]) vs. 85% (existing methods, 95% confidence interval: [83%, 87%]), with a significant difference (p<0.001).

[0105] Physical conflict rate: 0% (this invention, 95% confidence interval: [0%, 0.05%]) vs. 0.3% (existing method, 95% confidence interval: [0.2%, 0.4%]), the difference was significant (p<0.001).

[0106] The reliability of the three-stage coordination architecture system was 99.7% (this invention, 95% confidence interval: [99.4%, 99.9%]) vs. 92% (existing methods, 95% confidence interval: [90%, 94%]), with a significant difference (p<0.001).

[0107] Example 3 drone swarm coordination Scenario Description: 100 drones need to coordinate their flight in a complex urban environment, avoiding collisions with buildings, other drones, and ground obstacles, requiring absolutely no physical conflicts.

[0108] Three-stage coordination architecture system parameters Number of agents: ; Maximum relative speed: ; Minimum safe distance: ; Control update cycle: ; Minimum computation time: ; Computational minimum computational entropy:

[0109] (14) Coordination process: Initial state of the three-stage coordination architecture system: The three-stage coordination architecture system is in a supercritical state.

[0110] Multi-round collision detection: First round (verification): First preset time step 20 potential conflicts were detected; Second round (verification): Second preset time step 12 pairs of real conflicts were confirmed; Third round (verification): Third preset time step The verification was thorough and complete.

[0111] Final collision verification: Perform full-collision detection on the final trajectories of all 100 drones, with a detection time window. (0.5 seconds in the future), detection time step Verification passed.

[0112] Performance metrics: Physical collision rate: 0% (0 collisions in 10,000 tests, 95% confidence interval: [0%, 0.03%]) Conflict detection completeness: 99.9% (95% confidence interval: [99.7%, 100%]) Coordination delay: <30ms (95% confidence interval: [22ms, 28ms]) Intelligent agent capacity: Simultaneous coordination of >100 drones (test range: 50-150 drones) Example 4 Scenario Description: 50 robots need to coordinate their movement in a factory environment to complete material handling tasks, with absolutely no physical conflicts required.

[0113] Three-stage coordination architecture system parameters: Number of agents:

[0114] Maximum relative speed:

[0115] Minimum safe distance:

[0116] Control update cycle:

[0117] Minimum computation time:

[0118] Computational minimum computational entropy: (15) Coordination process: Initial state of the three-stage coordination architecture system: The three-stage coordination architecture system is in a subcritical state, triggering the conflict detection enhancement mechanism.

[0119] Enhanced collision detection: Increase the sampling density of collision detection: time step from t=0.2s decreases to t=0.1s.

[0120] Increase the number of optimization iterations: from 10 to 25.

[0121] Prioritize high-priority robots: use simplified trajectories for low-priority robots.

[0122] Final collision verification: Perform full-match, full-collision detection on the final trajectories of all 50 robots, with a detection time window. (Next 2 seconds), detection time step Verification passed.

[0123] Performance metrics: Physical collision rate: 0% (0 collisions in 10,000 tests, 95% confidence interval: [0%, 0.03%]) Conflict detection completeness: 98.8% (95% confidence interval: [98.2%, 99.4%]) Coordination delay: <80 ms (95% confidence interval: [65 ms, 78 ms]) The reliability of the three-stage coordination architecture system is 99.5% (95% confidence interval: [99.2%, 99.8%).

[0124] In a more specific embodiment provided by the present invention, the present invention also provides an agent coordination system based on computational process conflict detection and consensus guarantee, for executing an agent coordination method based on computational process conflict detection and consensus guarantee, comprising: The system initialization and status monitoring module is used to initialize the three-stage coordination architecture and monitor the computing capacity ratio of the three-stage coordination architecture system in real time. The first processing stage module is used to perform multiple rounds of conflict detection with decreasing time steps on the confirmed agent trajectory based on the computing capacity ratio in the first processing stage, and generate a set of conflict-free confirmed trajectories. The third processing stage module is used to introduce conflict avoidance constraints in the optimization calculation of the trajectory to be optimized, with reference to the set of confirmed trajectories without conflict, and to generate an optimized trajectory that does not conflict with the set of confirmed trajectories. The consensus reaching and verification module is used to integrate the set of conflict-free confirmed trajectories, the optimized trajectories, and the pre-coordination strategy to form a final coordination scheme, and to perform a final conflict verification on the final coordination scheme; if the verification fails, a conflict elimination strategy is triggered. The execution and monitoring module is used to execute the final coordination scheme and monitor the operation status of the intelligent agent in real time. When an abnormal status or distance is detected, a safety response is triggered.

