Multi-rotor aircraft cluster collaborative obstacle avoidance control method and system

Abstract

The invention discloses a multi-rotor aircraft cluster collaborative obstacle avoidance control method and system, and the method comprises the following steps: obtaining the data of each multi-rotor aircraft and a surrounding obstacle in a cluster, and calculating the local threat degree value of each aircraft and the obstacle; acquiring neighbor data to calculate a neighbor influence coefficient, and outputting a dynamically adjusted collaborative threat diffusion value according to the local threat degree value of the aircraft and the neighbor influence coefficient; the collaborative threat diffusion value is combined with target point attraction and cluster cohesion parameters, and a fuzzy reasoning algorithm is input to carry out multi-target decision to generate a three-axis behavior decision vector; the decision is that threat levels are divided by a clustering algorithm for obstacle avoidance, a local obstacle avoidance strategy is generated according to an area where an aircraft is located and aircraft power constraints are combined, and a fault-tolerant set is determined; the probability of potential conflicts among obstacle avoidance paths of the aircrafts is calculated through a collision detection algorithm, and decision compensation is carried out by adopting a game theory algorithm when collaborative conflicts exist. According to the invention, the precision of collaborative obstacle avoidance of the multi-rotor aircraft cluster is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of aircraft, in particular to a multi-rotor aircraft cluster cooperative obstacle avoidance control method and system. BACKGROUND

[0002] With the rapid development of unmanned aerial vehicle technology, multi-rotor aircraft has been widely used in logistics distribution, disaster relief, agricultural plant protection, aerial performance and urban air traffic and other fields. In these complex task scenarios, a single unmanned aerial vehicle is difficult to meet the requirements of large range, high efficiency and strong robustness, so multi-rotor aircraft cluster becomes a research hotspot. However, in the dense dynamic environment, the realization of cluster autonomous flight faces great challenges, and cooperative obstacle avoidance control is one of the core problems. Traditional obstacle avoidance methods are mostly designed for single machines, and it is difficult to deal with multi-target conflict and cooperative decision-making problems in cluster environment. Based on this, the present application provides a multi-rotor aircraft cluster cooperative obstacle avoidance control method and system. SUMMARY

[0003] The present application provides a multi-rotor aircraft cluster cooperative obstacle avoidance control method, characterized in that it comprises:

[0004] S10, acquiring real-time data of each multi-rotor aircraft in the cluster and sensing distance data of surrounding obstacles, and determining local threat degree values of each aircraft and obstacles by using a fuzzy comprehensive evaluation algorithm;

[0005] S20, calculating neighbor influence coefficients according to neighbor position and speed data obtained from the communication topology in the cluster, and outputting a dynamically adjusted cooperative threat diffusion value according to the local threat degree value of the aircraft itself and the neighbor influence coefficient;

[0006] S30, combining obstacle point attraction and cluster cohesion parameters with the cooperative threat diffusion value, inputting a fuzzy reasoning system to generate a three-axis behavior decision vector for multi-target decision-making;

[0007] S40, if the decision is obstacle avoidance, using a clustering algorithm to divide areas with different threat levels, generating a local obstacle avoidance strategy according to the area where the aircraft is located and the aircraft's kinetic constraints, and determining a fault-tolerant set;

[0008] S50, based on the global environment map, calculating the potential conflict probability between the obstacle avoidance paths of each aircraft by a collision detection algorithm, and if there is a cooperative conflict, using a game theory algorithm for decision compensation to solve the conflict and maintain the cooperative flight of the cluster.

[0009] The multi-rotor aircraft cluster cooperative obstacle avoidance control method as described above, wherein the fuzzy comprehensive evaluation algorithm is used to determine the local threat degree values of each aircraft and obstacles. Specifically, the following sub-steps are included:

[0010] According to the aircraft and obstacle position, proximity index, motion tendency angle index are calculated, and environment complexity index is calculated according to environment information;

[0011] The proximity index, motion tendency angle index, and environment complexity index are used as input factors for fuzzy comprehensive evaluation, and the local threat degree value is determined by using the Mamdani type fuzzy reasoning mechanism.

[0012] The multi-rotor aircraft cluster cooperative obstacle avoidance control method as described above, wherein the neighbor influence coefficient is calculated according to the neighbor position and speed data obtained from the communication topology in the cluster, and the dynamically adjusted cooperative threat diffusion value is output according to the local threat degree value of the aircraft itself and the neighbor influence coefficient. Specifically, the following sub-steps are included:

[0013] Based on the communication topology, neighbor position and speed information is obtained, and relative distance attenuation effect and motion direction consistency are fused to generate neighbor influence coefficient;

[0014] The local threat degree of the neighbor and its influence coefficient are weighted and aggregated, and the stability index of the motion state is combined to calculate the neighborhood influence adjustment amount;

[0015] The self-threat degree and neighborhood influence adjustment amount are enhanced by a nonlinear coupling function, a cooperative threat value with diffusion characteristics is constructed, and dynamic guidance of the cluster obstacle avoidance behavior by high-threat individuals is realized.

