Low-altitude logistics unmanned aerial vehicle path optimization and multi-machine cooperative distribution scheduling method and system

Abstract

The application discloses a low-altitude logistics unmanned aerial vehicle path optimization and multi-machine cooperative distribution scheduling method and system, and relates to the field of modern aviation logistics. The application obtains real-time operation parameters and order insertion attributes of the unmanned aerial vehicle; extracts local micro-meteorological wind resistance characteristics based on a flight control differential torque decoupling mechanism, and updates a resistance coefficient matrix; calculates dynamic space-time computing power excluding internal resistance heat loss in combination with residual available power and maximum available load; reconstructs multi-machine task handover space boundary by using a dynamic weighted Thiessen polygon according to the resistance matrix and the computing power; iteratively optimizes by using an improved ant colony algorithm introducing wind resistance, power penalty and load negative feedback pheromone in the boundary; and outputs and issues an optimal cooperative distribution matrix and a dynamic obstacle avoidance trajectory. The application deeply couples instantaneous physical constraints of the unmanned aerial vehicle and local environmental resistance, effectively avoids multi-machine cooperative conflicts and path deadlocks, and realizes safe scheduling with reduced consumption and increased efficiency.

Description

Technical Field

[0001] This invention relates to the field of modern aviation logistics technology, and in particular to a method and system for low-altitude logistics drone path optimization and multi-drone collaborative delivery scheduling. Background Technology

[0002] With the rapid development of the urban low-altitude economy, drones are widely used in urban last-mile logistics and emergency material delivery. In complex, high-density urban microenvironments (such as the canyon wind effect created by towering buildings), multi-drone collaborative delivery networks often face extremely limited scenarios with sudden high-priority orders (such as medical emergency supplies orders).

[0003] Existing drone scheduling methods suffer from the following technical shortcomings: When scheduling and planning routes, the systems often rely excessively on macro-weather data provided by meteorological bureaus and the static remaining battery power displayed on the drone's dial. When encountering sudden localized micro-weather changes (such as sudden crosswinds) or unexpected order insertions, existing technologies fail to deeply couple the drone's actual carrying capacity, which decreases non-linearly with wind resistance and load, with the complex dynamic low-altitude urban environment. In handling emergency order insertions, the lack of a mechanism to dynamically redraw task handover boundaries based on real-time environmental resistance and the actual available computing power of each drone leads to frequent task conflicts, high redundant energy consumption, and even forced landings due to battery depletion during multi-drone operations. Summary of the Invention

[0004] This invention provides a method and system for low-altitude logistics drone path optimization and multi-drone collaborative delivery scheduling. The technical problem to be solved is that the existing scheduling method cannot take into account both the wind resistance of the local micro-environment and the dynamic carrying capacity of the drone, resulting in low collaborative efficiency and high energy consumption when dealing with sudden order insertions.

[0005] To achieve the aforementioned objectives, the technical solution adopted by this invention is: a method for low-altitude logistics drone path optimization and multi-drone collaborative delivery scheduling, comprising:

[0006] S1. Obtain the real-time operating parameters of each UAV in the multi-UAV collaborative network, including the three-dimensional position of the UAV body, and receive the sudden insertion attributes of sudden insertion, wherein the sudden insertion attributes include the insertion target coordinates and physical weighing mass.

[0007] S2. Obtain the local micro-meteorological wind resistance characteristics of the airspace where the multi-machine collaborative network is located, and refresh the drag coefficient matrix of the three-dimensional spatiotemporal grid based on the local micro-meteorological wind resistance characteristics.

[0008] S3. Calculate the dynamic spatiotemporal computing power of each UAV in the current flight cycle by combining the remaining available power and maximum available load in the real-time operation parameters.

[0009] S4. Based on the resistance coefficient matrix and the dynamic spatiotemporal computing power, reconstruct the task handover space boundary of each delivery responsibility area of ​​the package in the multi-machine collaborative network;

[0010] S5. Within the boundary of the task handover space, an improved ant colony algorithm is used to perform path optimization iteration, outputting the optimal collaborative delivery matrix for the sudden order insertion attribute and the corresponding dynamic obstacle avoidance trajectory of each UAV, and sending the dynamic obstacle avoidance trajectory to the corresponding UAV to instruct the flight control system of the corresponding UAV to perform obstacle avoidance flight.

[0011] Further, in step S2, the local micro-meteorological wind resistance characteristics of the airspace where the multi-machine cooperative network is located are obtained, including:

[0012] S21. Real-time monitoring of the six-degree-of-freedom physical flight attitude of the preceding UAV in the forward detection route in three-dimensional space, and extracting the pulse width modulation duty cycle difference value of each rotor motor under the state of maintaining the horizontal balance of the body.

[0013] S22. Based on the difference in pulse width modulation duty cycle, filter out the reference static torque generated by the weight of the airframe and separate the net yaw compensation torque specifically used to resist lateral environmental wind disturbance as multi-rotor differential torque decoupling data.

[0014] S23. Input the multi-rotor differential torque decoupling data into a preset aerodynamic inverse transformation model for inverse calculation, and output the local micro-meteorological wind resistance characteristics corresponding to the three-dimensional spatial position.

[0015] Furthermore, step S23 specifically includes:

[0016] S231, according to the formula Obtain the local wind speed drag scalar, where R is the local wind speed drag scalar. The net yaw compensation torque in the multi-rotor differential torque decoupling data, where S is the frontal area coefficient and K is a preset aerodynamic calibration constant parameter. air density;

[0017] S232. The local wind speed drag scalar is combined with the aircraft yaw angle attitude data to convert it into a three-dimensional direction vector, generating the local micro-meteorological wind drag feature containing absolute direction attributes.