[0125] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for agent coordination based on conflict detection and consensus guarantee in the computation process, characterized in that, Includes the following steps: S1 initializes the three-stage coordinated architecture system that executes the first, second, and third processing stages sequentially, and monitors the computing capacity ratio in real time. S2, In the first processing stage, based on the computing capacity ratio, perform multiple rounds of conflict detection with decreasing time steps on the confirmed agent trajectory to generate a set of conflict-free confirmed trajectories; S3, In the second processing stage, taking the set of confirmed trajectories without conflict as a reference, conflict avoidance constraints are introduced into the optimization calculation of the trajectory to be optimized to generate an optimized trajectory that does not conflict with the set of confirmed trajectories. S4, in the third processing stage, forward conflict prediction is performed based on the state prediction model and the computing capacity ratio, and an advance coordination strategy is generated based on the prediction results; S5, integrate the set of confirmed trajectories without conflict, the optimized trajectory, and the advance coordination strategy to form a final coordination scheme, and perform a final conflict verification on the final coordination scheme; If verification fails, the conflict resolution strategy is triggered. S6. Execute the final coordination scheme and monitor the agent's operating status in real time. When an abnormal status and / or distance abnormality is detected, trigger a security response.

2. The agent coordination method based on computational process conflict detection and consensus guarantee as described in claim 1, characterized in that, The specific process of S2 is as follows: S2.1, coarse-grained detection performed at a first preset time step identifies agents as potential conflict pairs and forms a potential conflict set; S2.2, perform fine-grained detection on the potential conflict set with a second preset time step that is smaller than the first preset time step to form a real conflict set; S2.3, cross-validation detection is performed with a third preset time step that is smaller than the second preset time step to form a set of validation conflicts; S2.4, Based on the preset agent priority, adjust the trajectories of low-priority agents involved in the real conflict set; S2.5, Repeat step S2.2 on the adjusted trajectory until the verification conflict set is empty, and obtain the final conflict-free trajectory; S2.6, output the final conflict-free trajectory as a set of conflict-free confirmed trajectories to S3.

3. The agent coordination method based on conflict detection and consensus guarantee in the computation process according to claim 2, characterized in that, The cross-validation detection process is specifically as follows: All potential conflict pairs in the potential conflict set are re-examined; Spatiotemporal location verification is performed on all real conflict pairs in the real conflict set; The potential conflict set and the actual conflict set are cross-compared. When an inconsistency is found, the results of the first two rounds of detection are confirmed using the third time step.

4. The agent coordination method based on computational process conflict detection and consensus guarantee as described in claim 2, characterized in that, The specific process of S3 is as follows: S3.1, Receive a set of conflict-free confirmed trajectories; S3.2, Construct a trajectory optimization model. The objective function of the trajectory optimization model is to minimize the sum of control cost and terminal cost, and introduce conflict avoidance constraints into the trajectory optimization model. S3.3, Based on the trajectory optimization model, perform iterative calculation on the current agent trajectory to be optimized, and after each iteration, detect in real time the conflict between the current optimized trajectory and all trajectories in the set of confirmed trajectories without conflict; S3.4 If a conflict is detected in S3.3, the corresponding conflict avoidance constraints are strengthened in the trajectory optimization model, and iterative optimization is re-executed. S3.5, Real-time monitoring of computing capacity ratio. When the real-time monitoring computing capacity ratio exceeds a preset threshold, a conflict detection enhancement mechanism is triggered. The conflict detection enhancement mechanism includes at least one of the following: reducing the sampling time step of conflict detection and increasing the number of trajectory optimization iterations. S3.6 If, after triggering the conflict detection enhancement mechanism, it is still impossible to generate an optimized trajectory that does not conflict with the set of confirmed trajectories that does not conflict through iterative optimization, then a system degradation strategy is triggered. The system degradation strategy includes reducing the number of agents or reducing the speed of agent movement. S3.7, an optimized trajectory that satisfies all constraints and is not in conflict with the confirmed trajectory set without conflict is obtained through iterative optimization.

