Multi-uav logistics transportation collision prediction method

Abstract

The present application relates to a kind of multi-unmanned aerial vehicle logistics transportation collision prediction methods, method is generated by the dynamic coordination flight path of logistics transportation point space distribution and unmanned aerial vehicle real-time flight data, solve the trajectory conflict problem caused by the mismatch of traditional path planning and logistics node layout;Based on trajectory intersection area detection and minimum relative distance calculation accurate identification conflict space-time range;Through the uniform mapping of heterogeneous unmanned aerial vehicle obstacle avoidance action is realized in protocol obstacle avoidance rule base, eliminate the instruction conflict risk in multi-protocol environment;Combining the difference of obstacle avoidance action direction, distance attenuation factor and relative speed multi-dimensional parameter fusion calculation collision probability, significantly improve risk prediction accuracy;The method effectively solves the collision hazard caused by path planning mismatch, protocol heterogeneous conflict and risk model deviation in logistics scenario, realizes the safety protection of multi-unmanned aerial vehicle collaborative transportation.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) transportation management technology, and specifically to a method for predicting collisions in multi-UAV logistics transportation. Background Technology

[0002] With the rapid development of urban air mobility and smart logistics, multi-drone collaborative transportation systems are becoming a key technology for solving the "last mile" delivery problem. In logistics transportation scenarios, multiple drones need to simultaneously perform cargo transportation tasks, building a dense flight network between logistics hubs, warehouses, and delivery points. This high-density operation mode exposes drones to significant collision risks in areas where transportation paths intersect. Especially when multiple drones approach logistics transfer stations simultaneously, the spatial overlap and temporal staggering of their flight trajectories can easily lead to air traffic conflicts.

[0003] In existing technologies, mainstream collision prediction methods have three significant drawbacks: First, traditional path planning schemes fail to fully consider the spatial distribution characteristics of logistics transportation points, resulting in inconsistencies between flight paths and logistics node layouts, increasing the probability of trajectory intersections; second, in logistics scenarios using drones from different manufacturers, heterogeneous communication protocols make it difficult to unify and coordinate obstacle avoidance commands, often leading to command conflicts that exacerbate collision risks; and finally, most risk prediction models lack fusion analysis of multi-dimensional parameters such as differences in obstacle avoidance command directions and relative speeds, resulting in significant deviations between prediction results and actual risks. Summary of the Invention

[0004] Based on this, the purpose of this invention is to provide a multi-UAV logistics transportation collision prediction method that can collaboratively optimize logistics route planning, adapt to heterogeneous protocol environments, and accurately predict collision risks.

[0005] The objective of this invention is achieved through the following solution:

[0006] In a first aspect, the present invention provides a method for predicting collisions in multi-UAV logistics transportation, comprising the following steps:

[0007] S1: Obtain the coordinate data of the logistics transportation point and the real-time flight data of each drone, calculate the spatial relationship between the logistics transportation point and the flight distance of the drone, and generate the flight path data of each drone.

[0008] S2: Detect conflict areas in flight path data, calculate the minimum relative distance between two adjacent UAV trajectories, filter trajectories intersecting, and generate conflict area information and minimum distance data;

[0009] S3: Query obstacle avoidance actions for conflict area information, call the protocol obstacle avoidance rules to query the protocol action mapping relationship according to the relative azimuth angle, and generate the original obstacle avoidance action data;

[0010] S4: Perform collision risk prediction on the original obstacle avoidance action data and minimum distance data, calculate the conflict probability by combining the difference in obstacle avoidance action direction and relative speed, and generate collision risk probability data.

[0011] In one embodiment, S1 of the multi-UAV logistics transportation collision prediction method provided by the present invention specifically includes the following steps:

[0012] S11: Obtain the coordinate data of logistics transportation points and the real-time flight data of each drone, perform spatial distribution analysis on the coordinate data, calculate the distance relationship between the drone and each logistics transportation point based on the geographical location of the logistics transportation points, and generate drone delivery range data.

[0013] S12: Perform path optimization processing on the drone delivery range data, call the dynamic programming algorithm to generate multiple alternative transportation routes for each drone, and generate a set of alternative paths for each drone.

[0014] S13: Perform shortest path filtering on the set of candidate paths for each UAV, select the path with the shortest transport distance as the final flight path, and generate flight path data for each UAV. The flight path data is used to indicate the transport trajectory of the UAV.

[0015] In one embodiment, S2 of the multi-UAV logistics transportation collision prediction method provided by the present invention specifically includes the following steps:

[0016] S21: Perform trajectory function modeling on the flight path data of each UAV, construct a quadratic polynomial trajectory prediction function based on the current position, velocity vector and acceleration vector of each UAV, and generate trajectory function data for each UAV.

[0017] S22: Perform relative distance calculation on the trajectory function data of two adjacent UAVs, solve for the minimum Euclidean distance between the trajectories of the two UAVs within a predetermined time window, and generate minimum distance data. The minimum distance data is used to indicate the closest approach distance between the two UAVs during flight.

[0018] S23: Perform conflict area identification processing on the minimum distance data. If the minimum distance is less than the safe distance threshold, mark the corresponding time and spatial location as a conflict area and generate conflict area information. The conflict area information is used to indicate the spatiotemporal range where obstacle avoidance decisions are needed.

[0019] In one embodiment, S3 of the multi-UAV logistics transportation collision prediction method provided by the present invention specifically includes the following steps:

[0020] S31: Calculate the relative azimuth of the conflict area information, extract the real-time position coordinate data of the two UAVs in the conflict area, and calculate the average relative azimuth angle of the continuous time series.

[0021] S32: Perform discrete partitioning mapping on the average relative azimuth angle, mapping continuous angle values ​​to a preset set of discrete azimuth angle partitions to generate target azimuth angle partitions;

[0022] S33: Based on the protocol obstacle avoidance rule data, perform action query on the target azimuth partition, retrieve predefined obstacle avoidance action vectors according to the UAV's protocol type and azimuth partition, and generate raw obstacle avoidance action data.

[0023] In one embodiment, S4 of the multi-UAV logistics transportation collision prediction method provided by the present invention specifically includes the following steps:

[0024] S41: Perform motion direction analysis on the original obstacle avoidance motion data, calculate the cosine of the angle between the obstacle avoidance motion vectors of the two UAVs, and generate the motion direction difference value.

[0025] S42: Perform distance impact assessment processing on the minimum distance data, call the exponential decay function to calculate the impact factor of the minimum distance on the collision risk, and generate the distance impact factor;

[0026] S43: Perform collision probability synthesis processing on the difference value of the action direction and the distance influence factor, combine it with the relative velocity modulus to calculate the collision risk probability, and generate collision risk probability data. The collision risk probability data is used to indicate the collision risk level of the UAV.

[0027] In one embodiment, the calculation formula for the collision risk probability data of the multi-UAV logistics transportation collision prediction method provided by the present invention is as follows:

[0028] P coll =σ(α*D*F d +β*v rel )

[0029] Among them, P coll For collision risk probability data, σ is the Sigmoid function, α and β are weighting coefficients, D is the difference in action direction, and F... d As the distance influence factor, v rel This represents the relative velocity magnitude.

[0030] In one embodiment, after step S4 of the multi-UAV logistics transportation collision prediction method provided by the present invention, it further includes:

[0031] S51: Perform risk level judgment processing on collision risk probability data. When the risk probability exceeds the preset risk threshold, trigger the obstacle avoidance command compilation process to generate a high-risk warning signal.

