A collision-free global path planning method for mobile robots

Abstract

This application provides a collision-free global path planning method for mobile robots, belonging to the field of mobile robot path planning. It addresses the problems in related technologies, such as reliance on statistical correlation leading to misjudgment of intent, low efficiency in role allocation, and high collision risk in dynamic scenarios. This method achieves agent intent reasoning by constructing a mental graph containing causal relationships, performs efficient role allocation based on quantum role superposition states, optimizes path search by combining a conflict risk field, triggers role collapse when the conflict risk exceeds a threshold, and ensures cooperation between adjacent agents through quantum entanglement cooperative states. This scheme improves the accuracy of intent prediction and the efficiency of role allocation, achieving collision-free and efficient global path planning in dynamic environments.

Description

Technical Field

[0001] This application relates to the field of mobile robot path planning, and more particularly to a collision-free global path planning method for mobile robots. Background Technology

[0002] Existing multi-agent path planning methods for mobile robots often rely on the statistical correlation of trajectory data for decision-making, making it difficult to distinguish between real causal relationships and random associations, which can easily lead to misjudgment of intent. Role allocation often adopts a serial trial-and-error mode, which is inefficient in high-density dynamic scenarios and has a high conflict rate, making it difficult to meet the needs of collision-free global path planning in complex environments. Summary of the Invention

[0003] This application provides a collision-free global path planning method for mobile robots, which can achieve efficient collision-free path planning for mobile robots in complex scenarios by combining causal cognition and quantum parallel decision-making.

[0004] In a first aspect, this application provides a collision-free global path planning method for a mobile robot, comprising: constructing a causal mental graph to realize intention reasoning between agents, wherein the causal mental graph includes causal relationships between agents, a set of target beliefs, a set of behavioral parameters, and a cooperative credit score; assigning roles to agents based on a quantum role superposition state, wherein the quantum role superposition state includes the probability amplitude of multiple roles; and generating a collision-free global path by combining the result of the intention reasoning with the result of the role assignment.

[0005] By adopting the above technical solutions, the intentions of intelligent agents can be accurately inferred with the help of causal mind graphs, overcoming the limitations of traditional statistical association. At the same time, the parallelism of quantum role superposition states is used to improve the efficiency of role allocation. The combination of the two provides a reliable foundation for generating collision-free global paths and effectively addresses the path planning challenges in complex scenarios.

[0006] Furthermore, the construction of the causal mental graph includes: extracting causal event pairs from the agent's historical trajectory data, wherein the causal event pairs include cause events and result events; performing statistical significance verification on the causal event pairs to filter out valid causal relationships that meet a preset significance threshold; and updating the causal mental graph based on the valid causal relationships.

[0007] By adopting the above technical solutions, the authenticity and validity of causal relationships in causal mind maps are ensured, the interference of false associations on intention reasoning is reduced, the accuracy of intention reasoning is improved, and a more reliable basis is provided for subsequent path planning.

[0008] Furthermore, the extraction of causal event pairs from the agent's historical trajectory data includes: ensuring that the timestamp of the causal event is earlier than the timestamp of the result event through a time synchronization mechanism, wherein the synchronization accuracy of the time synchronization mechanism is not lower than a preset synchronization time threshold.

[0009] By adopting the above technical solution, the temporal relationship of causal events is strictly guaranteed, causal reversal caused by time synchronization errors is avoided, the accuracy of the extracted causal event pairs is ensured, and the foundation for the effective construction of causal relationships is laid.

[0010] Furthermore, the intent reasoning includes: generating a counterfactual trajectory of the agent based on counterfactual reasoning, wherein the counterfactual trajectory is a possible predicted trajectory of the agent assuming that no specific cause event has occurred; and calculating a target probability distribution by combining the counterfactual trajectory with the actual trajectory of the agent, wherein the target probability distribution is used to characterize the potential target of the agent and the corresponding confidence level.

[0011] By adopting the above technical solution, counterfactual reasoning can compensate for the lack of consideration for non-occurring events in traditional intention reasoning, and more comprehensively analyze the agent's intention, making the calculation of the target probability distribution more accurate and improving the reliability of intention prediction.

[0012] Furthermore, the probability amplitude of each role in the quantum role superposition state is determined in the following way: the probability amplitude of each role is calculated based on the marginal contribution value of the agent, the role fitness, and the causal effect. The marginal contribution value is calculated using the Shapley value, and the role fitness is used to characterize the degree of matching between the agent's behavioral parameters and the role.

[0013] By adopting the above technical solutions, the probability amplitude of roles is determined by considering multiple factors, making the quantum role superposition state more consistent with the actual situation and cooperation needs of the intelligent agent, and improving the rationality and effectiveness of role allocation.