5. The agent coordination method based on conflict detection and consensus guarantee in the computation process according to claim 1, characterized in that, The specific process of S4 is as follows: S4.1, use a state prediction model to predict the state of each agent within a future look-ahead time window to obtain the predicted state; S4.2, detect potential conflicts between any agent pair within the look-ahead time window based on the predicted state, and calculate the probability of future conflicts based on the real-time monitored computing capacity ratio; S4.3 When S4.2 detects a potential conflict and / or the calculated probability of a future conflict exceeds a preset threshold, calculate an advance coordination strategy to avoid the potential conflict; S4.4, The advance coordination strategy generated in S4.3 is fed back as input to the trajectory optimization calculation in S3.

6. The agent coordination method based on conflict detection and consensus guarantee in the computation process according to claim 1, characterized in that, The specific process for performing final conflict verification on the final coordination scheme in S5 is as follows: The trajectories of all agents in the final coordination scheme are subjected to full-match and full-conflict detection throughout the entire execution cycle using a time step. The all-pair all-collision detection traverses all agent pairs and calculates the distance between the predicted positions of agents at each discrete time point; If at any point in time the distance is less than the preset minimum safe distance, the verification is deemed to have failed. If the verification passes, the final coordination scheme is mathematically proven to be free of physical conflicts during execution. If the verification of S5.2 fails, the conflict elimination strategy is triggered.

7. The agent coordination method based on computational process conflict detection and consensus guarantee as described in claim 6, characterized in that, The conflict resolution strategy includes at least one of the following: Strategy 1: Revert to the conflict-free coordination result of the previous coordination cycle; Strategy 2: Trigger emergency coordination and re-execute S2 to S4; Strategy 3: Trigger system downgrade, reduce the number of agents and / or reduce the maximum speed of agents, and then re-execute S2 to S4.

8. The agent coordination method based on conflict detection and consensus guarantee in the computation process according to claim 1, characterized in that, The specific process of S6 is as follows: S6.1, Control all agents to execute according to the trajectory determined in the final coordination scheme; S6.2, calculates in real time the deviation between the actual position and the desired position of each agent, as well as the actual distance between any two agents; S6.3, trigger a security response based on the monitoring results of S6.2: If a state deviation is detected to exceed the first preset threshold but not the higher second preset threshold, then the corresponding agent is fine-tuned within the current execution cycle. If the detected state deviation exceeds the second preset threshold, emergency coordination is triggered, and S2 to S5 are re-executed; If the actual distance is detected to be less than the minimum safe distance minus the safety margin, an emergency braking action is triggered for the corresponding agent pair.

9. The agent coordination method based on conflict detection and consensus guarantee in the computation process according to claim 1, characterized in that, Also includes S7: The critical phase transition of a three-stage coordination architecture system is detected by finite size scaling analysis, and the critical point of the three-stage coordination architecture system is determined by calculating the cumulative amount of Binder under different agent numbers. When the computing capacity ratio monitored in real time in S1 exceeds the critical point, it is predicted that the three-stage coordination architecture system will fail. The corresponding self-healing strategy is triggered based on the predicted fault type.

10. An agent coordination system based on computational process conflict detection and consensus guarantee, used to execute the agent coordination method based on computational process conflict detection and consensus guarantee as described in any one of the claims, characterized in that, include: The system initialization and status monitoring module is used to initialize the three-stage coordination architecture and monitor the computing capacity ratio of the three-stage coordination architecture system in real time. The first processing stage module is used to perform multiple rounds of conflict detection with decreasing time steps on the confirmed agent trajectory based on the computing capacity ratio in the first processing stage, and generate a set of conflict-free confirmed trajectories. The third processing stage module is used to introduce conflict avoidance constraints in the optimization calculation of the trajectory to be optimized, with reference to the set of confirmed trajectories without conflict, and to generate an optimized trajectory that does not conflict with the set of confirmed trajectories. The consensus reaching and verification module is used to integrate the set of conflict-free confirmed trajectories, the optimized trajectories, and the pre-coordination strategy to form a final coordination scheme, and to perform final conflict verification on the final coordination scheme. If verification fails, the conflict resolution strategy is triggered. The execution and monitoring module is used to execute the final coordination scheme and monitor the operation status of the intelligent agent in real time. When an abnormal status or distance is detected, a safety response is triggered.

Images