[0032] S52: Perform obstacle avoidance action optimization processing on collision risk probability data and original obstacle avoidance action data, calculate the optimal obstacle avoidance action by minimizing collision probability and action offset, and generate optimal obstacle avoidance action data;

[0033] S53: Performs protocol instruction conversion processing on high-risk warning signals and optimal obstacle avoidance action data. Based on the target UAV's protocol type, it compiles the action vector into binary instruction code to generate UAV obstacle avoidance instructions. These instructions are used to instruct the UAV to perform three-dimensional obstacle avoidance actions.

[0034] Secondly, this invention provides a multi-UAV logistics transportation collision prediction system, which is configured with the following modules:

[0035] The flight path generation module is used to acquire the coordinate data of logistics transportation points and the real-time flight data of each UAV, calculate the spatial relationship between the logistics transportation points and the flight distance of the UAVs, and generate the flight path data of each UAV.

[0036] The conflict zone detection module is used to detect conflict zones in flight path data, calculate the minimum relative distance between two adjacent UAV trajectories, filter trajectories intersecting, and generate conflict zone information and minimum distance data.

[0037] The obstacle avoidance action mapping module is used to query obstacle avoidance actions based on conflict area information, call the protocol obstacle avoidance rules to query the protocol action mapping relationship according to the relative azimuth angle, and generate the original obstacle avoidance action data.

[0038] The collision risk prediction module is used to predict collision risks from raw obstacle avoidance action data and minimum distance data. It combines the differences in obstacle avoidance action direction and relative speed to calculate the probability of conflict and generate collision risk probability data.

[0039] Thirdly, this application provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement any of the above-mentioned multi-UAV logistics transportation collision prediction methods.

[0040] Fourthly, this application provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements any of the above-mentioned multi-UAV logistics transportation collision prediction methods.

[0041] In summary, the multi-UAV logistics transportation collision prediction method provided in this application achieves deep adaptation between flight paths and logistics node layout through dynamic collaborative processing of the spatial distribution of logistics transportation points and real-time UAV flight data, effectively solving the trajectory conflict problem caused by the disconnect between traditional path planning and delivery networks. Based on trajectory intersection area detection and minimum relative distance calculation mechanism, it can accurately identify the spatiotemporal range of conflicts, thereby achieving a dual improvement in the advance warning and coverage of conflict warnings. By using a protocol obstacle avoidance rule base to uniformly map and transform the obstacle avoidance actions of heterogeneous UAVs, it can eliminate the risk of command conflicts in multi-protocol environments and ensure the reliability of collaborative obstacle avoidance of UAVs from different manufacturers in mixed deployment scenarios. By integrating a multi-dimensional parameter analysis model of obstacle avoidance action direction differences, distance attenuation factors, and relative speeds, it can achieve dynamic and accurate prediction of collision risk probability, significantly improving the safety protection capability in complex logistics scenarios.

[0042] To better understand and implement this invention, the following detailed description is provided in conjunction with the accompanying drawings. Attached Figure Description

[0043] Figure 1 A flowchart illustrating a multi-UAV logistics transportation collision prediction method provided in this application embodiment;

[0044] Figure 2 This is a schematic diagram of the process for generating collision risk probability data provided in an embodiment of this application;

[0045] Figure 3 This is a schematic diagram of the structure of a multi-UAV logistics transportation collision prediction system provided in another embodiment of this application. Detailed Implementation

[0046] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Preferred embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a thorough and complete understanding of the disclosure of the invention.

[0047] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0048] In one embodiment, such as Figure 1As shown, a method for predicting collisions in multi-UAV logistics transportation is provided. This embodiment illustrates the application of this method to a terminal. It is understood that this method can also be applied to a server, or to a system including both a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0049] S1: Obtain the coordinate data of the logistics transportation point and the real-time flight data of each drone, calculate the spatial relationship between the logistics transportation point and the flight distance of the drone, and generate the flight path data of each drone.

[0050] Specifically, the coordinate data of logistics transportation points includes the latitude and longitude coordinates or three-dimensional coordinates (x, y, z) of fixed nodes such as logistics distribution points, warehouses, and delivery target points in a relative coordinate system, obtained through a pre-set geographic information database or a manual input system. Real-time flight data of the UAV is collected through the GNSS module, IMU inertial measurement unit, and onboard sensors carried by the UAV, specifically including: UAV number, real-time position coordinates (x, y, z). t y t , z t The system can measure instantaneous flight speed, heading angle, flight altitude, and remaining battery power, etc.; the data is transmitted in real time to the processing terminal at the ground control center via a wireless communication module.

[0051] To accurately calculate the spatial relationship between the logistics transport point and the drone's flight distance, the system employs a spatial geometric algorithm. Using the logistics transport point as a reference, a spatial coordinate system is constructed centered on that point, and the drone's real-time position coordinates are mapped onto this coordinate system. Then, by calculating the Euclidean distance between the drone's position coordinates and the logistics transport point's coordinates, the straight-line spatial distance between them is determined. Simultaneously, by combining the drone's flight direction and velocity vector, the system analyzes the relative positional change trend between the drone and the logistics transport point, determining whether the drone is flying towards the logistics transport point, away from it, or circling around it in the surrounding area.

[0052] Preferably, based on the above spatial relationship analysis results and the preset flight mission requirements such as cargo delivery sequence and expected arrival time, a path planning algorithm can be used to generate flight path data for each UAV. The flight path data records the complete flight trajectory of the UAV from the starting point to the destination in detail, presented as a series of continuous coordinate point sequences. It also includes parameter suggestions such as flight speed and flight altitude for each segment of the trajectory, aiming to ensure that the UAV avoids known obstacles and high-risk areas as much as possible while meeting mission requirements, and efficiently completes the cargo transportation mission.

[0053] S2: Detect conflict areas in flight path data, calculate the minimum relative distance between two adjacent UAV trajectories, filter trajectories intersecting, and generate conflict area information and minimum distance data.

[0054] Specifically, the system can use a three-dimensional mesh partitioning method to divide the UAV's flight airspace into several equal mesh cells, with the size of the mesh cells set according to the UAV's minimum safe distance. The system maps the flight paths of all UAVs to the partitioned mesh cells, traverses each mesh cell, and checks whether there are two or more paths passing through the same mesh cell simultaneously.

[0055] For detected overlapping grid cells, the system further analyzes the time windows of the involved UAVs passing through those grid cells to determine if there is an intersection between the time windows. For path pairs with time intersection, the system extracts the parameterized equations of the two paths, where the parameterized equations represent position coordinates with time as the variable. Preferably, the system can solve for the shortest distance between the spatial curves corresponding to the two parameterized equations using a numerical iteration method to obtain the minimum relative distance between the trajectories of two adjacent UAVs. Based on a preset minimum safe distance threshold, conflict areas are identified. Specifically, when the minimum relative distance is less than this threshold, the corresponding grid cell area is determined to be a conflict area. Simultaneously, the system records the boundary coordinates of the conflict area, the UAV identifiers involved, and the time window to form conflict area information, and stores the minimum relative distance and the corresponding time point, generating minimum distance data.

[0056] S3: Query obstacle avoidance actions for conflict area information, call the protocol obstacle avoidance rules to query the protocol action mapping relationship according to the relative azimuth angle, and generate the original obstacle avoidance action data.