[0016] The multi-rotor aircraft cluster cooperative obstacle avoidance control method as described above, wherein the cooperative threat diffusion value is combined with obstacle point attraction and cluster cohesion parameters, and input into a fuzzy reasoning system to generate a three-axis behavior decision vector for multi-objective decision making. Specifically, the following sub-steps are included:

[0017] Compare the cooperative threat diffusion value with the obstacle point attraction value to construct a dynamic obstacle avoidance trigger flag to determine whether the current aircraft is in an emergency obstacle avoidance demand state;

[0018] Generate a multi-behavior candidate set through direction modeling and weight evaluation algorithm;

[0019] Generate a three-axis behavior decision vector through fuzzy reasoning and vector fusion algorithm.

[0020] The multi-rotor aircraft cluster cooperative obstacle avoidance control method as described above, wherein the local obstacle avoidance strategy is generated according to the region where the aircraft is located and the dynamic constraints of the aircraft. Specifically, the following sub-steps are included:

[0021] According to the danger level of the region where the aircraft is located, the three-axis behavior decision vector, and the obstacle size, match the obstacle avoidance behavior in the multi-rotor aircraft obstacle avoidance behavior library;

[0022] The matched obstacle avoidance behavior is determined according to the actual situation of the obstacle and the dynamic constraint of the aircraft.

[0023] The multi-rotor aircraft cluster cooperative obstacle avoidance control method as described above, wherein the matched obstacle avoidance behavior is determined according to the actual situation of the obstacle and the dynamic constraint of the aircraft. Specifically, the following sub-steps are included:

[0024] The motion amplitude is scaled based on the actual size of the obstacle, the behavior direction vector is fine-tuned according to the relative position of the obstacle during the aircraft travel, and the maneuver intensity is intensified according to the emergency degree of the obstacle threat;

[0025] Meanwhile, the real-time residual capacity state of the aircraft and the dynamic limit are fused to generate refined obstacle avoidance instructions that meet the current environmental constraints.

[0026] The multi-rotor aircraft cluster cooperative obstacle avoidance control method as described above, wherein based on the global environment map, the potential conflict probability between the obstacle avoidance paths of each aircraft is calculated by a collision detection algorithm, and if there is a cooperative conflict, a game theory algorithm is used for decision compensation to solve the conflict and maintain the cooperative flight of the cluster. Specifically, the following sub-steps are included:

[0027] By discretizing the aircraft obstacle avoidance path and detecting the close-range conflict points at the same time point, a conflict probability matrix between the aircraft pairs is generated by combining the weighted calculation of the regional danger coefficient;

[0028] A potential game model of differentiated payoff function is constructed, and the heading correction and speed compensation instructions of the aircraft are output by solving the Nash equilibrium through iteration optimization;

[0029] After executing the compensation instructions, the conflict probability is monitored at a fixed period, and local re-planning is dynamically triggered until the global conflict probability is continuously below the set threshold, ensuring that the cluster reaches the target without conflict.

[0030] The application also provides a multi-rotor aircraft cluster cooperative obstacle avoidance control system, comprising:

[0031] The acquisition data and threat calculation module acquires real-time data of each multi-rotor aircraft in the cluster and the sensing distance data of the surrounding obstacles, and determines the local threat degree value of each aircraft and the obstacle by using a fuzzy comprehensive evaluation algorithm;

[0032] The neighbor influence and cooperative threat module calculates the neighbor influence coefficient according to the neighbor position and speed data obtained from the communication topology in the cluster, and outputs the dynamically adjusted cooperative threat diffusion value according to the local threat degree value of the aircraft and the neighbor influence coefficient;

[0033] The decision generation module combines the cooperative threat diffusion value with the obstacle point attraction and cluster cohesion parameters, inputs the fuzzy reasoning system to generate a three-axis behavior decision vector for multi-objective decision making.

[0034] The division and generation strategy module: the clustering algorithm is used to divide the areas with different threat levels, and the local obstacle avoidance strategy is generated according to the area where the aircraft is located and the dynamic constraints of the aircraft, and the fault tolerance set is determined.

[0035] The conflict detection and decision compensation module: based on the global environment map, the potential conflict probability between the obstacle avoidance paths of each aircraft is calculated through the collision detection algorithm, the game theory algorithm is used for decision compensation when there is a cooperative conflict, the conflict is solved and the cooperative flight of the cluster is maintained.

[0036] The application has the advantages that the precision of the multi-rotor aircraft cluster cooperative obstacle avoidance is improved. BRIEF DESCRIPTION OF DRAWINGS

[0037] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments described in the present application, and other drawings can also be obtained by those skilled in the art based on these drawings.

[0038] Figure 1 It is a multi-rotor aircraft cluster cooperative obstacle avoidance control method flow chart provided by the first embodiment of the present application.

[0039] Figure 2 It is a multi-rotor aircraft cluster cooperative obstacle avoidance control system schematic diagram provided by the second embodiment of the present application. DETAILED DESCRIPTION

[0040] The technical solutions in the embodiments of the present application will be described clearly and completely in the following with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0041] Embodiment one

[0042] As shown in Figure 1 The first embodiment of the present application provides a multi-rotor aircraft cluster cooperative obstacle avoidance control method, which comprises:

[0043] S10, the real-time data of each multi-rotor aircraft in the cluster and the sensing distance data of the surrounding obstacles are obtained, and the fuzzy comprehensive evaluation algorithm is used to determine the local threat degree value of each aircraft and the obstacle.