[0018] S233. Write the local micro-meteorological wind resistance characteristics into the corresponding spatial grid nodes of the three-dimensional environmental blank matrix, and use the Gaussian filtering algorithm to perform spatial interpolation smoothing on the unsampled adjacent nodes to output the drag coefficient matrix.

[0019] Furthermore, step S3 specifically includes:

[0020] S31. Extract the real-time battery internal resistance temperature, remaining available power, and instantaneous bus discharge current of each UAV from the real-time operating parameters.

[0021] S32, According to the formula Calculate the initial spatiotemporal computing power scalar for each UAV, where C is the initial spatiotemporal computing power scalar. For the current moment The remaining available power, From the current moment up to the predicted time The nonlinear internal resistance heat loss electrical energy integral, I is the instantaneous bus discharge current, W is the maximum available load, and r is the mapped resistance value obtained by conversion through the preset temperature resistance conversion function;

[0022] S33. Perform global extreme value normalization on the initial spatiotemporal computing power scalar throughout the multi-machine collaborative network, and output the normalized dynamic spatiotemporal computing power.

[0023] Furthermore, step S4 specifically includes:

[0024] S41. Set the dynamic spatiotemporal computing power as the central expansion capability index of the corresponding geographical node in the multi-machine collaborative network, and use it as the expansion denominator term for generating dynamic weighted Thiessen polygon rules.

[0025] S42. Taking the three-dimensional position of each UAV as the pole, according to the formula... Calculate the spatial expansion cost distance, where D is the spatial expansion cost distance and d is the Euclidean physical measured distance. C represents the value corresponding to the drag coefficient matrix, and C represents the dynamic spatiotemporal computing power.

[0026] S43. Compare the spatial expansion cost distances of adjacent poles, generate dynamic intersecting dividing lines at the intersection of spatial surfaces with equal costs, and perform continuous topological closure splicing to form the task handover space boundary.

[0027] Further, in step S5, an improved ant colony algorithm is used for path optimization iteration within the task handover space boundary, including:

[0028] S51. In the improved ant colony algorithm, a custom comprehensive cost heuristic function is introduced, which includes the wind resistance penalty parameter in the drag coefficient matrix and the power exponential decay penalty component.

[0029] S52. Construct a negative feedback pheromone local rapid volatilization rule parameter based on the product of the actual mounted effective payload mass data and the local micro-meteorological wind resistance characteristics;

[0030] S53. Based on the comprehensive cost heuristic function and the negative feedback pheromone local rapid evaporation rule parameter, the optimal collaborative delivery matrix is ​​generated by multi-generation optimization iteration of the physical delivery route within the task handover space boundary.

[0031] Further, in step S51, a custom comprehensive cost heuristic function is introduced into the improved ant colony algorithm, including:

[0032] S511. Calculate the spatial Euclidean distance between the current 3D spatial node and the target adjacent raster node to be expanded, and generate the basic path length heuristic factor.

[0033] S512. Extract the modulus of the local micro-meteorological wind resistance characteristics from the drag coefficient matrix as a wind resistance penalty parameter, and generate a power exponential decay penalty component based on the dynamic spatiotemporal computing power.

[0034] S513, according to the formula Generate the comprehensive cost heuristic function, where Let be the comprehensive cost heuristic function, and d be the basic path length heuristic factor. C represents the wind resistance penalty parameter, and C represents the dynamic spatiotemporal computing power. These are preset fixed weighting coefficients.

[0035] Furthermore, step S52 specifically includes:

[0036] S521. After a single blind path search iteration, extract the corresponding actual mounted effective payload mass data and the modulus of the local micro-meteorological wind resistance characteristics corresponding to the path three-dimensional grid.

[0037] S522. Multiply the actual mounted effective load mass data with the modulus of the local micro-meteorological wind resistance characteristics by tensor values ​​to obtain the high-risk load resistance product spatial variable.

[0038] S523, Substitute the high-risk load resistance product space variable into the pheromone update formula. ,in The updated pheromone values. M represents the base value of the pheromone from the original path, and M represents the actual mass of the mounted payload. This refers to the modulus value of the local micro-meteorological wind resistance characteristics. This is the preset environmental volatility coefficient.

[0039] Another objective of this invention is to provide a low-altitude logistics drone path optimization and multi-drone collaborative delivery scheduling system, comprising:

[0040] The system operates a data acquisition module, which acquires real-time operating parameters of each UAV in the multi-UAV collaborative network, including the three-dimensional position of the UAV body, and receives the sudden insertion attributes of sudden insertions, wherein the sudden insertion attributes include the target coordinates of the insertion and the physical weighing mass.

[0041] An environmental resistance construction module acquires the local micro-meteorological wind resistance characteristics of the airspace where the multi-machine collaborative network is located, and refreshes the resistance coefficient matrix of the three-dimensional spatiotemporal grid based on the local micro-meteorological wind resistance characteristics.

[0042] The dynamic computing power evaluation module combines the remaining available power and maximum available load in the real-time operating parameters to calculate the dynamic spatiotemporal computing power of each UAV in the current flight cycle.

[0043] The handover boundary reconstruction module reconstructs the task handover spatial boundary of each delivery responsibility area of ​​the package in the multi-machine collaborative network based on the resistance coefficient matrix and the dynamic spatiotemporal computing power.

[0044] The collaborative scheduling optimization module uses an improved ant colony algorithm to iterate path optimization within the task handover space boundary, outputs the optimal collaborative delivery matrix for the sudden order insertion attribute and the corresponding dynamic obstacle avoidance trajectory of each UAV, and sends the dynamic obstacle avoidance trajectory to the corresponding UAV to instruct the flight control system of the corresponding UAV to perform obstacle avoidance flight.