[0014] Furthermore, it also includes: when the conflict risk value between intelligent agents exceeds a preset risk threshold, measuring the quantum role superposition state and causing the quantum role superposition state to collapse into a determined role.

[0015] By adopting the above technical solutions, roles can be quickly determined in situations with high conflict risk, avoiding collisions caused by role uncertainty, balancing role exploration and collision avoidance, and improving the real-time performance and safety of path planning.

[0016] Furthermore, the conflict risk value is calculated through a conflict risk field, which comprehensively considers the spatial distance between agents and the target distribution entropy. The target distribution entropy is used to characterize the uncertainty of the potential target of the agent.

[0017] By adopting the above technical solutions, a comprehensive assessment of conflict risks is conducted, taking into account both spatial distance factors and the risks brought about by target uncertainty. This makes the calculation of conflict risk values ​​more comprehensive and accurate, providing a reliable basis for role collapse decision-making.

[0018] Furthermore, it also includes: constructing a quantum entangled cooperative state of adjacent agents, wherein the quantum entangled cooperative state contains a joint probability amplitude of complementary role pairs, wherein the complementary role pairs are combinations of roles capable of cooperating.

[0019] By adopting the above technical solutions, we can promote the collaborative work of neighboring intelligent agents, avoid conflicts at the role level, improve the efficiency of collaboration between intelligent agents, and reduce the coordination costs in path planning.

[0020] Furthermore, the construction of the causal mental graph also includes: dynamically updating the causal mental graph through a federated learning framework, which includes edge nodes and cloud nodes. The edge nodes are responsible for calculating local causal relationships, and the cloud nodes are responsible for aggregating global causal relationships.

[0021] By adopting the above technical solutions, efficient dynamic updates of causal mind maps can be achieved, taking into account both the timeliness of local computation and the comprehensiveness of global aggregation, thereby enhancing the adaptability of causal mind maps to environmental changes.

[0022] Furthermore, the generation of a collision-free global path includes: performing path search based on the A* algorithm, wherein the heuristic function of the A* algorithm integrates the path length and the integral value of the conflict risk field, and the integral value of the conflict risk field is the integral of the conflict risk value within a preset time interval.

[0023] By adopting the above technical solutions, path search not only considers path length but also fully incorporates conflict risk factors. The generated paths can effectively avoid collisions while ensuring efficiency, thus improving the safety and rationality of the paths.

[0024] In summary, this application has at least the following beneficial effects:

[0025] 1. A collision-free path planning method for mobile robots that combines causal cognition and quantum parallel decision-making is provided, which improves planning efficiency and safety;

[0026] 2. Through accurate causal reasoning and efficient role allocation, collaboration and intent understanding among intelligent agents are enhanced;

[0027] 3. Multi-dimensional optimization of path generation further improves the reliability and adaptability of the path.

[0028] It should be understood that the description in the Summary Section is not intended to limit the key or essential features of the embodiments of this application, nor is it intended to restrict the scope of this application. Other features of this application will become readily apparent from the following description. Attached Figure Description

[0029] The above and other features, advantages, and aspects of the embodiments of this application will become more apparent from the accompanying drawings and the following detailed description. In the drawings, the same or similar reference numerals denote the same or similar elements, wherein:

[0030] Figure 1 A schematic diagram of an exemplary operating environment in which embodiments of this application can be implemented is shown.

[0031] Figure 2 A flowchart of a collision-free global path planning method for a mobile robot according to an embodiment of this application is shown. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0033] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0034] This application provides a collision-free global path planning method for mobile robots, which effectively improves the path planning efficiency and collision-free performance of robot collaboration in complex scenarios by integrating causal cognition and quantum parallel decision-making technology.

[0035] This application discloses a collision-free global path planning method for mobile robots.

[0036] Figure 1 A schematic diagram of an exemplary operating environment in which embodiments of this application can be implemented is shown.

[0037] Reference Figure 1 The operating environment includes a mobile robot cluster, an environmental perception network, a computing and processing system, a communication transmission network, and a physical environment infrastructure. These components form a closed-loop support system through data interaction links.

[0038] A mobile robot cluster consists of multiple robot bodies with autonomous movement capabilities. Each robot is equipped with a drive chassis (for executing the planned path) and a central control unit (for running the core logic of the path planning algorithm), serving as the execution carrier for path planning.

[0039] The environmental perception network consists of a lidar, an event camera, and an inertial measurement unit (IMU). The lidar and event camera are deployed on the top of the robot to collect high-density environmental point clouds and dynamic event data, respectively. The IMU is integrated into the robot chassis to acquire motion state parameters. The three are connected to the central control unit through an internal data bus to provide environmental and self-state inputs for path planning.