[0044] The motion state parameters of the aircraft itself are measured according to an inertial measurement unit of the multi-rotor aircraft. The position, speed and attitude information of the aircraft are acquired, including the coordinate position of the aircraft in space, linear speed, acceleration, angular speed, pitch angle, roll angle and yaw angle, etc.

[0045] The laser radar carried by the multi-rotor aircraft transmits and receives tens of thousands of laser pulses per second, thereby constructing three-dimensional point cloud data of the surrounding environment. Through processing and analysis of the point cloud data, the distance between the aircraft and the obstacle, the approximate shape, size and relative position data of the obstacle relative to the aircraft are obtained. In combination with the camera of the multi-rotor aircraft, the computer vision algorithm is used, including feature extraction, target detection and recognition technology, to identify the type of obstacle.

[0046] The specific steps of calculating the local threat degree value are as follows:

[0047] S11, according to the position of the aircraft and the obstacle, the proximity index and the motion trend angle index are calculated, and the environmental complexity index is calculated according to the environmental information.

[0048] According to the real-time three-dimensional position coordinates of each multi-rotor aircraft and the position of the obstacle detected by the airborne sensor, the physical distance between the aircraft and the obstacle is calculated according to the Euclidean distance formula, which is used as the basic data of original threat perception. The physical distance is mapped using an S-shaped membership function. The smaller the distance, the closer the value of the proximity index after mapping is to 1, indicating a higher threat. According to the current speed vector of the aircraft and the direction vector from the current position to the obstacle, the cosine value of the included angle is calculated, and the motion trend angle index is mapped by linear transformation. This index reflects whether the aircraft is moving towards the obstacle. The factors affecting the flight state of the aircraft, such as environmental visibility, weather, wind direction and wind speed, are weighted and calculated as the environmental complexity index.

[0049] S12, the proximity index, the motion trend angle index and the environmental complexity index are used as input factors for fuzzy comprehensive evaluation, and the local threat degree value is determined by using the Mamdani type fuzzy reasoning mechanism.

[0050] The proximity index, the motion tendency angle index, and the environment complexity index are normalized to map their value ranges to the interval [0, 1] as the input of the fuzzy system. For each input factor, a corresponding language variable is defined, and the actual value of each factor is fuzzified into three fuzzy sets, i.e., low, medium, and high, by using a triangular or trapezoidal membership function. Then, a multi-input and single-output fuzzy rule base is constructed, and the rule form is “if the proximity is high, the motion tendency angle is high, and the environment complexity is high, then the local threat degree is high”. According to statistical learning, a number of fuzzy inference rules covering all combination situations are set. In the inference process, the Mamdani-type fuzzy inference mechanism is used, the antecedent part of each rule is calculated by using the minimum operation to obtain the activation strength, and the consequent part corresponds to the output fuzzy set. The activated output fuzzy sets are combined by using the AND operation to form the fuzzy distribution of the comprehensive output. Then, the defuzzification of the combined fuzzy set is performed by using the centroid method to calculate its weighted average value in the domain, and thus the accurate local threat degree value is obtained, which is located in the interval [0, 1] and is used to quantify the threat degree of the current obstacle faced by the aircraft.

[0051] S20, the neighbor influence coefficient is calculated according to the neighbor position and speed data obtained from the communication topology in the cluster, and a dynamically adjusted cooperative threat diffusion value is output according to the local threat degree value of the aircraft itself and the neighbor influence coefficient.

[0052] S21, the neighbor position and speed information is obtained based on the communication topology, the relative distance attenuation effect and the motion direction consistency are fused, and the neighbor influence coefficient is generated.

[0053] Based on the distributed communication topology structure of the cluster, the aircraft judges the neighbors within the effective communication radius according to the received communication signal strength, identifies the reachable neighbor node set of the current aircraft, and obtains the real-time position and speed vector of each neighbor individual through periodic state broadcasting. The relative distance between the aircraft i and the neighbor aircraft j is calculated using the Euclidean distance formula according to the positions of the aircraft i and the neighbor aircraft j, and the speed vector angle is calculated according to the speed vectors of the aircraft i and the neighbor aircraft j. The distance attenuation effect and the consistency degree of the speed direction are fused, and a nonlinear weighting mechanism is used to generate the neighbor influence coefficient, which is used to measure the influence strength of the neighbor j on the behavior of the aircraft i.

[0054] Specifically, the neighbor influence coefficient calculation formula is where d ij is the relative distance between the aircraft i and the neighbor aircraft j, R c is the effective communication radius, θ ij is the speed vector angle of the aircraft i and the neighbor aircraft j, and cosθ ijis the degree of alignment of the velocity direction of aircraft i and neighbor aircraft j, γ is a tunable parameter used to enhance the influence of the alignment direction.