[0045] Furthermore, the operational data acquisition module includes:

[0046] The message unpacking unit receives the sudden order logistics data stream message and extracts the target coordinates of the insertion target and the extreme delivery time threshold from the sudden order logistics data stream message.

[0047] A mass sensing unit acquires the physical weighing sensing electrical signal of the package, and processes the physical weighing sensing electrical signal through analog-to-digital conversion to generate a digital quantity as the physical weighing mass.

[0048] The attribute assembly unit splices the target coordinates of the inserted order, the extreme delivery time threshold, and the physical weighing mass into a low-level data packet to generate a structured burst order attribute, which is then sent to the collaborative scheduling optimization module.

[0049] The beneficial effects of this invention are as follows:

[0050] 1. This invention actively acquires the local micro-meteorological wind resistance characteristics of the airspace where the multi-machine collaborative network is located and refreshes the drag coefficient matrix, breaking the dependence of traditional scheduling on external macro weather data. This enables the system to respond to complex urban micro-environmental changes in seconds, providing a high-precision three-dimensional environmental foundation for underlying scheduling.

[0051] 2. This invention calculates dynamic spatiotemporal computing power by combining the remaining available power and the maximum available load, and introduces the nonlinear decay characteristics of physical batteries affected by load and internal resistance into the mathematical model to obtain the actual order-taking capacity of drones, thus completely avoiding the risk of low-power / low-load drones being wrongly assigned orders.

[0052] 3. Based on real-time drag coefficient matrix and dynamic spatiotemporal computing power, this invention reconstructs the task handover spatial boundary of each delivery responsibility area of ​​the package; this mechanism eliminates the ambiguity of task boundaries in multi-machine networks from a mathematical and geometric perspective, realizes flexible territory division driven by both computing power and environment, and ensures the most efficient takeover of sudden order interruptions by drones;

[0053] 4. This invention employs an improved ant colony algorithm for multi-generation optimization iteration within the reconstructed boundary, which not only outputs the optimal collaborative delivery matrix but also directly sends the generated dynamic obstacle avoidance trajectory to the corresponding UAV's flight control system for execution, achieving end-to-end energy-saving and efficiency-enhancing control from perception to physical action execution. Attached Figure Description

[0054] Figure 1 This is a flowchart illustrating the low-altitude logistics drone path optimization and multi-drone collaborative delivery scheduling method provided in an embodiment of the present invention.

[0055] Figure 2 This is a block diagram of the module structure of the low-altitude logistics drone dynamic path optimization and multi-drone collaborative delivery scheduling system provided in an embodiment of the present invention;

[0056] Figure 3 This is a schematic diagram illustrating the principle of multi-machine collaborative network based on dynamic weighted Thiessen polygon reconstruction task handover space boundary in an embodiment of the present invention;

[0057] Figure 4 This is a schematic diagram illustrating the improved ant colony algorithm performing obstacle avoidance path optimization iteration in the resistance coefficient matrix in an embodiment of the present invention. Detailed Implementation

[0058] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings. Identical components are indicated by the same reference numerals.

[0059] It should be noted that the terms “front,” “back,” “left,” “right,” “up,” and “down” used in the following description refer to the directions shown in the attached diagram, while the terms “inside” and “outside” refer to the directions toward or away from the geometric center of a specific component, respectively.

[0060] To make the content of this invention easier to understand, the technical solutions in the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings.

[0061] like Figure 1 As shown, this invention provides a method for low-altitude logistics drone path optimization and multi-drone collaborative delivery scheduling, including the following steps:

[0062] Step S1: Obtain the real-time operating parameters of each UAV in the multi-UAV collaborative network, including the three-dimensional position of the UAV body, and receive the sudden insertion attributes of sudden insertion, wherein the sudden insertion attributes include the insertion target coordinates and physical weighing mass.

[0063] Explanation: This feature is the fundamental data input port for the entire dispatch system. A multi-drone collaborative network refers to a delivery communication system composed of multiple drones and a cloud server. In the initial stages of operation, the system must monitor the real-time geographic coordinates and current operational status of each drone. When an emergency task is accessed, the system must obtain the task's three-dimensional geographic location and the actual weight of the package. Physical weighing mass is a core prerequisite variable for subsequent calculations of drone energy consumption and flight drag, as well as for triggering pheromone decay mechanisms.

[0064] Technical effect: Ensures that the system has a global perspective, providing accurate basic physical basis for subsequent spatiotemporal computing power assessment and multi-machine collaborative allocation.

[0065] Example: In a densely populated high-rise residential area, a drone logistics delivery network has been deployed. Currently, the cloud-based dispatch center is tracking the three-dimensional positions of drones A, B, and C within the network in real time. Suddenly, the system receives an order for emergency medical supplies to be delivered to a nearby target community. The system immediately records the destination coordinates of the order and the physical weight of the medical supply box.

[0066] Step S2: Real-time monitoring of the six-degree-of-freedom physical flight attitude of the preceding UAV in the forward detection route in three-dimensional space, extracting the pulse width modulation duty cycle difference value of each rotor motor under the state of maintaining the horizontal balance of the body; filtering out the reference static torque generated by the self-weight of the body based on the pulse width modulation duty cycle difference value, and separating the net yaw compensation torque specifically used to resist the lateral environmental wind disturbance as the multi-rotor differential torque decoupling data.

[0067] Explanation: This feature is one of the core innovations, aiming to solve the challenge of sensing meteorological drag in the micro-environment of urban high-rise buildings. Mature UAV flight control systems employ multi-rotor differential control technology. When encountering crosswinds, the aircraft automatically adjusts the rotational speeds of different rotors (i.e., changing the pulse-width modulation duty cycle of the motor's electrical signal) to resist wind force in order to avoid being overturned. This principle reverses this flight control mechanism, separating the extra force used to resist wind force (i.e., net yaw compensation torque) from the total force. This allows the UAV to act as a high-precision weather detector while delivering goods, eliminating the need for external weather radar.