[0040] The computing system adopts an edge-cloud collaborative architecture. Edge computing nodes (deployed in key locations in the physical environment) are connected to the robot control unit via wired links and are responsible for parallel processing of causal reasoning and quantum role state calculation. The cloud server interacts with the edge nodes through a communication network for the aggregation and updating of the global causal relationship library.

[0041] The communication transmission network includes a 5G-Advanced wireless communication module (integrated into the robot and edge nodes) and a fiber optic synchronization network. The former is used to transmit real-time trajectory data and role status information, while the latter enables cross-device nanosecond-level time synchronization to ensure the accuracy of the timing of causal events.

[0042] The physical environment must meet the requirements of a pre-defined spatial partition (e.g., a 50m × 50m grid), be equipped with a UWB positioning base station to provide a global coordinate reference, and ensure that the density of dynamic entities within the environment does not exceed 50 per m². Static obstacles must possess physical contours that can be identified by LiDAR, providing a modelable spatial basis for the path planning algorithm. These components interact through data to form a complete "perception-computation-decision-execution" chain, supporting the implementation of a collision-free global path planning method for mobile robots.

[0043] Figure 2 A flowchart illustrating a collision-free global path planning method for a mobile robot according to an embodiment of this application is shown. This method can be... Figure 1 The runtime environment is executed within it.

[0044] Reference Figure 2 The method specifically includes the following steps:

[0045] S1: Construct a causal mental graph to realize intention reasoning between agents. The causal mental graph includes causal relationships between agents, a set of goal beliefs, a set of behavioral parameters, and a collaborative credit score.

[0046] In this step, the method of constructing a causal mental graph includes: extracting causal event pairs from the agent's historical trajectory data, wherein the causal event pairs include cause events and result events; performing statistical significance verification on the causal event pairs to screen out effective causal relationships that meet a preset significance threshold; and updating the causal mental graph based on the effective causal relationships.

[0047] The extraction of causal event pairs from the agent's historical trajectory data includes: ensuring that the timestamp of the causal event is earlier than the timestamp of the result event through a time synchronization mechanism, wherein the synchronization accuracy of the time synchronization mechanism is not lower than a preset synchronization time threshold (e.g., 50ns, implemented based on the IEEE 1588 PTPv2 protocol). Specifically, causal event pairs are defined as triples. ,in For cause events (such as "battery level") "), For result events (such as "turning to the charging area"). The time difference between the two events , must meet (If the value is too short, it may be due to measurement error; if it is too long, other interference factors may exist.) By fusing the dynamic change sequence captured by the event camera with the position data from the lidar, precise extraction can be achieved. and Spatial coordinates and timestamps to ensure Calculation error .

[0048] When performing statistical significance verification on the causal event pair, the hypothesis testing framework is adopted: null hypothesis for" and "No causal relationship", alternative hypothesis for" lead to By calculating the causal effect value ,in This indicates that the event E was actively forced to occur (excluding interference from other confounding factors). This indicates that E should not occur. The causal effect value quantifies the difference in probability of B occurring when "intervention E occurs" and "intervention E does not occur" (reflecting the causal influence of E on B); when And the corresponding value At that time, refuse This is considered a valid causal relationship. A preset significance threshold is used. Typically, a value of 0.4 is used (based on calibration using 100,000 sets of historical data to ensure...). False positive rate at confidence level ).

[0049] Goal and Belief Set Its construction requires combining counterfactual reasoning results, and its core is computational intelligent agents. right Potential target probability distribution ,Right now For intelligent agents For intelligent agents Potential target probability (i.e.) by (The probability of being the actual target). For each candidate target (the first in the discretized target space) Candidate targets (such as target points / regions on a map) are processed by a variational autoencoder (VAE). Historical trajectory Encoding yields latent vectors The decoder outputs the initial target probability. Combining this with the counterfactual correction term, the final target probability is:

[0050]

[0051] in, The determination method includes statistical baseline probability, calculated through a variational autoencoder (VAE): using the encoder of the VAE... Historical trajectory Encoding yields latent vectors (Key features of the compressed trajectory), output by the VAE decoder. (measure and (match degree), and then through Convert to probability; ICE To intervene in causal effects, quantify "coercion". by For the target "trajectory" Degree of impact (reflecting the target) (causal relationship with trajectory) Let be the magnification factor. Used to balance "statistical correlation" "Historical connections" and "causal determinism" (intervention effects embodied in ICE); A dummy variable for iterating through all candidate targets (and molecules) (Belonging to the same target set, used for normalizing probabilities).