[0055] S22, the local threat degree of the neighbor is weighted and aggregated, and the stability index of the motion state of the aircraft itself is combined to calculate the neighborhood influence adjustment amount.

[0056] According to the influence coefficient of each neighbor, the local threat degree value reported by the neighbor is weighted and fused, and the state response sensitivity factor is calculated by combining the current aircraft's own motion acceleration change rate and attitude jitter energy. Then, the external threat input and internal state stability are comprehensively considered to obtain the neighborhood influence adjustment amount, and the specific calculation formula is

[0057] Where N i is the neighbor set of aircraft i, w ij is the neighbor influence coefficient of aircraft j on i, T j is the local threat degree value of neighbor aircraft j, is the state response sensitivity factor of aircraft i, is the linear acceleration vector of aircraft i, is the module length of linear acceleration, is the angular velocity vector of aircraft i, is the square module length of angular velocity, reflecting the attitude jitter energy. λ1 is the adjustment coefficient of linear acceleration influence, and λ2 is the adjustment coefficient of angular velocity influence.

[0058] S23, the self-threat degree and the neighborhood influence adjustment amount are enhanced by a nonlinear coupling function, a cooperative threat value with diffusion characteristics is constructed, and the dynamic guidance of high-threat individuals to the cluster obstacle avoidance behavior is realized.

[0059] The local threat degree value of the current aircraft itself is taken as the core driving source, and the neighborhood influence adjustment amount is nonlinearly coupled and operated. A cooperative threat enhancement function with positive feedback characteristics is designed. The exponential gain term and the saturation constraint mechanism are introduced, so that the high-threat individual can generate stronger obstacle avoidance signal radiation to its surrounding, and the cooperative threat diffusion value is obtained.

[0060] The specific formula is Where T i is the local threat degree value of aircraft i, α is the gain coefficient, which controls the amplification multiple of the neighbor influence on the self-threat, C i is the neighborhood influence adjustment amount, is the S-shaped activation function, β is the amplitude adjustment coefficient, k is the slope parameter, τ is the trigger threshold, if T i < τ, this term tends to 0 and is almost not activated, and if T i > τ, this term rapidly rises to close to β·Ci The strong diffusion mode is started when the threat of the aircraft i is high enough.

[0061] S30, the threat diffusion value is combined with the obstacle attraction and the cluster cohesion parameters, and is input into a fuzzy reasoning system to generate a three-axis behavior decision vector.

[0062] The aircraft monitors the remaining energy state of its battery, and when the energy is sufficient, the standard parameters are used to calculate the energy efficiency compensation; when the energy is lower than the threshold, the compensation mechanism is introduced to moderately reduce the attraction weight, so as to balance the flight efficiency and the target approaching demand. The distance and the direction consistency between the aircraft and the obstacle are weighted and combined, and then multiplied by the energy efficiency compensation. The result is processed by upper limit truncation to ensure that the output value is within the range of 0 to 1, and the obstacle attraction value is obtained.

[0063] The three-dimensional distance from the aircraft to each neighbor is calculated, and then the average of these distances is calculated. The average distance is processed by a hyperbolic tangent function to obtain the cluster spatial tightness. The aircraft collects the speed vector data of the neighbors, calculates the average speed vector of the neighbor group, and then compares the consistency degree of the direction of the own speed with the average speed direction. The spatial tightness and the speed consistency are weighted and combined, multiplied by the neighbor density compensation factor, and the result is processed by upper limit truncation to ensure that the output value is within the range of 0 to 1, and the cluster cohesion value is obtained.

[0064] Generating a three-axis behavior decision vector specifically includes the following steps:

[0065] S31, the threat diffusion value is compared with the obstacle attraction value to construct a dynamic obstacle avoidance trigger flag to determine whether the current aircraft is in an emergency obstacle avoidance demand state.

[0066] The threat diffusion value is compared with the obstacle attraction value to construct a dynamic obstacle avoidance trigger flag to determine whether the current aircraft is in an emergency obstacle avoidance demand state.

[0067] S32, a multi-behavior candidate set is generated by a direction modeling and weight evaluation algorithm.

[0068] After confirming the need to perform obstacle avoidance behavior, enter the multi-target coordination phase. First, analyze the spatial distribution of surrounding obstacles and other aircraft, and use the negative gradient synthesis method to calculate an exclusion direction that should be avoided first, as the dominant direction of obstacle avoidance behavior; second, according to the geometric relationship between the current pose of the aircraft and the task target point, determine the guide direction pointing to the target as the direction suggestion of the target-seeking behavior; based on the preset formation configuration and the relative position of the adjacent aircraft, calculate the deviation degree of the current formation structure, and generate the cohesion direction for restoring the ideal configuration as the behavior suggestion of formation keeping.

[0069] Quantitative evaluation of the rationality of each behavior in the current situation: the weight of obstacle avoidance behavior is obtained by nonlinear enhancement of the cooperative threat diffusion value, ensuring that high-risk situations are responded to first; the weight of target-seeking behavior is adjusted in combination with the target attraction and the current motion stability of the aircraft to prevent blind progress in intense maneuvering; the weight of formation keeping is determined according to the decay characteristics of the formation structure deviation energy, and the greater the deviation, the stronger the restoring force. Finally, a multi-behavior candidate set containing three unit direction vectors and their corresponding rationality weights is formed as input data for the next stage of fusion decision.