[0057] Specifically, when a conflict zone that meets preset conditions exists in the conflict zone information, the system retrieves the protocol obstacle avoidance rules from the storage module. These rules include communication protocol adaptation standards for drones from different manufacturers and corresponding obstacle avoidance action specifications. Based on this, the system calculates the relative azimuth angles between adjacent drones within the conflict zone. The calculation method involves establishing a local coordinate system with one drone as the origin, solving for the azimuth angle value based on the position coordinates of the other drone, and then partitioning the system according to the magnitude of the relative azimuth angle. The partitioning criteria are four intervals: 0°-90°, 90°-180°, 180°-270°, and 270°-360°, with each interval corresponding to a different azimuth region.

[0058] Specifically, the system queries the preset protocol action mapping relationship in the protocol obstacle avoidance rules based on the partition to which the relative azimuth angle belongs. The protocol action mapping relationship is the standardized obstacle avoidance action corresponding to drones from different manufacturers under different partitions. At the same time, the system converts the obstacle avoidance actions under heterogeneous communication protocols into action commands in a unified format, generating raw obstacle avoidance action data. The raw obstacle avoidance action data includes the drone identification involved, the type of obstacle avoidance action, the start time of action execution, and action parameters. The action parameters cover yaw angle, climb rate, and speed adjustment value.

[0059] S4: Perform collision risk prediction on the original obstacle avoidance action data and minimum distance data, calculate the conflict probability by combining the difference in obstacle avoidance action direction and relative speed, and generate collision risk probability data.

[0060] Specifically, the system extracts the obstacle avoidance action direction of each UAV from the raw obstacle avoidance action data, and calculates the angle between the obstacle avoidance action directions of adjacent UAVs through vector operations. The angle value is used to characterize the difference in obstacle avoidance action direction. Based on the real-time flight speed of the UAVs, the system can use the relative velocity calculation formula to solve for the relative velocity vector between adjacent UAVs. The magnitude of the relative velocity vector reflects the rate at which the two UAVs approach or move away from each other.

[0061] Preferably, the system uses the difference in obstacle avoidance action direction, relative speed, and minimum distance as input parameters, which are then substituted into a preset collision risk prediction model. The collision risk prediction model employs a weighted summation algorithm, where the collision probability is equal to the sum of the products of the weight coefficients corresponding to each parameter and the parameter values. The weight coefficients are obtained through training with historical collision data. The system calculates the collision probability between adjacent drones using this model, with the probability ranging from 0 to 1. Simultaneously, based on the magnitude of the collision probability, the system classifies the collision risk into three levels: high risk, medium risk, and low risk. The calculation results are then integrated to generate collision risk probability data, which includes the involved drone identification pairs, the collision probability values, and the corresponding risk levels.

[0062] In summary, the multi-UAV logistics transportation collision prediction method provided in this application achieves deep adaptation between flight paths and logistics node layout through dynamic collaborative processing of the spatial distribution of logistics transportation points and real-time UAV flight data, effectively solving the trajectory conflict problem caused by the disconnect between traditional path planning and delivery networks. Based on trajectory intersection area detection and minimum relative distance calculation mechanism, it can accurately identify the spatiotemporal range of conflicts, thereby achieving a dual improvement in the advance warning and coverage of conflict warnings. By using a protocol obstacle avoidance rule base to uniformly map and transform the obstacle avoidance actions of heterogeneous UAVs, it can eliminate the risk of command conflicts in multi-protocol environments and ensure the reliability of collaborative obstacle avoidance of UAVs from different manufacturers in mixed deployment scenarios. By integrating a multi-dimensional parameter analysis model of obstacle avoidance action direction differences, distance attenuation factors, and relative speeds, it can achieve dynamic and accurate prediction of collision risk probability, significantly improving the safety protection capability in complex logistics scenarios.

[0063] The method provided in this application can achieve global optimization of flight path planning and logistics delivery needs, thereby systematically reducing the probability of trajectory conflicts; it can complete the standardized conversion of heterogeneous protocol action commands, ensuring the consistency and executability of obstacle avoidance decisions; and it can construct a multi-dimensional dynamic risk assessment system, thereby significantly improving the accuracy of collision warnings. The entire process, through a three-pronged approach of deep integration with logistics scenarios, enhanced protocol compatibility, and optimized risk models, effectively addresses collision safety hazards in high-density drone logistics transportation, providing reliable safety assurance for intelligent logistics systems.

[0064] In one embodiment, S1 of the multi-UAV logistics transportation collision prediction method provided by the present invention specifically includes the following steps:

[0065] S11: Obtain coordinate data of logistics transportation points and real-time flight data of each drone, perform spatial distribution analysis on the coordinate data, calculate the distance relationship between the drone and each logistics transportation point based on the geographical location of the logistics transportation points, and generate drone delivery range data.

[0066] Specifically, the system receives coordinate data of logistics transportation points through a data interface. This coordinate data uses the WGS-84 geodetic coordinate system and includes three-dimensional coordinate information such as longitude, latitude, and altitude for various nodes, including logistics distribution centers, regional warehouses, and last-mile delivery points. The data format is a structured data table, and each record includes a unique node identifier, coordinate parameters, and a node type label. Simultaneously, the system collects real-time flight data from each UAV through its onboard GNSS module and airborne sensors. This flight data includes real-time position coordinates, instantaneous flight speed, current heading angle, and remaining range. The data sampling interval is 1 second, and the data is transmitted via a wireless communication link to a ground-based multi-UAV logistics transportation collision prediction system.

[0067] Preferably, the system performs spatial distribution analysis on the received coordinate data. It can employ the K-means clustering algorithm to divide the logistics transportation points into regions. Using node coordinates as feature values, the transportation points are divided into multiple spatial clusters according to a preset number of clusters. The centroid coordinates of each cluster are calculated as the regional center reference point, generating a spatial distribution heatmap and cluster boundary polygon data. Based on this analysis, the system calculates the distance relationship between the UAV and each logistics transportation point. The straight-line distance from the UAV's real-time position to each transportation point is calculated using a spatial distance formula. Simultaneously, the system incorporates Earth curvature correction parameters to correct the long-distance calculation results, ensuring distance accuracy.

[0068] The system generates drone delivery range data based on the distance calculation results. This data constructs a spatial sphere model with the drone's current location as the center and the remaining flight range as the radius. All logistics transportation points within the sphere are then selected to form a structured dataset containing transportation point identifiers, distance values, and cluster identifiers.

[0069] S12: Perform path optimization processing on the drone delivery range data, call the dynamic programming algorithm to generate multiple alternative transportation routes for each drone, and generate a set of alternative paths for each drone.

[0070] Specifically, the system processes the drone delivery range data, extracts the logistics transportation points within the delivery range as path planning nodes, determines the starting point, ending point and passing nodes according to the logistics task requirements, and constructs a node connection graph, where the weight of the edge in the graph is the actual reachable distance between two nodes.

[0071] Preferably, the system uses a dynamic programming algorithm to generate alternative routes. The algorithm uses the node access order as the decision variable and minimizes the total path distance as the optimization objective, taking into account node access constraints such as time window requirements for some delivery points and the node number limit corresponding to the maximum payload of the drone. In this embodiment, the system can use a state-compressed dynamic programming method, defining the state as "current node + set of visited nodes". The shortest path distance from the starting node to each state is calculated by establishing a state transition equation. The state transition process needs to determine whether the next node is within the delivery range and has not been visited.