[0052] Behavioral parameter set Includes intelligent agents Motion characteristic parameters, such as velocity distribution , turning curvature These parameters are extracted from historical trajectories using maximum likelihood estimation and are updated every 30 minutes based on new data, using the following update formula: For smoothing coefficients, (The average velocity of the new trajectory) ensures that the parameters are dynamically adjusted as the behavior pattern changes.

[0053] Collaborative Credit Score The initial value is set to 0.5, and it updates dynamically based on the interaction results: when Actual behavior and based on When the predictions are consistent, When unforeseen conflicts or collaboration failures occur, This score directly affects the model aggregation weights in subsequent federated learning; agents with higher credit ratings have a higher weight in the global aggregation of local causal relationships.

[0054] In this step, the construction of the causal mental graph further includes: dynamically updating the causal mental graph using a federated learning framework. The federated learning framework includes edge nodes and cloud nodes. The edge nodes are responsible for calculating local causal relationships, and the cloud nodes are responsible for aggregating global causal relationships. Specifically, each edge node (corresponding to a 50m × 50m spatial partition) maintains a local causal relationship database. This includes valid causal pairs within the partition over the past 30 minutes. Edge nodes compute local model parameters every 5 minutes. (like probability distribution parameters The distribution parameters are uploaded to the cloud via a secure aggregation protocol. Cloud nodes aggregate global parameters using a weighted average method. The weight ( This represents the average collaborative credit of agents within a partition, ensuring that local knowledge from high-credit regions accounts for a higher proportion of the global model. (The aggregated data is then used.) It will synchronize back to each edge node to update the local causal mental graph. The total communication volume of the whole process is... Each iteration (achieved through model parameter compression) has a latency of <1 second.

[0055] In this step, the intention reasoning includes: generating a counterfactual trajectory of the agent based on counterfactual reasoning, wherein the counterfactual trajectory is a possible predicted trajectory of the agent assuming that no specific cause event has occurred; and calculating a target probability distribution by combining the counterfactual trajectory with the actual trajectory of the agent, wherein the target probability distribution is used to characterize the potential target of the agent and the corresponding confidence level.

[0056] Generate counterfactual trajectories At that time, it was described as an "uncaused event". Given this premise, the solution is obtained using a modified A* algorithm:

[0057]

[0058] in, For intelligent agents The counterfactual trajectory (simulating the occurrence of "causeless event E") "The trajectory to be traveled" (used for causal intervention analysis). The mathematical operation is "searching for the candidate trajectory P that minimizes the summation term (the core is finding the "optimal counterfactual trajectory")". The position of candidate trajectory P at time step t is determined by iterating through all possible trajectories and verifying whether they meet the constraints. For intelligent agents The actual trajectory at time t (historical observation data, serving as a benchmark for "closeness to reality"). The Euclidean distance is used to calculate the positional deviation between the candidate trajectory and the actual trajectory, ensuring that the counterfactual trajectory "does not deviate from the historical trend". For the preset counterfactual loss function (when With "nothing" The greater the deviation from the expected path, the higher the CF value. Preset counterfactual weights (to ensure that the trajectory closely follows historical trends and conforms to "no (constraints).

[0059] Combining counterfactual trajectories with actual trajectories When calculating the target probability distribution, the initial target probability is adjusted by comparing the deviations of the two at key decision points (such as intersections). For example, if the counterfactual trajectory shows... Originally it was going to However, the actual trajectory is due to Turning ,but Confidence improvement ( For correction factor, The maximum possible deviation is used to normalize d, so that This ultimately forms a probability distribution containing 5-8 candidate targets, and satisfies... ,in, For intelligent agents right goal The confidence level increase (due to trajectory deviation caused by E, the target probability needs to be corrected). The total deviation between the actual trajectory and the counterfactual trajectory (the sum of deviations at key decision points such as intersections, reflecting the degree of influence of E). For intelligent agents right by The probability of the target (after correction, it must satisfy a valid distribution with a total confidence level of 1).

[0060] Through the above process, causal mind mapping can dynamically integrate causal relationships, goal beliefs, behavioral characteristics, and collaboration history, providing a quantitative basis for intention reasoning among agents and significantly improving the accuracy of intention prediction in high-density dynamic scenarios.

[0061] S2: Assign roles to agents based on quantum role superposition states, wherein the quantum role superposition states contain the probability amplitudes of multiple roles.

[0062] The probability amplitude of each role in the quantum role superposition state is determined in the following way: the probability amplitude of each role is calculated based on the marginal contribution value of the agent, the role fitness, and the causal effect. The marginal contribution value is calculated using the Shapley value, and the role fitness is used to characterize the degree of matching between the agent's behavioral parameters and the role.