[0070] S33, generate a three-axis behavior decision vector through fuzzy reasoning and vector fusion algorithm.

[0071] The rationality weights of obstacle avoidance, target-seeking and formation keeping are used as fuzzy reasoning inputs, and membership functions are used to convert them into low, medium and high fuzzy language variables to complete the fuzzification process; according to the pre-set fuzzy rule base, perform reasoning operations, and the rule content reflects the behavior priority logic in different situations. Each activated rule outputs a suggested motion direction and its influence strength. Using the weighted synthesis method of vector space, all rule output direction suggestions are vector superimposed according to their activation strength to generate a comprehensive motion direction. The direction vector is normalized to form a three-axis behavior decision vector, whose three components represent the motion direction suggestions of the aircraft in the front-back, left-right and up-down three spatial axes, respectively, for generating speed commands to achieve the optimal balance between obstacle avoidance, task advancement and group coordination.

[0072] S40, if the decision is obstacle avoidance, use clustering algorithm to divide the areas with different threat levels, and according to the area where the aircraft is located and the dynamic constraints of the aircraft, generate a local obstacle avoidance strategy to determine the fault tolerance set.

[0073] S41, use clustering algorithm to divide the space where the obstacles are located into areas with different threat levels.

[0074] To facilitate clustering, the entire flight area is divided into three-dimensional grid cells. Each grid cell records whether it contains an obstacle center point, the number of obstacles within the cell, the average threat level, and the spatial connectivity status. This grid map serves as the input data field for the clustering algorithm, which uses a density clustering algorithm to regionalize obstacle locations. The algorithm automatically identifies connected areas with dense obstacle distribution and groups these areas into a single category, labeling them as high-risk areas. Areas with isolated or sparsely distributed obstacles are labeled as medium-risk areas, and areas with small or no obstacles are labeled as low-risk areas. The clustering process considers spatial proximity and continuity, ensuring that each divided area is physically coherent.

[0075] S42. Generate a local obstacle avoidance strategy based on the area where the aircraft is located and the dynamic constraints of the aircraft.

[0076] S421. Match obstacle avoidance behaviors in the multirotor aircraft obstacle avoidance behavior library according to the danger level of the area where the aircraft is located, the three-axis behavior decision vector, and the obstacle size.

[0077] The specific matching formula is: Among them, B match is the obstacle avoidance behavior in the matched aircraft obstacle avoidance behavior library, L is the aircraft obstacle avoidance behavior library, η d is the weight of direction matching, is the three-axis decision vector, is the direction vector of the i-th obstacle avoidance behavior in the obstacle avoidance behavior library, is a vector The model, is a vector The modulus of η t is the weight of threat level matching, e is a natural constant, ΔA is the difference between the current threat level and the baseline threat level of the i-th behavior in the behavior library, σ t is the adjustment parameter of threat fitness, which controls the influence range of threat level difference, η s is the weight of obstacle matching, s obs is the equivalent size of the current obstacle, is the obstacle size referenced by the i-th obstacle avoidance behavior in the aircraft behavior library, and δ is a smoothing constant.

[0078] S422: Determine the specific execution of the matched obstacle avoidance behavior based on the actual situation of the obstacle and the dynamic constraints of the aircraft.

[0079] According to the matched obstacle avoidance behavior type, combined with the real-time perceived obstacle size, position distribution and threat urgency, the preset behavior parameters are dynamically adjusted in three stages: the motion amplitude is scaled based on the actual size of the obstacle, the large obstacle automatically increases the 20% fly radius, and the behavior direction vector is fine-tuned according to the relative position of the obstacle during the flight of the aircraft, and finally the maneuvering strength is strengthened according to the emergency degree of the obstacle threat. At the same time, the real-time residual power state of the aircraft and the dynamic limit are fused to generate refined obstacle avoidance instructions that meet the current environmental constraints, and the obstacle motion trend is continuously monitored during execution for rolling optimization.

[0080] S43, a three-dimensional ellipsoid basic safety envelope is constructed, the dynamic shrinking boundary is compensated by threat erosion, and the infeasible region is cut by combining the kinetic accessibility domain to generate the aircraft behavior fault tolerance set.

[0081] Based on the adjusted obstacle avoidance strategy and flight state, a three-dimensional ellipsoid basic safety envelope is constructed with the aircraft as the center and the strategy vector direction as the long axis. The longitudinal size is determined by the aircraft speed and the prediction time window, and the lateral size is determined by the physical size of the aircraft body and the set safety margin. Then three dynamic adjustments are made: threat erosion compensation, envelope boundary compression according to the depth of obstacle intrusion, and long axis reduction according to the depth of intrusion and the emergency degree; dynamic cutting, removing regions that intersect with obstacle trajectories and exceed the aircraft maneuvering ability; environmental adaptive optimization, expanding or shrinking the safety space boundary by integrating positioning accuracy, energy state and weather disturbance, and finally outputting the fault tolerance set containing attitude adjustment range and variable time window.