[0072] For each drone, the system generates multiple alternative transportation routes. The number is dynamically adjusted according to the number of nodes. For example, 5-8 routes are generated when the number of nodes is less than 10, and 8-12 routes are generated when the number of nodes is greater than 10. This ensures coverage of different combinations of node access sequences. The generated set of alternative routes is stored in the form of a list. Each route includes the node access sequence, total route distance, estimated time, and detailed distance of each segment. It also marks whether the route contains no-fly zone edge segments.

[0073] S13: Perform shortest path filtering on the set of candidate paths for each UAV, select the path with the shortest transport distance as the final flight path, and generate flight path data for each UAV. The flight path data is used to indicate the transport trajectory of the UAV.

[0074] Specifically, the system extracts data from the set of candidate paths for each UAV, obtains the coordinates of the waypoints for each candidate path and the distance parameters between each waypoint, and calculates the total transport distance for each candidate path using a distance accumulation algorithm. During the accumulation process, the straight-line distance between each waypoint needs to be corrected according to the curvature of the actual flight trajectory to ensure that the total distance calculation result is consistent with the actual flight mileage.

[0075] The system compares the total distance of all routes in the candidate path set for the same UAV and selects the path with the shortest total transport distance as the final flight path of the UAV. The final flight path is stored in the form of a waypoint sequence, which includes the three-dimensional coordinates of each waypoint, the estimated arrival time and flight speed limit, and marks the obstacle areas to be avoided and the forced turning points in the path.

[0076] In one embodiment, S2 of the multi-UAV logistics transportation collision prediction method provided by the present invention specifically includes the following steps:

[0077] S21: Perform trajectory function modeling on the flight path data of each UAV, construct a quadratic polynomial trajectory prediction function based on the current position, velocity vector and acceleration vector of each UAV, and generate trajectory function data for each UAV.

[0078] Specifically, the system analyzes the flight path data of each UAV, extracts the three-dimensional coordinates of all waypoints included in each path, the estimated arrival time of each waypoint, and the flight speed limits for different flight segments. It also obtains the current position coordinates of each UAV through a real-time data interface, with the coordinates expressed in longitude, latitude, and altitude in the geodetic coordinate system. The system continuously samples the real-time flight data of the UAVs, with the sampling frequency set according to the flight speed of the UAVs. The higher the speed, the higher the sampling frequency. The sampled data includes the position coordinates at different times.

[0079] The system calculates the position difference between adjacent sampling points based on the continuously sampled position coordinate sequence. Combined with the sampling time interval, it obtains the UAV's velocity vector, which is decomposed into three components along the x, y, and z axes of the geodetic coordinate system. The system then performs the same continuous sampling and calculation on the velocity vector to obtain the acceleration vector. Based on the current position coordinates, velocity vector, and acceleration vector, a quadratic polynomial trajectory prediction function is constructed. The function expression is as follows:

[0080] x(t) = a1t 2+b1t+c1

[0081] y(t)=a2t 2 +b2t+c2

[0082] z(t) = a3t 2 +b3t+c3

[0083] Where t is the time variable, and a1, b1, and c1 represent the polynomial coefficients of the acceleration vector, velocity vector, and current position coordinates, respectively. Preferably, the system can use the least squares method to substitute the sampled position coordinates, velocity components, and acceleration components into the above expression to obtain the specific values ​​of each coefficient. The system integrates the function expressions of the three coordinate axes, the effective start point and end point of the function into trajectory function data, and stores the trajectory function data in a preset trajectory database. The trajectory function data is used in subsequent steps to calculate the relative distance between UAVs.

[0084] S22: Perform relative distance calculation on the trajectory function data of two adjacent UAVs, solve for the minimum Euclidean distance between the trajectories of the two UAVs within a predetermined time window, and generate minimum distance data. The minimum distance data is used to indicate the closest approach distance between the two UAVs during flight.

[0085] Specifically, the system performs pairwise comparisons of the trajectory function data of all UAVs to calculate the projection overlap of the initial paths of any two UAVs in three-dimensional space. The projection overlap is calculated as follows: the initial paths of the two UAVs are projected onto the xy plane, xz plane, and yz plane respectively, and the ratio of the overlap length of the projected trajectories to the length of their respective trajectories is calculated on each plane. The average of the three ratios is taken as the projection overlap in three-dimensional space.

[0086] In this embodiment, the operator presets a projection overlap threshold in the system. When the projection overlap of two drones is greater than the threshold, they are determined to be adjacent drones. Specifically, for each pair of adjacent drones, a predetermined time window is set. The start time of the time window is the current real-time of the system. At the same time, the system reads the maximum range parameter and average flight speed parameter of the drone, calculates the ratio of the maximum range to the average flight speed to obtain the maximum flight time, and adds the current system time to the maximum flight time as the end time of the time window.

[0087] Within a predetermined time window, the system samples the trajectory functions of the two UAVs at fixed time intervals. These intervals are set based on the UAVs' flight speed and a safe distance threshold; higher speeds and smaller thresholds result in shorter time intervals. The system substitutes each sampling moment into the trajectory functions of the two UAVs to obtain the corresponding position coordinates. It then uses the three-dimensional Euclidean distance formula to calculate the distance between the two position coordinates at that moment. The system compares the distance values ​​calculated for all sampling moments and selects the smallest distance value as the minimum Euclidean distance between the two UAV trajectories, recording the sampling moment corresponding to this minimum Euclidean distance.

[0088] Finally, the system integrates the identification information of the two UAVs, the minimum Euclidean distance value and the corresponding time into minimum distance data, and stores the minimum distance data in the distance database. The minimum distance data is used to indicate the closest distance that the two UAVs may reach during flight.

[0089] S23: Perform conflict area identification processing on the minimum distance data. If the minimum distance is less than the preset safe distance threshold, mark the corresponding time and spatial location as a conflict area and generate conflict area information. The conflict area information is used to indicate the spatiotemporal range where obstacle avoidance decision is required.

[0090] Specifically, the safe distance threshold is determined based on the drone's fuselage length, width, height, maximum flight speed, and obstacle avoidance system response time. Different drone models correspond to different safe distance thresholds, and the system matches the corresponding safe distance threshold based on the drone model involved in the minimum distance data.

[0091] Specifically, the system extracts all minimum distance data from the distance database and reads safe distance thresholds from the system's configuration file. These safe distance thresholds are determined based on the drone's fuselage length, width, height, maximum flight speed, and obstacle avoidance system response time. Different drone models correspond to different safe distance thresholds. The system matches the corresponding safe distance threshold to the drone model involved in the minimum distance data. Preferably, the system compares the minimum Euclidean distance in the minimum distance data with the matched safe distance threshold. When the minimum Euclidean distance is less than the safe distance threshold, the system determines that there is a collision risk between the drones.

[0092] The system queries the trajectory function data of the two UAVs in the trajectory database based on the time recorded in the minimum distance data, substitutes the time into the trajectory function, calculates the spatial position coordinates of the two UAVs at that time, takes that time as the center, and combines it with the response time of the UAV obstacle avoidance system to determine the time range. The start time of the time range is the time minus half of the response time, and the end time is the time plus half of the response time. The three-dimensional spatial range is determined with the midpoint of the spatial position coordinates of the two UAVs at that time as the center and the safe distance threshold as the radius.