[0063] Specifically, the quantum role superposition state is the parallel representation of multiple potential roles by an agent during the decision-making process, and its mathematical form is:

[0064]

[0065] in, The quantum state vector (right vector) describes the quantum state of the system. For a preset set of characters (usually containing) Navigator, Follower, Coordinator, Obstacle Avoider, Standby (A core role, which can be expanded according to the scenario) For the role ground state, The probability amplitude for the corresponding role satisfies the normalization condition. (Ensure the total probability is 1). Probability amplitude Size directly reflects the character The probability of being selected needs to be calculated by integrating three factors: marginal contribution value, role suitability, and causal effect, to form a multi-dimensional decision-making basis.

[0066] Calculation of marginal contribution value (Shapley value): Agent The marginal contribution value is obtained through the Shapley value. Quantification, at its core, is evaluation. The average value increment resulting from cooperation among all possible subsets of agents is given by the following formula:

[0067]

[0068] in, For intelligent agents The Shapley value (a quantitative indicator of marginal contribution). The set of agents in the current task. for Not included a subset of For subset The number of intelligent agents, for The total number of intelligent agents ( ), Characterization The incremental value of the subset after inclusion For subset The value of completing the task (defined as the reciprocal of the total path length planned by all agents within the subset, i.e.) (The shorter the path, the higher the value). Shapley value The larger, the more it indicates The more irreplaceable a entity is in collaboration, the higher its probability of playing a key role (such as a leader) will be.

[0069] Quantification of character fit: Character fit Used to measure intelligent agents behavioral parameter set With the character The degree of matching between needs and requirements is calculated using a weighted scoring method:

[0070]

[0071] in, These are behavioral parameter dimensions (such as speed stability, steering agility, and collaborative response speed). For the role For the first The weights of each parameter (e.g., the weight of "Navigator" on speed stability) Weighting of steering flexibility ), for The Each behavior parameter value, For the role For the first The optimal values ​​of each parameter For the normalization function (mapping parameter values ​​to) (The closer the interval is to the optimal value, the higher the score). For example, if the role... As the "navigator", its optimal speed stability value (variance ), Speed ​​stability ,but This indicates a high degree of compatibility.

[0072] Integration of causal effects: causal effects Quantify roles With intelligent agents The strength of the causal relationship of the current task objective is derived from the valid causal relationships in the causal mental map:

[0073]

[0074] in, To and Take on a role The set of related causal event pairs (such as the causal pair of "taking on the role of a navigator" and "reduced task completion time"). The causal effect value of this causal pair (defined in step S1) ). The larger the value, the more important the role. The stronger the causal relationship with the current task objective, the greater its probability amplitude needs to be.

[0075] Final calculation of probability amplitude: Considering the above three factors, the role probability amplitude The calculation formula is:

[0076]

[0077] To meet the normalization conditions, the probability amplitude of all characters needs to be normalized:

[0078]

[0079] Through this formula, roles with high marginal contribution, good fit, and strong causal correlation will obtain higher probability amplitudes, enabling the quantum role superposition state to accurately reflect the agent's collaborative potential and task requirements, and providing a reliable initial state for subsequent role collapse (when the risk of conflict exceeds the threshold).

[0080] Furthermore, the parallelism of quantum role superposition states is manifested in all roles. In terms of synchronous representation, the agent does not need to evaluate each role individually, but explores all possibilities simultaneously through superposition. Its computational efficiency increases linearly with the number of roles (thanks to the parallel processing capability of the quantum simulation acceleration unit), solving the efficiency bottleneck of traditional serial role trial and error in high-density scenarios.

[0081] S3: Combine the results of the intent reasoning with the results of the role assignment to generate a collision-free global path.

[0082] The specific method of this step includes: performing path search based on the A* algorithm, wherein the heuristic function of the A* algorithm integrates the path length and the integral value of the conflict risk field, and the integral value of the conflict risk field is the integral of the conflict risk value within a preset time interval.

[0083] The conflict risk field is used to quantify the collision risk between agents. Its calculation needs to comprehensively consider spatial distance and intent uncertainty, and the formula is as follows:

[0084]

[0085] in, For intelligent agents and At the point of spacetime The intensity of the conflict risk field, For spatial coordinates, For timestamps; coordinates With intelligent agents Predicted path In time The Euclidean distance; The preset weighting coefficients (spatial distance has a greater impact on immediate risk); The pre-acquired spatial attenuation coefficient (taken as 1.5 times the robot radius, e.g., when the radius is 30cm) for right Target distribution entropy ( A higher entropy value indicates The greater the uncertainty of the objective; As a smoothing factor, avoid The denominator is 0.