[0082] S50, based on the global environment map, the potential conflict probability between the obstacle avoidance paths of each aircraft is calculated by a collision detection algorithm. If there is a cooperative conflict, a game theory algorithm is used for decision compensation to solve the conflict and maintain the cooperative flight of the cluster.

[0083] S51, by discretizing the aircraft obstacle avoidance path and detecting close-range conflict points at the same time point, a conflict probability matrix between aircraft pairs is generated by combining weighted calculation of regional risk coefficients.

[0084] The global environment map generated by the cluster distributed communication topology structure is obtained, the obstacle avoidance prediction path with timestamp of each aircraft is discretized into 0.1 second interval space-time points, then the spatial distance of the aircraft at the same time is detected pair by pair, when the distance is less than the dynamic safety threshold, it is marked as a conflict point, and the conflict probability is calculated by combining the conflict point distribution: the proportion of conflict points in the total number of points is calculated, multiplied by the set regional risk coefficient, to generate an N×N-dimensional conflict probability matrix, where each element represents the collision risk value of a specific aircraft pair.

[0085] S52, construct a potential game model of the differentiated payoff function, solve the Nash equilibrium by iterative optimization, and output the heading correction and speed compensation instructions of the aircraft.

[0086] A game model is constructed for the aircrafts in conflict: define the strategy set path adjustment options, including adjusting speed, adjusting yaw angle, adjusting height; design a differentiated payoff function, which consists of path efficiency and conflict cost. The path efficiency is measured by the degree to which the aircraft approaches the target point along the adjusted path, and the conflict cost is represented by the conflict probability of the adjusted path multiplied by the regional danger coefficient. The weights of path efficiency and conflict cost are set according to the danger level of the region where the aircraft is located. For high-risk aircraft, more attention is paid to safety, and the weight of conflict cost is higher. For low-risk aircraft, more attention is paid to efficiency, and the weight of path efficiency is higher. The virtual optimal response algorithm is used to solve the Nash equilibrium: each aircraft adjusts its own strategy in turn, while the strategies of other aircrafts remain unchanged. Each time the aircraft selects the strategy that maximizes its payoff. This process is iterated until the strategy of all aircrafts changes less than a certain threshold or reaches the maximum number of iterations. According to the Nash equilibrium strategy, compensation instructions are generated for each aircraft, including speed adjustment, heading correction, height adjustment, etc.

[0087] S53, monitor the conflict probability at a fixed period after executing the compensation instructions, and dynamically trigger local re-planning until the global conflict probability is continuously below the set threshold, ensuring that the cluster reaches the target without conflict.

[0088] After the aircraft executes the compensation instructions, the path conflict probability is recalculated at a period of 0.5 seconds. When a conflict probability exceeding the set threshold is detected, local re-planning of the affected aircraft is triggered immediately, retaining 80% of the original path. After confirming that the global conflict probability is below the set threshold through continuous detection for three times, the energy-saving cruise mode is switched to. The whole process uses a rolling optimization mechanism until all aircrafts arrive at the target area without conflict.

[0089] Embodiment Two

[0090] As shown in Figure 2 Embodiment Two of the present application provides a multi-rotor aircraft cluster cooperative obstacle avoidance control system, which comprises:

[0091] Obtain data and calculate threat module: obtain real-time data of each multi-rotor aircraft in the cluster and sensing distance data of surrounding obstacles, and determine the local threat degree value of each aircraft and obstacle by using fuzzy comprehensive evaluation algorithm. Specifically, it is divided into the following sub-modules:

[0092] Calculation submodule: calculate the proximity index and motion trend angle index according to the positions of the aircraft and the obstacle, and calculate the environmental complexity index according to the environmental information.

[0093] Fuzzy inference submodule: The proximity index, the motion trend angle index, and the environment complexity index are taken as input factors of fuzzy comprehensive evaluation, and the local threat degree value is determined by using the Mamdani type fuzzy inference mechanism.

[0094] Neighbor influence and cooperative threat submodule: The neighbor influence coefficient is calculated according to the neighbor position and speed data obtained from the communication topology in the cluster, and the dynamically adjusted cooperative threat diffusion value is output according to the local threat degree value of the aircraft itself and the neighbor influence coefficient. It is specifically divided into the following submodules:

[0095] Neighbor influence submodule: The neighbor position and speed information is obtained based on the communication topology, and the neighbor influence coefficient is generated by fusing the relative distance attenuation effect and the motion direction consistency.

[0096] Neighbor influence submodule: The local threat degree of the neighbor and its influence coefficient are weighted and aggregated, and the neighbor influence adjustment amount is calculated in combination with the stability index of the motion state of the aircraft itself.

[0097] Cooperative threat submodule: The local threat degree of the aircraft itself and the neighbor influence adjustment amount are enhanced by a nonlinear coupling function, a cooperative threat value with diffusion characteristics is constructed, and the dynamic guidance of the high-threat individual to the cluster obstacle avoidance behavior is realized.