[0068] Technical effect: It breaks through the dependence of traditional scheduling on macro weather forecasts from external meteorological bureaus and realizes second-level self-sensing and self-extraction of local micro-environment meteorological data.

[0069] Example: Drone A is passing between two high-rise residential buildings and encounters a strong "canyon wind effect." To maintain its yaw stability, Drone A's flight control system automatically adjusts its rotor speed. The cloud system captures the difference in pulse width modulation duty cycle of the electrical signals between Drone A's rotor speeds in real time. After deducting the hovering force required for Drone A's own weight, it precisely isolates the net yaw compensation torque specifically designed to counteract this canyon wind.

[0070] Furthermore, the multi-rotor differential torque decoupling data is input into a preset aerodynamic inverse transformation model for inverse calculation, outputting the local micro-meteorological drag characteristics corresponding to the three-dimensional spatial location. According to the formula... The local wind speed drag scalar is obtained, and the local wind speed drag scalar is combined with the aircraft yaw angle attitude data to transform it into a three-dimensional direction vector, generating local micro-meteorological wind drag characteristics containing absolute direction attributes; Gaussian filtering algorithm is used to perform spatial interpolation smoothing on unsampled adjacent nodes to output the drag coefficient matrix.

[0071] Explanation: Building upon the previous steps, this feature further quantifies the extracted physical torque into standard aerodynamic drag parameters. In the formula, R represents the local wind speed drag scalar; The net yaw compensation torque in the multi-rotor differential torque decoupling data; S is the frontal area coefficient; K is the preset aerodynamic calibration constant parameter; Let be the air density. Using this formula, the magnitude of wind resistance can be accurately calculated. Then, combined with the drone's current nose-up orientation (yaw angle), it is transformed into a three-dimensional vector with wind direction. Finally, a Gaussian filtering algorithm is used for mathematical smoothing, filling in the blank data in the surrounding unflyed areas to form a continuous drag coefficient matrix.

[0072] Technical effect: It realizes the transformation of unstructured hardware data into standardized environmental constraint parameters, providing a precise mathematical matrix foundation for subsequent optimization algorithms.

[0073] Example: The system substitutes the net yaw compensation torque generated by UAV A into the formula to calculate that the wind speed drag in the residential building area is extremely high, and the wind direction is westward. The system then fills the wind drag vector into the corresponding grid on the map, and uses Gaussian filtering to deduce that the adjacent blank grids around the block also belong to the high wind drag danger zone, and finally outputs the latest drag coefficient matrix covering the entire target community.

[0074] Step S3: Extract the real-time battery internal resistance temperature, remaining usable power, and instantaneous bus discharge current of each UAV from the real-time operating parameters; according to the formula... Calculate the initial spatiotemporal computing power scalar for each UAV. Then, perform global extremum normalization on the initial spatiotemporal computing power scalar across the entire multi-UAV collaborative network to output the normalized dynamic spatiotemporal computing power.

[0075] Explanation: This feature represents an innovative aspect of evaluating single-drone carrying capacity. During high-intensity flight, the drone's battery heats up, consuming a significant amount of energy. Dispatching orders solely based on apparent remaining battery power could easily lead to a crash mid-flight. In the formula, C represents the initial spatiotemporal computing power scalar; For the current moment The remaining available power; From the current moment up to the predicted time The nonlinear internal resistance heat loss is the integral of the electrical charge; I is the instantaneous bus discharge current; W is the maximum available load; and r is the mapped resistance value obtained by converting through a preset temperature-resistance conversion function. This principle not only deducts the energy decay caused by battery heating, but also multiplies it by the maximum available load to obtain a scalar that comprehensively reflects its absolute order-taking capacity.

[0076] Technical effects: It completely eliminates the trap of inflated power levels caused by heat loss, achieves accurate mapping of physical battery constraints on the mathematical model, and ensures scheduling safety.

[0077] Example: Although Drone B's display shows 60% battery remaining, its continuous ascent has resulted in a high real-time internal resistance temperature and significant heat loss. After applying the above formula, the system calculates that Drone B's actual initial spatiotemporal computing power has been greatly reduced, and the system determines that it is no longer suitable to handle the heavy emergency order for medical supplies.

[0078] Step S4: Set the dynamic spatiotemporal computing power as the central expansion capability index of the corresponding geographical node in the multi-machine collaborative network, and use it as the expansion denominator term for generating the dynamic weighted Thiessen polygon rule; taking the three-dimensional position of each UAV as the pole, according to the formula... Calculate the spatial expansion cost distance, compare the spatial expansion cost distance of adjacent poles, generate dynamic intersecting dividing lines at the intersection of spatial surfaces with equal costs, and perform continuous topological closure splicing to form the task handover spatial boundary.

[0079] Explanation: Following the previous steps, this feature is used to determine which drone in the network should accept a sudden order. Traditional order acceptance logic only considers straight-line distance, while this solution introduces a dynamically weighted Thiessen polygon algorithm. In the formula, D is the spatial expansion cost distance; d is the Euclidean physical measured distance; Here, C represents the value corresponding to the drag coefficient matrix; C is the dynamic spatiotemporal computing power. The principle is: the system expands outwards from the three-dimensional position of each drone's body; the stronger the computing power C, the lower the surrounding wind resistance. The smaller the value, the lower the cost D of its outward expansion, and the larger the area it can cover. When the costs of expanding adjacent poles are equal, a dynamic intersecting dividing line is formed.