[0086] The integral value of the conflict risk field is the accumulation of risk values ​​within a preset time interval, used to assess the conflict risk throughout the entire path. The calculation formula is as follows:

[0087] Integral field of conflict risk field =

[0088] in, For the agent to reach the current node Time, To reach the target Time (by node) arrive The straight-line distance is estimated by dividing the average speed, i.e. for (historical average speed) For nodes on the path In time The corresponding position (obtained through linear interpolation). The larger this integral value, the closer the path is to the target location within the preset time interval. The higher the cumulative risk of a collision.

[0089] Heuristic function of the improved A* algorithm The formula combines path length and conflict risk integral, and is as follows:

[0090]

[0091] in, The value is the agent The heuristic function value at node n, original The traditional A* Euclidean distance heuristic function original The preset risk weighting coefficients (calibrated through 1000 sets of simulation experiments, ensuring that the proportion of risk factors in the total cost is approximately...) (balancing path efficiency and safety). This represents the summation of risk integrals over all other agents, ensuring a global collision-free environment.

[0092] The total cost function for path search is:

[0093]

[0094] in, To start from the starting point To the node Actual path length; RoleCost Quantify nodes based on the cost of character adaptation. With intelligent agents Current role The degree of matching (e.g., the "Navigator" role has a lower cost at main channel nodes, and the "Obstacle Avoider" role has a lower cost at edge region nodes) is calculated using the formula RoleCost. For nodes (Functional attributes of the area) This is the role weight coefficient, ensuring that the role allocation result affects path preference.

[0095] During the path generation process, it is necessary to dynamically adjust the path based on the intent reasoning results in real time. :when Goal and Belief Set During updates (e.g.) (From 0.3 to 0.8), its predicted path was recalculated. And update the conflict risk field simultaneously. and heuristic functions If the risk score of a candidate path exceeds a preset threshold... (like ,correspond Collision probability at confidence level If the path is not found, then the path should be pruned directly to avoid invalid searches.

[0096] Using the above method, the generated path can not only ensure that the agent reaches the target efficiently, but also avoid potential conflicts in advance based on intent reasoning and role division, thus achieving global collision-free cooperative movement.

[0097] The method further includes: when the conflict risk value between intelligent agents exceeds a preset risk threshold, measuring the quantum role superposition state and causing the quantum role superposition state to collapse into a determined role.

[0098] In this step, the conflict risk value is calculated through a conflict risk field, which comprehensively considers the spatial distance between agents and the target distribution entropy. The target distribution entropy is used to characterize the uncertainty of potential targets of agents.

[0099] Specifically, the calculation of the conflict risk value is based on the conflict risk field, selecting the maximum risk value within a preset time window (e.g., 5 seconds) for the agent as the judgment criterion. The formula is as follows:

[0100]

[0101] in, For the current time, Risk prediction window (based on the agent's average velocity) (Set to cover a 5m movement range). For intelligent agents In time The predicted location, for and The conflict risk field value at this spatiotemporal point (calculated in the same way as step S3) .when To preset risk thresholds, calibration was performed using 100,000 collision case studies to ensure... When a potential collision can be triggered, the measurement collapse of the quantum role superposition state is triggered.

[0102] Target distribution entropy It plays a key adjusting role in the calculation of conflict risk value, and its formula is as follows:

[0103]

[0104] in The number of candidate targets. For intelligent agents right by The confidence level of the target (from the target belief set in step S1). The entropy value ranges from... ,when Time (indicating) The target is highly uncertain; the target uncertainty term in the conflict risk field. It will increase, make It is easier to exceed the threshold This allows for early collapse to address scenarios with high uncertainty.

[0105] Quantum role superposition state The measurement process essentially involves calculating the squared magnitude of the probability amplitude for each role. Choose the role with the highest probability as the collapse result:

[0106]

[0107] in The roles are determined after the collapse. To ensure cooperation between roles, the collapse process needs to refer to the role states of neighboring agents, especially through quantum entanglement cooperative states. Achieved role association:

[0108]

[0109] in For sets of complementary role pairs (e.g., {(Navigator, Follower), (Coordinator, Executor)}), For the joint probability amplitude. When Collapse into a character hour, It will collapse into a state of entanglement first through the correlation. Complementary roles ,Right now and To avoid conflicts at the role allocation level (e.g., two agents will not collapse into "navigator" at the same time).

[0110] Collapsed Character This will be incorporated as a constraint into subsequent path planning, by adjusting the role adaptation cost. (See step S3) Implement dynamic updates to path preferences: For example, an agent that collapses into the role of an "obstacle avoider" will prioritize edge region nodes (with lower RoleCost values) in its path search, while a "navigator" will prioritize main channel nodes. This linkage mechanism between roles and paths ensures that the decisions made after collapse can directly affect the generation of collision-free paths, forming a closed loop of "conflict detection - role determination - path adjustment".