[0093] The system integrates the time range, spatial range, and identification information of the two involved UAVs into conflict area information, and transmits the conflict area information to the subsequent obstacle avoidance decision-making stage. The conflict area information is used to clarify the time and spatial range in which obstacle avoidance decisions need to be made.

[0094] In one embodiment, S3 of the multi-UAV logistics transportation collision prediction method provided by the present invention specifically includes the following steps:

[0095] S31: Calculate the relative azimuth of the conflict area information, extract the real-time position coordinate data of the two UAVs in the conflict area, and calculate the average relative azimuth angle of the continuous time series.

[0096] Specifically, the system analyzes the time range, spatial range, and the identifiers of the two involved UAVs in the conflict area information. The time range is represented by the start and end timestamps, and the spatial range is defined by three-dimensional coordinate boundary values. The system determines the time interval for collecting real-time position coordinate data based on the size of the conflict area and the flight speed of the UAVs; the smaller the spatial range and the higher the flight speed, the shorter the time interval should be.

[0097] Within a defined timeframe, the system synchronously collects real-time position coordinate data of two UAVs at set time intervals via a wireless data transmission link. The coordinate data can be longitude, latitude, and altitude values ​​in the WGS84 coordinate system. The system sorts the collected coordinate data by timestamp, forming a continuous time series of position coordinate sets. The set contains no fewer than 20 coordinate pairs to ensure calculation accuracy.

[0098] For example, the system uses one drone as a reference point, setting its position coordinates as (x0, y0, z0), and the position coordinates of the other drone as (x, y, z). It then converts the three-dimensional coordinates to planar coordinates using a coordinate transformation formula, obtaining planar coordinates (x0, y0) and (x, y). The system then calculates the relative azimuth angle using the arctangent function and corrects the calculation results according to the coordinate quadrants, obtaining the relative azimuth angle within the range of 0°-360°. The system performs an arithmetic mean calculation on all calculated relative azimuth angles in the continuous time series, removing outliers that deviate from the mean by more than a preset standard deviation during the calculation, resulting in the average relative azimuth angle, which is retained to two decimal places.

[0099] S32: Perform discrete partitioning mapping on the average relative azimuth angle, mapping continuous angle values ​​to a preset set of discrete azimuth angle partitions to generate target azimuth angle partitions.

[0100] Specifically, the system pre-defines a set of discrete azimuth angle partitions. This set divides the continuous angle range of 0°-360° into eight equal partitions, each corresponding to a 45° angle interval: 0°-45°, 45°-90°, ..., 315°-360°. Each partition is assigned a unique partition number, numbered from 1 to 8, corresponding to the aforementioned intervals sequentially. The system compares the calculated average relative azimuth angle with the angle intervals of each partition. If the angle value falls between the start and end values ​​of a certain interval, it is determined to belong to that partition. When the angle value is exactly equal to the end value of the interval, the system assigns it to the next partition, such as assigning 45° to the 45°-90° partition. Based on the comparison results, the system determines the partition number and corresponding angle interval to which the average relative azimuth angle belongs, generating the target azimuth angle partition. The target azimuth angle partition data includes the partition number and the start and end values ​​of the angle interval.

[0101] S33: Based on the protocol obstacle avoidance rule data, perform action query on the target azimuth partition, retrieve predefined obstacle avoidance action vectors according to the UAV's protocol type and azimuth partition, and generate raw obstacle avoidance action data.

[0102] Specifically, the system retrieves obstacle avoidance rule data, which is stored in a structured table. The table includes a protocol type field, an azimuth zone number field, and an obstacle avoidance motion vector field. The protocol type field corresponds to the communication protocol identifier of different manufacturers' drones. The obstacle avoidance motion vector field includes a three-dimensional direction vector and a motion amplitude value. The direction vector is represented as a unit vector (x, y, z), and the amplitude value is represented as an angle change or speed adjustment. The system parses the zone number in the target azimuth zone and simultaneously queries its built-in protocol type information through the drone's identifier. The protocol type information is stored in the drone's device profile data.

[0103] Specifically, the system filters a subset of matching records from the obstacle avoidance rule data based on the UAV's protocol type. Then, it retrieves the corresponding obstacle avoidance action vector from this subset based on the target azimuth partition number. During the retrieval process, the corresponding action vector is extracted when the protocol type field and partition number field are completely identical. The retrieved obstacle avoidance action vector is then associated with the corresponding UAV identifier. Simultaneously, the system determines the action execution time window by combining the time range in the conflict area information. The start time of the time window is no earlier than the start time of the conflict area time range, and the end time is no later than the end time of the conflict area time range. The system integrates the UAV identifier, the direction parameters and amplitude parameters of the obstacle avoidance action vector, and the time window to generate the raw obstacle avoidance action data.

[0104] In one embodiment, such as Figure 2 As shown, S4 of the multi-UAV logistics transportation collision prediction method provided by the present invention specifically includes the following steps:

[0105] S41: Perform motion direction analysis on the original obstacle avoidance motion data, calculate the cosine of the angle between the obstacle avoidance motion vectors of the two UAVs, and generate motion direction difference values.

[0106] Specifically, the system performs data verification on the motion vectors. The verification is based on the maximum speed adjustment threshold recorded in the UAV hardware parameters. If the absolute value of the parameter of any axis in the vector exceeds the threshold, it is determined to be invalid data. At this time, the system calls the historical average motion vector of the UAV model under the same conflict scenario to replace it. After the data verification is completed, the system extracts the motion vector parameters of the two UAVs and determines the directional attributes of the vectors in three-dimensional space.

[0107] Simultaneously, the system calculates the cosine of the angle between the two vectors. The calculation first determines the directional relationship between the vectors, then obtains the cosine value by the ratio of the vector dot product to the magnitude. This value reflects the degree of difference between the two action directions; a smaller value indicates a greater directional difference. The system defines this cosine value as the action direction difference value, generating action direction difference data containing the identifiers of the two UAVs, action vector parameters, and the cosine of the angle. The data is stored in binary format for easy and rapid subsequent retrieval.

[0108] S42: Perform distance impact assessment processing on the minimum distance data, call the exponential decay function to calculate the impact factor of the minimum distance on the collision risk, and generate the distance impact factor.

[0109] Specifically, the system acquires minimum distance data, extracts the minimum relative distance between the two drones, and simultaneously reads the system's preset safe distance threshold. This threshold is determined based on the maximum size of the drone and the flight environment level, with the threshold for urban areas being higher than that for suburban areas. In detail, the system determines the validity of the minimum distance value. If the value is negative or exceeds the preset maximum monitoring distance, a data correction mechanism is activated, replacing the negative value with the weighted average of the five most recent valid distance data sets. The weight increases over time, with the most recent data having the highest weight.

[0110] Preferably, the system calls an exponential decay function to process the effective minimum distance value. The input of this function is the ratio of the minimum distance to the safe distance threshold, and the output is the original influence factor. The smaller the distance, the larger the original influence factor. When the distance reaches the safe distance threshold, the original influence factor drops below 0.1. Simultaneously, the system normalizes the original influence factor, mapping it to a numerical range of 0 to 1, obtaining a distance influence factor. This distance influence factor includes the minimum distance value, the safe distance threshold, the normalized influence factor, and a data correction flag. The data correction flag is a Boolean value used to distinguish between the original data and the corrected data.