[0098] Decision generation module: The cooperative threat diffusion value is combined with the obstacle point attraction force and the cluster cohesion force parameters, and is input into the fuzzy inference system to generate a three-axis behavior decision vector for multi-objective decision making. It is specifically divided into the following submodules:

[0099] Judgment submodule: The cooperative threat diffusion value is compared with the obstacle point attraction force value, a dynamic obstacle avoidance trigger flag is constructed, and it is determined whether the current aircraft is in an emergency obstacle avoidance demand state.

[0100] Candidate submodule: A multi-behavior candidate set is generated by a direction modeling and weight evaluation algorithm.

[0101] Decision submodule: A three-axis behavior decision vector is generated by a fuzzy inference and vector fusion algorithm.

[0102] Division and generation strategy module: If the decision is obstacle avoidance, a clustering algorithm is used to divide the regions with different threat levels, a local obstacle avoidance strategy is generated according to the region where the aircraft is located and the kinematic constraints of the aircraft, and a fault-tolerant set is determined. It is specifically divided into the following submodules:

[0103] Division submodule: The space where the obstacles are located is divided into regions with different threat levels by using a clustering algorithm.

[0104] Obstacle avoidance submodule: A local obstacle avoidance strategy is generated according to the region where the aircraft is located and the kinematic constraints of the aircraft. It is specifically divided into the following submodules:

[0105] Matching behavior library submodule: according to the danger level of the area where the aircraft is located, the three-axis behavior decision vector, and the obstacle size, the obstacle avoidance behavior of the multi-rotor aircraft is matched with the obstacle avoidance behavior library.

[0106] Adjustment submodule: according to the actual situation of the obstacle and the dynamic constraints of the aircraft, the specific execution of the matched obstacle avoidance behavior is determined.

[0107] Fault-tolerant submodule: a three-dimensional ellipsoid basic safety envelope is constructed, the dynamic shrinking boundary is compensated by threat erosion, and the infeasible region is cut by combining the kinetic reachable set, to generate the fault-tolerant set of the aircraft behavior.

[0108] Conflict detection and decision compensation module: based on the global environment map, the potential conflict probability between the obstacle avoidance paths of each aircraft is calculated by the collision detection algorithm, and the game theory algorithm is used for decision compensation when there is a cooperative conflict, to solve the conflict and maintain the cooperative flight of the cluster. Specifically, the following submodules are included:

[0109] Conflict detection submodule: by discretizing the obstacle avoidance path of the aircraft and detecting the close-range conflict points at the same time point, the conflict probability matrix between the pairs of aircraft is generated by combining the weighted calculation of the regional danger coefficient.

[0110] Correction and compensation submodule: a potential game model of differentiated payoff function is constructed, and the Nash equilibrium is solved by iterative optimization to output the heading correction and speed compensation instructions of the aircraft.

[0111] Instruction execution and monitoring submodule: after executing the compensation instructions, the conflict probability is monitored at a fixed period, and the local re-planning is dynamically triggered until the global conflict probability is continuously below the set threshold, to ensure that the cluster reaches the target without conflict.

[0112] The above specific embodiments further illustrate the purpose, technical solutions and beneficial effects of the present application. It should be understood that the above description is only a specific embodiment of the present application and does not limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made on the basis of the technical solutions of the present application shall be included in the protection scope of the present application.

Claims

1. A multi-rotor aircraft cluster coordinated obstacle avoidance control method and system, characterized in that: include: S10, obtaining real-time data of each multi-rotor aircraft in the cluster and the perceived distance data of surrounding obstacles, and using a fuzzy comprehensive evaluation algorithm to determine the local threat level value between each aircraft and the obstacle; S20, obtaining neighbor position and velocity data based on the communication topology in the cluster to calculate the neighbor influence coefficient, and outputting a dynamically adjusted collaborative threat diffusion value based on the aircraft's own local threat value and the neighbor influence coefficient; S30, the collaborative threat diffusion value is combined with the obstacle point attraction and cluster cohesion parameters, and input into the fuzzy inference system to make multi-objective decisions to generate a three-axis behavior decision vector; S40, if the decision is to avoid obstacles, a clustering algorithm is used to divide the areas with different threat levels, and a local obstacle avoidance strategy is generated based on the area where the aircraft is located and the dynamic constraints of the aircraft, and a fault tolerance set is determined; S50. Based on the global environment map, the potential conflict probability between the obstacle avoidance paths of each aircraft is calculated through a collision detection algorithm. If there is a collaborative conflict, a game theory algorithm is used to make decision compensation to resolve the conflict and maintain the collaborative flight of the cluster.

2. The multi-rotor aircraft cluster coordinated obstacle avoidance control method according to claim 1, characterized in that: The fuzzy comprehensive evaluation algorithm is used to determine the local threat value of each aircraft and obstacle. It is divided into the following sub-steps: Calculate the proximity index and motion direction angle index based on the positions of the aircraft and obstacles, and calculate the environment complexity index based on the environmental information; The proximity index, motion tendency angle index and environment complexity index are used as input factors of fuzzy comprehensive evaluation, and the local threat value is determined using Mamdani type fuzzy inference mechanism.