[0111] Using the above method, the collapse of the quantum role superposition state retains the efficiency advantage of parallel exploration and can quickly converge to a determined role in high-risk scenarios, balancing the flexibility and security of planning. Furthermore, by combining multi-factor assessment of the conflict risk field with the synergistic constraint of quantum entanglement, the role collapse result is more consistent with the goal of global collision-free operation, significantly reducing the collision probability in high-density scenarios.

[0112] The method further includes: constructing a quantum entangled cooperative state of neighboring agents, wherein the quantum entangled cooperative state contains a joint probability amplitude of complementary role pairs, wherein the complementary role pairs are combinations of roles capable of cooperating.

[0113] The method further includes: constructing a quantum entangled cooperative state of neighboring agents, wherein the quantum entangled cooperative state contains a joint probability amplitude of complementary role pairs, wherein the complementary role pairs are combinations of roles capable of cooperating.

[0114] The determination of the complementary role pairs is based on the compatibility of role functions and historical collaboration data, specifically: from a preset set of roles. The middle screening meets the requirements Role combination To form a set of complementary roles .in The character compatibility rating is calculated using the following formula: For functional compatibility (e.g., the functional complementarity score for "Navigator" and "Follower" is 0.9, and for "Dual Navigator" it is 0.2). The historical collaboration success rate (calculated based on the proportion of conflict-free collaborations in the past 100 collaborations). For functional weights, A compatibility threshold (ensuring the selected roles contribute to collaboration efficiency) ).

[0115] The criterion for determining neighboring agents is the intersection of spatial distance and communication range: when agents... and Euclidean distance ( (Based on the effective transmission distance setting of the communication module), and collaborative credit scoring. When the cooperative credit score from step S1 is used to ensure the reliability of historical cooperation, the agent is identified as a neighboring agent, triggering the construction of a quantum entangled cooperative state.

[0116] The mathematical form of a quantum entangled cooperative state is a multi-particle superposition state:

[0117]

[0118] in The joint probability amplitude satisfies the normalization condition. The calculation of the joint probability amplitude integrates role suitability and collaborative credit:

[0119]

[0120] for For the role The degree of fit (from step S2). for and The collaborative credit score (from step S1) is used to give complementary role pairs with high fit and good collaborative credit a higher joint probability amplitude.

[0121] The dynamic update mechanism of quantum entangled cooperative states is as follows: every time... (Based on the agent's movement speed and the frequency of environmental changes), the distance between neighboring agents is recalculated. Compatibility rating ,like or If the entanglement is resolved, the state is untangled; otherwise, the probability amplitude is updated jointly based on the latest role suitability and collaborative credit. This ensures that the entangled state always reflects the current collaborative needs.

[0122] During role collapse, quantum entangled cooperative states achieve role coordination among neighboring agents through constraints on joint probability amplitudes: when The quantum role superposition state collapses into hour, The result of the collapse Must meet And the collapse probability is Positive correlation, that is:

[0123]

[0124] in To and A complementary set of roles. This mechanism ensures that the roles of neighboring agents always form an effective cooperative combination, fundamentally avoiding path collisions caused by role conflicts.

[0125] Based on the foregoing, this step's method constructs a causal mental graph containing causal relationships, a set of target beliefs, a set of behavioral parameters, and a collaborative credit score. This allows for the accurate differentiation between causal and random associations in agent behavior. Combined with counterfactual reasoning, it corrects the target probability distribution, thereby reducing the risk of path conflicts caused by misjudgment of intent. By using quantum role superposition states to represent multiple roles in parallel, and integrating marginal contribution values, role suitability, and causal effect probability amplitudes, role allocation aligns with both agent collaborative potential and task requirements, avoiding the efficiency losses of traditional serial trial-and-error. A conflict risk field integrates spatial distance and target uncertainty to quantify collision risk, and the risk integral of the A* algorithm heuristic function is optimized within a preset time interval, enabling path search to avoid accumulated risks while maintaining efficiency. When the conflict risk exceeds a threshold, a role is determined by quantum role superposition state collapse locking, combined with quantum entanglement cooperative states to constrain the complementary roles of adjacent agents, ensuring collaborative compatibility at the role level and reducing path intersections caused by role conflicts at their root. These techniques form a closed loop from intent understanding, role allocation, risk assessment to conflict resolution, ultimately achieving collision-free and efficient global path planning for agents in dynamic environments.

[0126] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, because according to the embodiments of this application, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to this application.