[0111] S43: Perform collision probability synthesis processing on the difference value of the action direction and the distance influence factor, combine it with the relative velocity modulus to calculate the collision risk probability, and generate collision risk probability data. The collision risk probability data is used to indicate the collision risk level of the UAV.

[0112] Specifically, the system extracts the difference in motion direction and the distance influence factor. Simultaneously, it obtains the absolute velocity vectors of the two drones through the drone's real-time data interface. The vector parameters are the velocity components along the x, y, and z axes, in meters per second. The system calculates the difference between the two vectors to obtain the relative velocity vector, and then uses a vector magnitude calculation method to obtain the relative velocity magnitude, in meters per second. The system calls preset weighting coefficients α and β. α ranges from 0.5 to 0.7, and β ranges from 0.3 to 0.5. The sum of the two coefficients is 1. The specific values ​​are determined based on the combination of drone types; the α value is higher for multi-rotor drone combinations than for fixed-wing drone combinations. Preferably, the formula for calculating the collision risk probability data is:

[0113] P coll =σ(α*D*F d +β*v rel )

[0114] Among them, P coll For collision risk probability data, σ is the Sigmoid function, α and β are weighting coefficients, D is the difference in action direction, and F... d As the distance influence factor, v relThe relative velocity modulus is used. The system calculates the collision risk probability according to a given formula. First, it calculates the product of α and the difference in motion direction and the distance influence factor. Then, it calculates the product of β and the relative velocity modulus. The two results are added together to obtain an intermediate variable. This intermediate variable is then input into the Sigmoid function to obtain the collision risk probability P. coll This value ranges from 0 to 1. The system determines this based on P. coll The system uses numerical values ​​to classify risk levels: 0 to 0.3 indicates low risk, 0.3 to 0.7 indicates medium risk, and 0.7 to 1 indicates high risk. Based on this processing method, the system generates collision risk probability data, including collision risk probability value, relative velocity magnitude, movement direction difference value, distance influence factor, weight coefficient, and risk level. This data is sent to the UAV control system via an encrypted transmission protocol to trigger corresponding obstacle avoidance strategies.

[0115] In one embodiment, after step S4 of the multi-UAV logistics transportation collision prediction method provided by the present invention, it further includes:

[0116] S51: Perform risk level judgment processing on collision risk probability data. When the risk probability exceeds the preset risk threshold, trigger the obstacle avoidance command compilation process to generate a high-risk warning signal.

[0117] Specifically, the system retrieves collision risk probability data, analyzes the drone identifier pairs, collision risk probabilities, and corresponding risk levels contained in the data, and reads a preset risk threshold from the configuration file. This threshold is set according to the safety level requirements of the drone's operating scenario, with different threshold values ​​corresponding to different scenarios. Specifically, the system compares the risk probability in the collision risk probability data with the preset risk threshold. When the risk probability is greater than the preset risk threshold, it is determined to be a high-risk state. The system triggers obstacle avoidance command compilation, generating a high-risk warning signal. The warning signal includes key information such as the trigger time, the identifier of the relevant drone, and the current collision risk probability. The system stores this warning signal in a log file and sends it in real time to the ground control station and the control terminals of the relevant drones via a communication interface, ensuring that operators can be aware of the current high-risk situation in a timely manner. Simultaneously with triggering the obstacle avoidance command compilation process, the system initializes the relevant parameters and data structures for obstacle avoidance action optimization, preparing for subsequent obstacle avoidance action optimization.

[0118] S52: Perform obstacle avoidance action optimization processing on collision risk probability data and original obstacle avoidance action data. Calculate the optimal obstacle avoidance action by minimizing the collision probability and action offset, and generate optimal obstacle avoidance action data.

[0119] Specifically, the system analyzes the collision risk probability calculation parameters in the collision risk probability data, as well as the obstacle avoidance action vector and the corresponding action offset benchmark value in the original obstacle avoidance action data. The action offset benchmark value is the upper limit of the deviation between the actual trajectory and the original planned trajectory when the UAV flies according to the original obstacle avoidance action.

[0120] Preferably, the system constructs an optimization model aimed at minimizing collision probability and action deviation. The model's constraints include the UAV's maximum turning angular velocity, maximum climb / descent rate, and action execution time limits. The system solves this optimization model iteratively, adjusting the values ​​of each component of the obstacle avoidance action vector while satisfying the constraints. This reduces the collision probability to below a preset risk threshold, while ensuring the action deviation is less than a baseline value. The system then organizes the solved action vectors into optimal obstacle avoidance action data, which includes the UAV identifier, the three-dimensional components of the action vector, the action execution start time, and the duration.

[0121] During the optimization process, the system calculates the adjusted collision risk probability and action offset in real time until a preset convergence condition is met, i.e., both the collision probability and action offset are minimized. The system generates optimal obstacle avoidance action data, including detailed information such as the optimized obstacle avoidance action type, execution order, duration, and expected degree of collision risk reduction. The system stores the optimal obstacle avoidance action data in memory and prepares it for protocol instruction conversion processing to ensure that the optimized obstacle avoidance action can be accurately transmitted to the UAV and executed.

[0122] S53: Performs protocol instruction conversion processing on high-risk warning signals and optimal obstacle avoidance action data. Based on the target UAV's protocol type, it compiles the action vector into binary instruction code to generate UAV obstacle avoidance instructions. These instructions are used to instruct the UAV to perform three-dimensional obstacle avoidance actions.

[0123] Specifically, the system queries the UAV protocol type database, retrieves the corresponding protocol type based on the UAV identifier, and the protocol type includes information such as communication frame format, command encoding rules and verification methods; and according to the retrieved protocol type, calls the corresponding protocol conversion algorithm to convert the motion vector in the optimal obstacle avoidance action data into command parameters that conform to the protocol specifications. The parameters include the adjustment amount in three-dimensional direction, the action execution speed and duration.

[0124] For example, the system compiles the instruction parameters into binary instruction code according to the frame structure specified in the protocol. During the compilation process, fields such as frame header, frame trailer, device address, and checksum are added to ensure the integrity and accuracy of the instruction during transmission. The system generates an obstacle avoidance instruction for the UAV, which includes the unique identifier of the target UAV, the binary instruction code, and the instruction execution priority. This instruction is sent to the target UAV through a wireless communication interface to instruct the UAV to perform corresponding obstacle avoidance actions in three-dimensional space.

[0125] During the conversion process, the system strictly adheres to the encoding rules specified in the protocol, such as mapping enumerated values ​​for action types, checking the numerical range of execution parameters, and converting data types. The system generates obstacle avoidance commands for the UAV, containing complete obstacle avoidance action information that can be directly parsed and executed by the UAV's flight control system. The system sends the generated obstacle avoidance commands to the target UAV via the communication module, simultaneously recording the command transmission status and timestamp to ensure the timeliness and reliability of the commands. After sending the commands, the system initiates a command execution status monitoring program to track the UAV's response to the obstacle avoidance commands in real time, providing a basis for subsequent flight status assessment and command adjustments.

[0126] In summary, the collision prediction method for multi-UAV logistics transportation provided in this application can achieve dynamic threshold response of collision risk probability, thereby solving the problem of delayed obstacle avoidance timing caused by risk response lag in traditional methods; the multi-objective optimization calculation based on collision probability and action offset can achieve precise and smooth adjustment of obstacle avoidance actions, effectively eliminating the risk of secondary collisions caused by sudden action changes; and the protocol-adaptive instruction conversion technology can achieve accurate compilation of instructions from heterogeneous UAV platforms, thereby solving the instruction compatibility problem in scenarios with multiple protocols.