3. The multi-rotor aircraft cluster coordinated obstacle avoidance control method according to claim 1, characterized in that: The neighbor influence coefficient is calculated based on the communication topology of the cluster, and the neighbor position and speed data are obtained. The dynamically adjusted collaborative threat diffusion value is output based on the aircraft's own local threat value and the neighbor influence coefficient. The specific steps are as follows: Obtain neighbor position and speed information based on communication topology, integrate relative distance attenuation effect and motion direction consistency to generate neighbor influence coefficient; The neighborhood influence adjustment amount is calculated by weighted aggregation of the neighbors' local threat degree and their influence coefficient, combined with the stability index of its own motion state; The self-threat degree and the neighborhood influence adjustment amount are positively feedback enhanced through a nonlinear coupling function to construct a collaborative threat value with diffusion characteristics, so as to achieve dynamic guidance of the cluster's obstacle avoidance behavior by high-threat individuals.

4. The multi-rotor aircraft cluster coordinated obstacle avoidance control method according to claim 1, characterized in that: The collaborative threat diffusion value is combined with the obstacle attraction and cluster cohesion parameters and input into the fuzzy inference system for multi-objective decision-making to generate a three-axis behavior decision vector. The specific steps are as follows: Compare the collaborative threat diffusion value with the obstacle attraction value to construct a dynamic obstacle avoidance trigger flag to determine whether the current aircraft is in an emergency obstacle avoidance requirement state; Generate multiple behavior candidate sets through direction modeling and weight evaluation algorithm; The three-axis behavior decision vector is generated through fuzzy reasoning and vector fusion algorithm.

5. The multi-rotor aircraft cluster coordinated obstacle avoidance control method according to claim 1, characterized in that: Generate a local obstacle avoidance strategy based on the area where the aircraft is located and the aircraft's dynamic constraints. This strategy is divided into the following sub-steps: Match the obstacle avoidance behaviors in the multi-rotor aircraft obstacle avoidance behavior library based on the danger level of the area where the aircraft is located, the three-axis behavior decision vector, and the obstacle size; The specific execution of the matched obstacle avoidance behavior is determined based on the actual situation of the obstacle and the dynamic constraints of the aircraft.

6. The multi-rotor aircraft cluster coordinated obstacle avoidance control method according to claim 5, characterized in that: The matching obstacle avoidance behavior is specifically executed based on the actual obstacle situation and the dynamic constraints of the aircraft. It is divided into the following sub-steps: The movement amplitude is scaled based on the actual size of the obstacle, the behavior direction vector is fine-tuned according to the relative position of the obstacle during the aircraft's movement, and the maneuver intensity is enhanced according to the urgency of the obstacle threat; At the same time, the real-time remaining power status of the aircraft and its dynamic limits are integrated to generate refined obstacle avoidance instructions that meet the current environmental constraints.

7. The multi-rotor aircraft cluster coordinated obstacle avoidance control method according to claim 1, characterized in that: Based on the global environment map, the collision detection algorithm calculates the potential conflict probability between the obstacle avoidance paths of each aircraft. If there is a collaborative conflict, the game theory algorithm is used to make decision compensation to resolve the conflict and maintain the coordinated flight of the cluster. The specific steps are as follows: By discretizing the aircraft's obstacle avoidance path and detecting close-range conflict points at the same time, combined with weighted calculation of regional risk coefficients, a conflict probability matrix between aircraft pairs is generated. Construct a potential game model with differentiated payoff functions, use iterative optimization to solve the Nash equilibrium, and output the aircraft's heading correction and speed compensation instructions; After executing the compensation instruction, the conflict probability is monitored at a fixed period, and local replanning is dynamically triggered until the global conflict probability remains below the set threshold, ensuring that the cluster reaches the target in a conflict-free and coordinated manner.

8. A multi-rotor aircraft cluster collaborative obstacle avoidance control system, characterized in that: include: Data acquisition and threat calculation module: obtains real-time data of each multi-rotor aircraft in the cluster and the perceived distance data of surrounding obstacles, and uses fuzzy comprehensive evaluation algorithm to determine the local threat value of each aircraft and obstacle; Neighbor Influence and Cooperative Threat Module: This module calculates the neighbor influence coefficient based on the communication topology of the cluster and obtains neighbor position and velocity data. It then outputs a dynamically adjusted cooperative threat diffusion value based on the aircraft's own local threat value and the neighbor influence coefficient. Generate decision module: The coordinated threat diffusion value is combined with the obstacle attraction and cluster cohesion parameters, and input into the fuzzy inference system to perform multi-objective decision-making to generate a three-axis behavior decision vector; Division and strategy generation module: If the decision is to avoid obstacles, a clustering algorithm is used to divide the areas with different threat levels. Based on the area where the aircraft is located and the aircraft's dynamic constraints, a local obstacle avoidance strategy is generated to determine the fault tolerance set; Conflict detection and decision compensation module: Based on the global environment map, the collision detection algorithm is used to calculate the potential conflict probability between the obstacle avoidance paths of each aircraft. If there is a collaborative conflict, the game theory algorithm is used for decision compensation to resolve the conflict and maintain the collaborative flight of the cluster.

Images