[0080] Technical effect: It eliminates task conflicts in multi-machine networks when dealing with sudden orders from a mathematical and geometric perspective, and realizes flexible boundary reconstruction driven by computing power.

[0081] Example: Due to the decline in computing power of drone B, its coverage area shrinks drastically; while drone C, which is standing by nearby with ample power and in a windless area, rapidly expands its boundary outward according to the formula. Ultimately, the task handover space boundary formed by the expansion of drone C successfully encompasses the coordinates of the insertion target located in the target cell, and the system then determines that drone C will accept the order.

[0082] Step S5: Within the task handover space boundary, an improved ant colony algorithm is used for path optimization iteration, according to the formula... Generate a comprehensive cost heuristic function.

[0083] Explanation: After determining which drone will accept the order, this feature is used to plan the safest and most energy-efficient flight path for it. Traditional ant colony optimization algorithms only consider distance factors. The innovation of this approach lies in a deep modification of the algorithm's underlying structure. In the formula, The comprehensive cost heuristic function; d is the basic path length heuristic factor; represents the drag penalty parameter term in the drag coefficient matrix; C represents the dynamic spatiotemporal computing power. This is a penalty component for the generated power exponential decay. The preset fixed weight coefficients. This function forces the virtual ant colony to actively avoid grids with high wind resistance when exploring paths, and drones with lower computing power will be subject to a more stringent power exponential decay penalty component.

[0084] Technical effect: Significantly improves the quality of global solution in complex environments and effectively avoids planning fatally dangerous paths that are "shortest in distance but consume more power against the wind".

[0085] Example: The system begins planning a route for drone C to its destination. In matrix iteration operations, due to the existence of heuristic formulas... This wind resistance penalty means that the virtual ant colony senses strong winds in the canyon that drone A just explored in the high-rise residential area ahead. Therefore, the algorithm actively abandons the straight shortcut through the two high-rise buildings and instead plans a slightly detour for safety.

[0086] Furthermore, a negative feedback pheromone local rapid evaporation rule parameter is constructed based on the correlation between the actual loaded effective payload mass data and the modulus product of local micro-meteorological wind resistance characteristics. The spatial variable of the high-risk load resistance product is substituted into the pheromone update formula. .

[0087] Explanation: In connection with the previous steps, this feature optimizes the anti-local deadlock mechanism of the improved ant colony algorithm. It is important to note that the actual mounted effective payload mass data (M), which is used as an input parameter in the formula, is essentially the physical weighing mass obtained in technical feature 1 within the system data flow. In the formula, The updated pheromone values; This represents the base value of the pheromone from the original path. The modulus value represents the local micro-meteorological wind resistance characteristics; This is the preset environmental volatility coefficient. The principle is as follows: if a drone carrying a heavy load forcibly flies through a high-drag area, the risk of crash and energy consumption is amplified exponentially. At this time, the system will drastically reduce the attracting pheromones along the path, output a negative feedback pheromone local rapid volatilization rule parameter, and forcibly warn other heavily loaded drones in the network to avoid the area.

[0088] Technical effect: It effectively overcomes the mathematical defect of traditional heuristic algorithms that are prone to getting stuck in local optimum deadlock in multi-constraint logistics scheduling, and greatly improves the robustness of the overall logistics network.

[0089] Example: Suppose that a few minutes ago, another drone D, fully loaded with heavy cargo (i.e., with an extremely large actual payload), forcibly flew over a windy area at an intersection. Based on the formula mentioned above, the system instantly and drastically reduced the path pheromones in that area to zero. Therefore, when planning its route, drone C, carrying a medical supply box, completely avoided the high-risk pheromone emission area left by drone D.

[0090] Furthermore, the system outputs the optimal collaborative delivery matrix for sudden order insertion attributes and the corresponding dynamic obstacle avoidance trajectory for each drone, and sends the dynamic obstacle avoidance trajectory to the corresponding drone to instruct the flight control system of the corresponding drone to perform obstacle avoidance flight.

[0091] Explanation: This feature is the final physical execution point in the entire methodology. After completing all data collection, feature extraction, boundary reconstruction, and optimal path iterative calculation in the cloud, the final determined optimal control command must be transmitted to the physical entity. The cloud server, through a wireless communication network, directly sends the dynamic obstacle avoidance trajectory, including the coordinates of the three-dimensional heading control points, to the assigned UAV, taking over and guiding its onboard flight control system's deflector motors to execute the corresponding obstacle avoidance maneuvers.

[0092] Technical effect: It ensures that the logical calculation results of the software cloud can be truly and accurately converted into physical execution actions of the hardware, achieving a complete closed-loop control from virtual scheduling to physical logistics delivery.

[0093] Example: Ultimately, the cloud-based dispatch center sends the calculated safest and least energy-consuming dynamic obstacle avoidance trajectory to drone C. Upon receiving the trajectory instruction, drone C's onboard flight control system immediately adjusts the rotation speed of its four rotors, allowing it to smoothly climb along the planned safe passage and successfully deliver emergency medical supplies to the target community customer, perfectly handling the challenges of complex urban weather conditions and urgent delivery.

[0094] To fully illustrate the above embodiments, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0095] This embodiment provides a method for low-altitude logistics drone path optimization and multi-drone collaborative delivery scheduling, the execution entity of which is typically a logistics scheduling server deployed in the cloud. The specific execution logic of this embodiment is as follows:

[0096] (1) Obtain real-time operational parameters, including the three-dimensional position of the drones, in the multi-drone collaborative network, and receive the sudden order insertion attributes. Among them, the sudden order insertion attributes include the target coordinates of the insertion and the physical weighing mass. This process is the basic data input port of the entire scheduling system. In practical application scenarios, such as in a multi-drone logistics collaborative network, the cloud scheduling center tracks the three-dimensional positions of drones A, B, and C in the network in real time. When the system receives an emergency medical supply order from a nearby target community, the system immediately records the target coordinates of the order insertion and the physical weighing mass of the medical supply box. The physical weighing mass is the core pre-variable for subsequent calculation of drone energy consumption and flight resistance.