[0127] This process can construct a closed-loop system encompassing risk warning, action optimization, and command generation, achieving a dual improvement in the timeliness and reliability of obstacle avoidance decisions. It can complete the standardized conversion of three-dimensional spatial action vectors, ensuring precise execution of obstacle avoidance commands in pitch, yaw, and roll dimensions. It can form a millisecond-level response chain from risk prediction to action execution, fundamentally enhancing safety protection capabilities in complex logistics scenarios. Through dynamic risk response, smooth action optimization, and adaptive protocol conversion, this mechanism effectively solves the command execution deviation problem in high-density UAV logistics transportation, providing highly reliable obstacle avoidance guarantees for heterogeneous UAV swarms.

[0128] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0129] Based on the same inventive concept, this application also provides a multi-UAV logistics transportation collision prediction system for implementing the multi-UAV logistics transportation collision prediction method described above. The solution provided by this system is similar to the implementation scheme described in the above method. Therefore, the specific limitations of one or more multi-UAV logistics transportation collision prediction system embodiments provided below can be found in the limitations of the multi-UAV logistics transportation collision prediction method described above, and will not be repeated here.

[0130] Preferably, such as Figure 3 As shown, the present invention provides a multi-UAV logistics transportation collision prediction system 600, which is configured with the following modules:

[0131] The flight path generation module 610 is used to acquire the coordinate data of the logistics transportation point and the real-time flight data of each UAV, calculate the spatial relationship between the logistics transportation point and the flight distance of the UAV, and generate the flight path data of each UAV.

[0132] The conflict area detection module 620 is used to detect conflict areas in flight path data, calculate the minimum relative distance between two adjacent UAV trajectories, filter trajectories intersecting, and generate conflict area information and minimum distance data.

[0133] The obstacle avoidance action mapping module 630 is used to query obstacle avoidance actions for conflict area information, call the protocol obstacle avoidance rules to query the protocol action mapping relationship according to the relative azimuth angle partition, and generate the original obstacle avoidance action data.

[0134] The collision risk prediction module 640 is used to predict collision risks from the original obstacle avoidance action data and minimum distance data. It combines the difference in obstacle avoidance action direction and relative speed to calculate the probability of conflict and generate collision risk probability data.

[0135] Preferably, the flight path generation module 610 provided in this application is configured with the following units:

[0136] The delivery range calculation unit is used to acquire the coordinate data of logistics transportation points and the real-time flight data of each drone, perform spatial distribution analysis on the coordinate data, calculate the distance relationship between the drone and each logistics transportation point based on the geographical location of the logistics transportation point, and generate drone delivery range data.

[0137] The alternative route generation unit is used to perform path optimization processing on the drone delivery range data, call the dynamic programming algorithm to generate multiple alternative transportation routes for each drone, and generate an alternative route set for each drone.

[0138] The shortest path filtering unit is used to filter the shortest path from the set of candidate paths for each UAV, select the path with the shortest transportation distance as the final flight path, and generate flight path data for each UAV. The flight path data is used to indicate the transportation trajectory of the UAV.

[0139] Preferably, the conflict zone detection module 620 provided in this application is configured with the following units:

[0140] The trajectory function modeling unit is used to perform trajectory function modeling on the flight path data of each UAV. Based on the current position, velocity vector and acceleration vector of each UAV, a quadratic polynomial trajectory prediction function is constructed to generate the trajectory function data of each UAV.

[0141] The relative distance calculation unit is used to perform relative distance calculation on the trajectory function data of two adjacent UAVs, solve the minimum Euclidean distance between the trajectories of the two UAVs within a predetermined time window, and generate minimum distance data to indicate the closest approach distance between the two UAVs during flight.

[0142] The conflict zone marking unit is used to identify conflict zones in the minimum distance data. If the minimum distance is less than the safe distance threshold, the corresponding time and spatial location are marked as conflict zones, and conflict zone information is generated. The conflict zone information is used to indicate the spatiotemporal range where obstacle avoidance decisions are needed.

[0143] Preferably, the obstacle avoidance motion mapping module 630 provided in this application is configured with the following units:

[0144] The relative orientation calculation unit is used to calculate the relative orientation of conflict area information, extract the real-time position coordinate data of two UAVs in the conflict area, and calculate the average relative orientation angle of a continuous time series.

[0145] The azimuth partitioning mapping unit is used to perform discrete partitioning mapping on the average relative azimuth angle, mapping continuous angle values ​​to a preset set of discrete azimuth angle partitions to generate target azimuth angle partitions.

[0146] The obstacle avoidance action retrieval unit is used to perform action queries on the target azimuth partition based on protocol obstacle avoidance rule data. It retrieves predefined obstacle avoidance action vectors according to the UAV's protocol type and azimuth partition, and generates raw obstacle avoidance action data.

[0147] Preferably, the collision risk prediction module 640 provided in this application is configured with the following units:

[0148] The motion direction difference calculation unit is used to perform motion direction analysis and processing on the original obstacle avoidance motion data, calculate the cosine value of the angle between the obstacle avoidance motion vectors of the two UAVs, and generate motion direction difference values.

[0149] The distance impact factor assessment unit is used to perform distance impact assessment on the minimum distance data, call the exponential decay function to calculate the impact factor of the minimum distance on the collision risk, and generate the distance impact factor.

[0150] The collision risk probability synthesis unit is used to perform collision probability synthesis processing on the difference value of the action direction and the distance influence factor, and calculate the collision risk probability by combining the relative velocity modulus to generate collision risk probability data. The collision risk probability data is used to indicate the collision risk level of the UAV.

[0151] Preferably, the multi-UAV logistics transportation collision prediction system 600 provided in this application is further configured with the following units:

[0152] The risk level judgment unit is used to judge the risk level of collision risk probability data. When the risk probability exceeds the preset risk threshold, the obstacle avoidance command compilation process is triggered to generate a high-risk warning signal.

[0153] The obstacle avoidance action optimization unit is used to optimize the obstacle avoidance action by processing the collision risk probability data and the original obstacle avoidance action data. It calculates the optimal obstacle avoidance action by minimizing the collision probability and the action offset, and generates the optimal obstacle avoidance action data.

[0154] The protocol instruction conversion unit is used to process the high-risk warning signal and the optimal obstacle avoidance action data into protocol instructions. According to the protocol type of the target UAV, the action vector is compiled into binary instruction code to generate UAV obstacle avoidance instructions. The UAV obstacle avoidance instructions are used to instruct the UAV to perform three-dimensional obstacle avoidance actions.

[0155] In one embodiment, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the above-described multi-UAV logistics transportation collision prediction method.

[0156] In one embodiment, this application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described multi-UAV logistics transportation collision prediction method.