[0127] In summary, this application has at least the following beneficial effects:

[0128] 1. By using causal mind mapping and counterfactual reasoning, we can accurately distinguish between causal and random associations in agent behavior, improve the accuracy of intent prediction, and reduce path conflicts caused by misjudgment of intent from the root.

[0129] 2. By leveraging the parallelism of quantum role superposition states and the complementary constraints of entangled cooperative states, role allocation and coordination can still be efficiently completed in high-density scenarios, breaking through the efficiency bottleneck of traditional serial role trial and error and improving the flexibility of multi-agent cooperation.

[0130] 3. The improved A* algorithm integrates the spatiotemporal integral of the conflict risk field, enabling the generated path to avoid the cumulative collision risk within a preset time interval while ensuring arrival efficiency, thus achieving a balance between safety and efficiency.

[0131] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of disclosure in this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the foregoing disclosed concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this application.

Claims

1. A collision-free global path planning method for a mobile robot, characterized in that, include: A causal mental graph is constructed to enable intention reasoning among agents. The causal mental graph includes causal relationships among agents, a set of goal beliefs, a set of behavioral parameters, and a collaborative credit score. Role allocation for intelligent agents is based on quantum role superposition states, where each quantum role superposition state contains the probability amplitudes of multiple roles. A collision-free global path is generated by combining the results of the intent reasoning and the role assignment. The construction of the causal mind map includes: Extract causal event pairs from the historical trajectory data of the agent, wherein the causal event pairs include causal events and result events; The causal event pairs are statistically significant, and valid causal relationships that meet the preset significance threshold are selected. The causal mental map is updated based on the effective causal relationships. The probability amplitude of each role in the quantum role superposition state is determined in the following way: The probability amplitude of each role is calculated based on the marginal contribution value of the agent, the role fit, and the causal effect. The marginal contribution value is calculated using the Shapley value, and the role fit is used to characterize the degree of matching between the agent's behavioral parameters and the role. The generation of a collision-free global path includes: Path search is performed based on the A* algorithm. The heuristic function of the A* algorithm integrates the path length and the integral value of the conflict risk field. The integral value of the conflict risk field is the integral of the conflict risk value within a preset time interval. The conflict risk field is used to quantify the collision risk between agents. The conflict risk value is calculated through the conflict risk field, which comprehensively considers the spatial distance between agents and the target distribution entropy. The target distribution entropy is used to characterize the uncertainty of potential targets for agents. The formula is: ； in The number of candidate targets. For intelligent agents right by The confidence level of the target comes from the set of target beliefs in intentional reasoning; When the conflict risk value between intelligent agents exceeds a preset risk threshold, the quantum role superposition state is measured, causing the quantum role superposition state to collapse into a definite role.

2. The collision-free global path planning method for mobile robots according to claim 1, characterized in that, The extraction of causal event pairs from the agent's historical trajectory data includes: The time synchronization mechanism ensures that the timestamp of the cause event is earlier than the timestamp of the result event, and the synchronization accuracy of the time synchronization mechanism is not lower than a preset synchronization time threshold.

3. The collision-free global path planning method for a mobile robot according to claim 1, characterized in that, The intent reasoning includes: The counterfactual trajectory of the agent is generated based on counterfactual reasoning. The counterfactual trajectory is the possible predicted trajectory of the agent assuming that no specific cause event has occurred. The target probability distribution is calculated by combining the counterfactual trajectory with the agent's actual trajectory. The target probability distribution is used to characterize the agent's potential targets and corresponding confidence levels.

4. The collision-free global path planning method for a mobile robot according to claim 1, characterized in that, Also includes: When the conflict risk value between intelligent agents exceeds a preset risk threshold, the quantum role superposition state is measured, causing the quantum role superposition state to collapse into a definite role.

5. The collision-free global path planning method for a mobile robot according to claim 4, characterized in that, The conflict risk value is calculated through a conflict risk field, which comprehensively considers the spatial distance between agents and the target distribution entropy. The target distribution entropy is used to characterize the uncertainty of potential targets of agents.

6. The collision-free global path planning method for a mobile robot according to claim 1, characterized in that, Also includes: Construct a quantum entangled cooperative state of neighboring agents, wherein the quantum entangled cooperative state contains the joint probability amplitude of complementary role pairs, and the complementary role pairs are combinations of roles capable of cooperating.

7. The collision-free global path planning method for a mobile robot according to claim 1, characterized in that, The construction of the causal mind map also includes: The causal mental graph is dynamically updated using a federated learning framework, which includes edge nodes and cloud nodes. The edge nodes are responsible for calculating local causal relationships, while the cloud nodes are responsible for aggregating global causal relationships.

Images