[0097] (2) Real-time monitoring of the six-degree-of-freedom physical flight attitude of the preceding UAV in the forward detection route in three-dimensional space, extracting the pulse width modulation duty cycle difference value of each rotor motor under the state of maintaining the horizontal balance of the body; filtering out the reference static torque generated by the self-weight of the body based on the pulse width modulation duty cycle difference value, and separating the net yaw compensation torque specifically used to resist the crosswind disturbance as the multi-rotor differential torque decoupling data. This mechanism reversely utilizes the flight control principle of multi-rotor differential control. When UAV A encounters a strong "canyon wind effect" while passing through two high-rise residential buildings, the system captures the difference in the rotational speed electrical signal between its rotors, accurately separates the net yaw compensation torque to counteract this crosswind, and makes the UAV itself act as a high-precision micro-meteorological detector.

[0098] (3) Input the multi-rotor differential torque decoupling data into the preset aerodynamic inverse transformation model for inverse calculation, and output the local micro-meteorological drag characteristics corresponding to the three-dimensional spatial position. Specifically, the system follows the formula The local wind speed drag scalar is obtained and transformed into a three-dimensional direction vector containing absolute direction attributes by combining it with the aircraft yaw angle attitude data. Subsequently, the local micro-meteorological wind drag characteristics are written into the corresponding spatial grid nodes of the three-dimensional environment blank matrix, and a Gaussian filtering algorithm is used to perform spatial interpolation smoothing on unsampled adjacent nodes to output the drag coefficient matrix.

[0099] (4) Combining the remaining available power and maximum available payload in the real-time operating parameters, calculate the dynamic spatiotemporal computing power of each UAV in the current flight cycle. The system extracts the real-time battery internal resistance temperature, remaining available power, and instantaneous bus discharge current of each UAV from the real-time operating parameters; according to the formula The initial spatiotemporal computing power scalar for each drone is calculated. This principle completely eliminates the "artificially high battery power" trap caused by excessive internal resistance and heat loss due to high battery temperature during high-load flight. The normalized output dynamic spatiotemporal computing power comprehensively reflects the drone's absolute order-taking capability.

[0100] (5) Based on the resistance coefficient matrix and dynamic spatiotemporal computing power, reconstruct the task handover spatial boundary of each delivery responsibility area in the multi-machine collaborative network. Please refer to [link to relevant documentation]. Figure 3 Taking UAVs A33, B34, and C37 as examples, the system sets the dynamic spatiotemporal computing power as the expanded denominator, and calculates the spatial expansion cost distance with the three-dimensional position of the aircraft as the pole. For example... Figure 3 As shown, drone B 34 experiences a severe shrinkage in its polygon coverage area due to decreased computing power; drone A 33, near the "local high wind resistance zone 32," shows its polygon edges concave inwards due to a sharp increase in wind resistance costs; while drone C 37, with ample battery power and located in a windless zone, exhibits rapid outward expansion of its polygon. When the cost of expansion at adjacent poles is equal, a situation is formed... Figure 3The bolded dynamic intersection dividing line 35. Ultimately, the expanded mission handover space boundary of UAV C 37 successfully encompassed the coordinates of the sudden insertion target 36, and the system determined that UAV C 37 would accept the mission.

[0101] (6) Within the task handover space boundary, an improved ant colony algorithm is used for path optimization iteration, outputting the optimal collaborative delivery matrix for sudden order insertion attributes and the corresponding dynamic obstacle avoidance trajectories for each UAV. Please refer to [reference needed]. Figure 4 In the drag coefficient matrix 41 of the three-dimensional spatiotemporal grid, this invention introduces a custom comprehensive cost heuristic function into the improved ant colony algorithm, according to the formula... Forcing the virtual ant colony to actively avoid Figure 4 A localized high-drag micro-meteorological zone 42, with a shading line in the middle. Simultaneously, a negative feedback pheromone local rapid evaporation rule parameter is constructed based on the product of the actual payload mass data and the local micro-meteorological drag characteristics. For example... Figure 4 As shown, if a fully loaded drone has previously flown over the area, the resulting high-risk load resistance product spatial variable will be substituted into the formula. A drastic reduction in pheromones. This causes... Figure 4 The “early exploration path 43” in the algorithm forms a rapid disconnection path 45 when encountering high resistance and extremely low pheromone (avoiding deadlock). Finally, the algorithm converges smoothly and outputs an optimal cooperative obstacle avoidance trajectory 44 that perfectly bypasses the high-risk area.

[0102] (7) The dynamic obstacle avoidance trajectory is sent to the corresponding UAV to instruct the flight control system of the corresponding UAV to perform obstacle avoidance flight. The cloud server sends the dynamic obstacle avoidance trajectory containing the coordinates of the three-dimensional heading control point directly to UAV C through the wireless communication network, takes over and guides the deflection motor of its airborne flight control system, and successfully delivers emergency medical supplies to the customer, achieving closed-loop control from virtual scheduling to physical logistics delivery.

[0103] Based on the above methodological logic, this embodiment also provides a dynamic path optimization and multi-drone collaborative delivery scheduling system for low-altitude logistics drones. Please refer to [link / reference]. Figure 2 The architecture of the various virtual software modules of this system is as follows:

[0104] The data acquisition module 100 acquires real-time operational parameters, including the three-dimensional position of the drone, of each drone in the multi-drone collaborative network, and receives the burst insertion attributes of burst insertions.