[0157] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0158] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0159] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in this application, and these should all be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for predicting collisions in multi-UAV logistics transportation, characterized in that, Includes the following steps: S1: Obtain the coordinate data of the logistics transportation point and the real-time flight data of each drone, calculate the spatial relationship between the logistics transportation point and the flight distance of the drone, and generate the flight path data of each drone. S2: Detect conflict areas in the flight path data, calculate the minimum relative distance between two adjacent UAV trajectories, filter the trajectory intersection areas, and generate conflict area information and minimum distance data; S3: Perform obstacle avoidance action query on the conflict area information, call the protocol obstacle avoidance rules to query the protocol action mapping relationship according to the relative azimuth angle partition, and generate the original obstacle avoidance action data; S4: Perform collision risk prediction on the original obstacle avoidance action data and the minimum distance data, and calculate the conflict probability by combining the difference in obstacle avoidance action direction and relative speed to generate collision risk probability data; The step of querying obstacle avoidance actions for the conflict area information, invoking protocol obstacle avoidance rules to query protocol action mapping relationships based on relative azimuth angle partitions, and generating original obstacle avoidance action data includes: S31: Calculate the relative azimuth of the conflict area information, extract the real-time position coordinate data of the two UAVs in the conflict area, and calculate the average relative azimuth angle of the continuous time series. S32: Perform discrete partitioning mapping on the average relative azimuth angle, mapping continuous angle values ​​to a preset set of discrete azimuth angle partitions to generate target azimuth angle partitions; S33: Based on the protocol obstacle avoidance rule data, perform action query on the target azimuth partition, retrieve predefined obstacle avoidance action vectors according to the UAV's protocol type and azimuth partition, and generate original obstacle avoidance action data; The step of performing collision risk prediction on the original obstacle avoidance action data and the minimum distance data, and calculating the collision probability by combining the difference in obstacle avoidance action direction and relative speed, to generate collision risk probability data includes: S41: Perform motion direction analysis on the original obstacle avoidance motion data, calculate the cosine of the angle between the obstacle avoidance motion vectors of the two UAVs, and generate motion direction difference values. S42: Perform distance impact assessment processing on the minimum distance data, call the exponential decay function to calculate the impact factor of the minimum distance on the collision risk, and generate the distance impact factor; S43: Perform conflict probability synthesis processing on the difference value of the action direction and the distance influence factor, and calculate the collision risk probability by combining the relative velocity modulus to generate collision risk probability data. The collision risk probability data is used to indicate the collision risk level of the UAV. The calculation formula for the collision risk probability data is as follows: ； in, This is collision risk probability data. For the Sigmoid function, , These are the weighting coefficients. This represents the difference in the direction of movement. Distance is the influencing factor. This represents the relative velocity magnitude.

2. The method according to claim 1, characterized in that, S1 includes: S11: Obtain the coordinate data of logistics transportation points and the real-time flight data of each drone, perform spatial distribution analysis on the coordinate data, calculate the distance relationship between the drone and each logistics transportation point based on the geographical location of the logistics transportation points, and generate drone delivery range data. S12: Perform path optimization processing on the drone delivery range data, call the dynamic programming algorithm to generate multiple alternative transportation routes for each drone, and generate a set of alternative paths for each drone. S13: Perform shortest path filtering on the candidate path set for each UAV, select the path with the shortest transportation distance as the final flight path, and generate flight path data for each UAV. The flight path data is used to indicate the transportation trajectory of the UAV.

3. The method according to claim 1, characterized in that, S2 includes: S21: Perform trajectory function modeling processing on the flight path data of each UAV, construct a quadratic polynomial trajectory prediction function based on the current position, velocity vector and acceleration vector of each UAV, and generate trajectory function data of each UAV; S22: Perform relative distance calculation processing on the trajectory function data of two adjacent UAVs, solve the minimum Euclidean distance between the trajectories of the two UAVs within a predetermined time window, and generate minimum distance data. The minimum distance data is used to indicate the closest approach distance between the two UAVs during flight. S23: Perform conflict area identification processing on the minimum distance data. If the minimum distance is less than the preset safe distance threshold, mark the corresponding time and spatial location as a conflict area and generate conflict area information. The conflict area information is used to indicate the spatiotemporal range where obstacle avoidance decision is required.

4. The method according to claim 1, characterized in that, Following S4, the following is also included: S51: Perform risk level judgment processing on the collision risk probability data. When the risk probability exceeds the preset risk threshold, trigger the obstacle avoidance command compilation process to generate a high-risk warning signal. S52: Perform obstacle avoidance action optimization processing on the collision risk probability data and the original obstacle avoidance action data, calculate the optimal obstacle avoidance action by minimizing the collision probability and action offset, and generate the optimal obstacle avoidance action data; S53: Perform protocol instruction conversion processing on the high-risk warning signal and the optimal obstacle avoidance action data, compile the action vector into binary instruction code according to the protocol type of the target UAV, and generate UAV obstacle avoidance instructions. The UAV obstacle avoidance instructions are used to instruct the UAV to perform three-dimensional space obstacle avoidance actions.

5. A multi-UAV logistics transportation collision prediction system, characterized in that, The system includes: The flight path generation module is used to acquire the coordinate data of logistics transportation points and the real-time flight data of each UAV, calculate the spatial relationship between the logistics transportation points and the flight distance of the UAVs, and generate the flight path data of each UAV. The conflict area detection module is used to detect conflict areas in the flight path data, calculate the minimum relative distance between two adjacent UAV trajectories, filter the trajectory intersection areas, and generate conflict area information and minimum distance data. The obstacle avoidance action mapping module is used to query the obstacle avoidance action information of the conflict area, call the protocol obstacle avoidance rules to query the protocol action mapping relationship according to the relative azimuth angle, and generate the original obstacle avoidance action data. The collision risk prediction module is used to predict the collision risk of the original obstacle avoidance action data and the minimum distance data, and calculate the collision probability by combining the difference in obstacle avoidance action direction and relative speed to generate collision risk probability data. The step of querying obstacle avoidance actions for the conflict area information, invoking protocol obstacle avoidance rules to query protocol action mapping relationships based on relative azimuth angle partitions, and generating original obstacle avoidance action data includes: S31: Calculate the relative azimuth of the conflict area information, extract the real-time position coordinate data of the two UAVs in the conflict area, and calculate the average relative azimuth angle of the continuous time series. S32: Perform discrete partitioning mapping on the average relative azimuth angle, mapping continuous angle values ​​to a preset set of discrete azimuth angle partitions to generate target azimuth angle partitions; S33: Based on the protocol obstacle avoidance rule data, perform action query on the target azimuth partition, retrieve predefined obstacle avoidance action vectors according to the UAV's protocol type and azimuth partition, and generate original obstacle avoidance action data; The step of performing collision risk prediction on the original obstacle avoidance action data and the minimum distance data, and calculating the collision probability by combining the difference in obstacle avoidance action direction and relative speed, to generate collision risk probability data includes: S41: Perform motion direction analysis on the original obstacle avoidance motion data, calculate the cosine of the angle between the obstacle avoidance motion vectors of the two UAVs, and generate motion direction difference values. S42: Perform distance impact assessment processing on the minimum distance data, call the exponential decay function to calculate the impact factor of the minimum distance on the collision risk, and generate the distance impact factor; S43: Perform conflict probability synthesis processing on the difference value of the action direction and the distance influence factor, and calculate the collision risk probability by combining the relative velocity modulus to generate collision risk probability data. The collision risk probability data is used to indicate the collision risk level of the UAV. The calculation formula for the collision risk probability data is as follows: ； in, This is collision risk probability data. For the Sigmoid function, , These are the weighting coefficients. This represents the difference in the direction of movement. Distance is the influencing factor. This represents the relative velocity magnitude.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 4.

Images