[0105] The environmental resistance construction module 200 acquires the local micro-meteorological wind resistance characteristics of the airspace where the multi-machine collaborative network is located, and refreshes the resistance coefficient matrix of the three-dimensional spatiotemporal grid based on the local micro-meteorological wind resistance characteristics.

[0106] The dynamic computing power assessment module 300 combines the remaining available power and maximum available payload in the real-time operating parameters to calculate the dynamic spatiotemporal computing power of each UAV in the current flight cycle.

[0107] The handover boundary reconstruction module 400 reconstructs the task handover spatial boundary of each delivery responsibility area of ​​the package in the multi-machine collaborative network based on the resistance coefficient matrix and dynamic spatiotemporal computing power.

[0108] The collaborative scheduling optimization module 500 uses an improved ant colony algorithm to iterate path optimization within the task handover space boundary, outputs the optimal collaborative delivery matrix for sudden order insertion attributes and the corresponding dynamic obstacle avoidance trajectory of each UAV, and sends the dynamic obstacle avoidance trajectory to the corresponding UAV to instruct the flight control system of the corresponding UAV to perform obstacle avoidance flight.

[0109] In addition, to execute the computational logic of the aforementioned system modules, embodiments of the present invention also provide an electronic device, which may be a data scheduling server deployed in a logistics cloud. The electronic device includes at least one processor and a memory communicatively connected to the processor.

[0110] The memory stores computer instructions that can be executed by the processor. The processor executes these computer instructions during runtime to implement the various logic and algorithm formulas for dynamic path optimization and multi-drone collaborative delivery scheduling of low-altitude logistics drones as described in the above method embodiments. Specifically, the processor can be a central processing unit (CPU), or it can be an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or a neural network processor (NPU) used to accelerate the matrix operations of the improved ant colony algorithm described above.

[0111] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for low-altitude logistics drone path optimization and multi-drone collaborative delivery scheduling, characterized in that, include: S1. Obtain the real-time operating parameters of each UAV in the multi-UAV collaborative network, including the three-dimensional position of the UAV body, and receive the sudden insertion attributes of sudden insertion, wherein the sudden insertion attributes include the insertion target coordinates and physical weighing mass. S2. Obtain the local micro-meteorological wind resistance characteristics of the airspace where the multi-machine collaborative network is located, and refresh the drag coefficient matrix of the three-dimensional spatiotemporal grid based on the local micro-meteorological wind resistance characteristics. S3. Calculate the dynamic spatiotemporal computing power of each UAV in the current flight cycle by combining the remaining available power and maximum available load in the real-time operation parameters. S4. Based on the resistance coefficient matrix and the dynamic spatiotemporal computing power, reconstruct the task handover space boundary of each delivery responsibility area of ​​the package in the multi-machine collaborative network; S5. Within the boundary of the task handover space, an improved ant colony algorithm is used to perform path optimization iteration, outputting the optimal collaborative delivery matrix for the sudden order insertion attribute and the corresponding dynamic obstacle avoidance trajectory of each UAV, and sending the dynamic obstacle avoidance trajectory to the corresponding UAV to instruct the flight control system of the corresponding UAV to perform obstacle avoidance flight.

2. The method for low-altitude logistics drone path optimization and multi-drone collaborative delivery scheduling according to claim 1, characterized in that, In step S2, the local micro-meteorological wind resistance characteristics of the airspace where the multi-machine cooperative network is located are obtained, including: S21. Real-time monitoring of the six-degree-of-freedom physical flight attitude of the preceding UAV in the forward detection route in three-dimensional space, and extracting the pulse width modulation duty cycle difference value of each rotor motor under the state of maintaining the horizontal balance of the body. S22. Based on the difference in pulse width modulation duty cycle, filter out the reference static torque generated by the weight of the airframe and separate the net yaw compensation torque specifically used to resist lateral environmental wind disturbance as multi-rotor differential torque decoupling data. S23. Input the multi-rotor differential torque decoupling data into a preset aerodynamic inverse transformation model for inverse calculation, and output the local micro-meteorological wind resistance characteristics corresponding to the three-dimensional spatial position.

3. The method for low-altitude logistics drone path optimization and multi-drone collaborative delivery scheduling according to claim 2, characterized in that, Step S23 specifically includes: S231, according to the formula Obtain the local wind speed drag scalar, where R is the local wind speed drag scalar. The net yaw compensation torque in the multi-rotor differential torque decoupling data, where S is the frontal area coefficient and K is a preset aerodynamic calibration constant parameter. air density; S232. The local wind speed drag scalar is combined with the aircraft yaw angle attitude data to convert it into a three-dimensional direction vector, generating the local micro-meteorological wind drag feature containing absolute direction attributes. S233. Write the local micro-meteorological wind resistance characteristics into the corresponding spatial grid nodes of the three-dimensional environmental blank matrix, and use the Gaussian filtering algorithm to perform spatial interpolation smoothing on the unsampled adjacent nodes to output the drag coefficient matrix.

4. The method for low-altitude logistics drone path optimization and multi-drone collaborative delivery scheduling according to claim 1, characterized in that, Step S3 specifically includes: S31. Extract the real-time battery internal resistance temperature, remaining available power, and instantaneous bus discharge current of each UAV from the real-time operating parameters. S32, According to the formula Calculate the initial spatiotemporal computing power scalar for each UAV, where C is the initial spatiotemporal computing power scalar. For the current moment The remaining available power, From the current moment up to the predicted time The nonlinear internal resistance heat loss electrical energy integral, I is the instantaneous bus discharge current, W is the maximum available load, and r is the mapped resistance value obtained by conversion through the preset temperature resistance conversion function; S33. Perform global extreme value normalization on the initial spatiotemporal computing power scalar throughout the multi-machine collaborative network, and output the normalized dynamic spatiotemporal computing power.

5. The method for low-altitude logistics drone path optimization and multi-drone collaborative delivery scheduling according to claim 1, characterized in that, Step S4 specifically includes: S41. Set the dynamic spatiotemporal computing power as the central expansion capability index of the corresponding geographical node in the multi-machine collaborative network, and use it as the expansion denominator term for generating dynamic weighted Thiessen polygon rules. S42. Taking the three-dimensional position of each UAV as the pole, according to the formula... Calculate the spatial expansion cost distance, where D is the spatial expansion cost distance and d is the Euclidean physical measured distance. C represents the value corresponding to the drag coefficient matrix, and C represents the dynamic spatiotemporal computing power. S43. Compare the spatial expansion cost distances of adjacent poles, generate dynamic intersecting dividing lines at the intersection of spatial surfaces with equal costs, and perform continuous topological closure splicing to form the task handover space boundary.

6. The method for low-altitude logistics drone path optimization and multi-drone collaborative delivery scheduling according to claim 1, characterized in that, In step S5, an improved ant colony algorithm is used for path optimization iteration within the task handover space boundary, including: S51. In the improved ant colony algorithm, a custom comprehensive cost heuristic function is introduced, which includes the wind resistance penalty parameter in the drag coefficient matrix and the power exponential decay penalty component. S52. Construct a negative feedback pheromone local rapid volatilization rule parameter based on the product of the actual mounted effective payload mass data and the local micro-meteorological wind resistance characteristics; S53. Based on the comprehensive cost heuristic function and the negative feedback pheromone local rapid evaporation rule parameter, the optimal collaborative delivery matrix is ​​generated by performing multi-generation optimization iterations on the physical delivery route within the task handover space boundary.

7. The method for low-altitude logistics drone path optimization and multi-drone collaborative delivery scheduling according to claim 6, characterized in that, In step S51, a custom comprehensive cost heuristic function is introduced into the improved ant colony algorithm, including: S511. Calculate the spatial Euclidean distance between the current 3D spatial node and the target adjacent raster node to be expanded, and generate the basic path length heuristic factor. S512. Extract the modulus of the local micro-meteorological wind resistance characteristics from the drag coefficient matrix as a wind resistance penalty parameter, and generate a power exponential decay penalty component based on the dynamic spatiotemporal computing power. S513, according to the formula Generate the comprehensive cost heuristic function, where Let be the comprehensive cost heuristic function, and d be the basic path length heuristic factor. C represents the wind resistance penalty parameter, and C represents the dynamic spatiotemporal computing power. These are preset fixed weighting coefficients.

8. The method for low-altitude logistics drone path optimization and multi-drone collaborative delivery scheduling according to claim 6, characterized in that, Step S52 specifically includes: S521. After a single blind path search iteration, extract the corresponding actual mounted effective payload mass data and the modulus of the local micro-meteorological wind resistance characteristics corresponding to the path three-dimensional grid. S522. Multiply the actual mounted effective load mass data with the modulus of the local micro-meteorological wind resistance characteristics by tensor values ​​to obtain the high-risk load resistance product spatial variable. S523, Substitute the high-risk load resistance product space variable into the pheromone update formula. ,in The updated pheromone values. M represents the base value of the pheromone from the original path, and M represents the actual mass of the mounted payload. This refers to the modulus value of the local micro-meteorological wind resistance characteristics. This is the preset environmental volatility coefficient.

9. A system applied to the low-altitude logistics drone path optimization and multi-drone collaborative delivery scheduling method according to any one of claims 1 to 8, characterized in that, The system includes: The system operates a data acquisition module, which acquires real-time operating parameters of each UAV in the multi-UAV collaborative network, including the three-dimensional position of the UAV body, and receives the sudden insertion attributes of sudden insertions, wherein the sudden insertion attributes include the target coordinates of the insertion and the physical weighing mass. An environmental resistance construction module acquires the local micro-meteorological wind resistance characteristics of the airspace where the multi-machine collaborative network is located, and refreshes the resistance coefficient matrix of the three-dimensional spatiotemporal grid based on the local micro-meteorological wind resistance characteristics. The dynamic computing power evaluation module combines the remaining available power and maximum available load in the real-time operating parameters to calculate the dynamic spatiotemporal computing power of each UAV in the current flight cycle. The handover boundary reconstruction module reconstructs the task handover spatial boundary of each delivery responsibility area of ​​the package in the multi-machine collaborative network based on the resistance coefficient matrix and the dynamic spatiotemporal computing power. The collaborative scheduling optimization module uses an improved ant colony algorithm to iterate path optimization within the task handover space boundary, outputs the optimal collaborative delivery matrix for the sudden order insertion attribute and the corresponding dynamic obstacle avoidance trajectory of each UAV, and sends the dynamic obstacle avoidance trajectory to the corresponding UAV to instruct the flight control system of the corresponding UAV to perform obstacle avoidance flight.

10. The low-altitude logistics drone path optimization and multi-drone collaborative delivery scheduling system according to claim 9, characterized in that, The operational data acquisition module includes: The message unpacking unit receives the sudden order logistics data stream message and extracts the target coordinates of the insertion target and the extreme delivery time threshold from the sudden order logistics data stream message. A mass sensing unit acquires the physical weighing sensing electrical signal of the package, and processes the physical weighing sensing electrical signal through analog-to-digital conversion to generate a digital quantity as the physical weighing mass. The attribute assembly unit splices the target coordinates of the inserted order, the extreme delivery time threshold, and the physical weighing mass into a low-level data packet to generate a structured burst order attribute, which is then sent to the collaborative scheduling optimization module.

Images