Automatic path planning method and device based on multi-constraint dynamic optimization

Abstract

The invention provides an automatic path planning method and device based on multi-constraint dynamic optimization, and relates to the technical field of unmanned aerial vehicle automatic path planning. According to the method, environment information is collected in real time, multi-dimensional constraints such as electric quantity, speed, obstacles and task priority are converted into mathematical conditions, and multi-constraint collaborative analysis is achieved; the dynamic environment interaction module is used for detecting environment change in real time and triggering local rapid re-planning, so that the response time is remarkably shortened; performing weighted comprehensive scoring of length, energy consumption, stability and task time consumption on each candidate path through a multi-objective optimization module to realize balanced optimization; and meanwhile, multi-task priority ordering, path connection optimization, task point regionalization clustering and multi-machine collaborative allocation are supported, so that the operation efficiency and safety in a complex scene are effectively improved. The method has the advantages of being high in multi-constraint adaptability, fast in dynamic response, balanced in optimization target, efficient in multi-task planning and the like, and is suitable for various practical application scenes such as unmanned aerial vehicle inspection, rescue and logistics delivery.

Description

Technical Field

[0001] This invention relates to the field of automatic path planning technology for unmanned aerial vehicles (UAVs), and specifically to an automatic path planning method and apparatus based on multi-constraint dynamic optimization. Background Technology

[0002] With the rapid development of drone technology, automatic path planning has become a key supporting technology for realizing autonomous drone flight. In complex application scenarios such as power line inspection, disaster relief, logistics transportation, and agricultural monitoring, drones need to complete flight missions efficiently and safely in complex and dynamic environments. However, current mainstream path planning algorithms still have many limitations and cannot fully meet the above-mentioned practical operational requirements.

[0003] First, traditional algorithms often consider only a single constraint, such as planning based solely on the shortest path or static obstacle avoidance, while ignoring the synergistic effects of multi-dimensional constraints. In actual flight, drones not only need to avoid obstacles but also consider factors such as remaining battery power, flight speed limits, mission time windows, and environmental weather conditions. If obstacle avoidance is the sole objective, a path may be planned that, while avoiding collision risk, consumes excessive energy, leading to premature battery depletion and mission failure. If only the shortest path is considered, the inclusion of numerous sharp turns or climbs may reduce flight stability, even exceeding the safety thresholds of the drone's structure or control system, affecting equipment lifespan and operational accuracy.

[0004] Secondly, existing algorithms are insufficient in their ability to respond to dynamically changing environments. Most algorithms rely on pre-acquired static environmental data for offline planning, lacking the ability to perceive and quickly replan dynamic obstacles (such as other aircraft temporarily entering the flight area, temporary facilities, sudden airflow, etc.). When the environment changes, the algorithm usually needs to recalculate the global path, which is time-consuming and causes the drone to be unable to adjust its course in time, resulting in a high risk of collision or loss of control.

[0005] Furthermore, existing technologies suffer from imbalances in multi-objective optimization. Some algorithms excessively pursue the optimization of a single metric, such as the shortest path length, while neglecting other key performance indicators such as flight energy consumption, mission duration, trajectory smoothness, and overall stability. This single-objective optimization may result in a planned path that is not applicable in practice. For example, although the distance may be the shortest, frequent maneuvers (such as numerous sharp turns, climbs / descents) can lead to a surge in energy consumption and drastic changes in flight attitude, which not only shortens the endurance but may also affect the performance of airborne sensors (such as detection sensors) or actuators (such as aerial cameras).

[0006] Furthermore, existing algorithms are poorly adapted to multi-task scenarios. When a drone needs to visit multiple target task points sequentially (such as inspecting multiple designated locations or dropping materials at specific points), traditional methods often struggle to reasonably coordinate the task execution order and flight path connections, easily leading to problems such as redundant flight paths, inefficient task sequences, and excessively long overall operation times, failing to meet the practical engineering requirements for efficiency and orderliness.

[0007] In view of the above, this application is hereby submitted. Summary of the Invention

[0008] This invention provides an automatic path planning method and apparatus based on multi-constraint dynamic optimization, which can at least partially improve the above-mentioned problems.

[0009] To achieve the above objectives, the present invention adopts the following technical solution: An automatic path planning method based on multi-constraint dynamic optimization includes: The system acquires multi-dimensional information collected by a preset environmental perception module, and transforms the multi-dimensional information to obtain a set of mathematical constraints. The task constraints of the UAV are determined based on the mathematical constraint set, and the energy consumption data of each candidate path is calculated using the FFEM module based on the preset UAV parameters, UAV flight status and task constraints. The Dueling module is called to perform weight allocation based on the task scenario and energy consumption data of each candidate path. The allocated candidate paths are then comprehensively scored, and the path with the highest comprehensive score is selected as the optimal path, while the remaining paths are selected as candidate paths. When the drone flies along the optimal path, the multi-dimensional information collected at the moment is compared with the multi-dimensional information collected at the previous moment. When an environmental change is detected, the INEM module is used to update the mathematical constraint group to achieve dynamic obstacle avoidance.

[0010] The present invention also provides an automatic path planning device based on multi-constraint dynamic optimization, comprising: The conversion unit is used to acquire multi-dimensional information collected by the preset environmental perception module, and to convert the multi-dimensional information to obtain a set of mathematical constraints. The energy consumption calculation unit is used to determine the mission constraints of the UAV based on the mathematical constraint set, and to calculate the energy consumption data of each candidate path based on the preset UAV parameters, UAV flight status and mission constraints using the FFEM module. The path planning unit calls the Dueling module to perform weight allocation processing based on the task scenario and the energy consumption data of each candidate path, performs a comprehensive score on each candidate path after allocation, selects the path with the highest comprehensive score as the optimal path, and selects the remaining paths as candidate paths. The dynamic obstacle avoidance unit is used to compare the multi-dimensional information collected at the current moment with the multi-dimensional information collected at the previous moment when the UAV is flying according to the optimal path. When it is determined that there is an environmental change, the INEM module is used to update the mathematical constraint group to achieve dynamic obstacle avoidance.

[0011] In summary, the present invention constructs an overall architecture comprising an environment perception layer, a constraint resolution layer, a dynamic optimization layer, and a path output layer. Through the collaborative work of the flight energy consumption modeling module, the dynamic environment interaction module, and the multi-objective optimization module, it achieves highly adaptable planning for complex and dynamic flight environments. Specifically, the present invention transforms real-time information such as battery power, speed, obstacles, and task priority into computable mathematical constraints through the constraint resolution layer, overcoming the problem of single constraints in traditional algorithms and ensuring path safety and task reachability. The dynamic environment interaction module can perceive environmental changes in real time and trigger rapid local path replanning when dynamic obstacles are detected, greatly shortening response time and enhancing the flight reliability of UAVs in dynamic scenarios. The multi-objective optimization module introduces a configurable weight allocation mechanism to comprehensively score and select the best among path length, flight energy consumption, attitude stability, and task time, achieving balanced optimization among multiple objectives and taking into account both the economy and practicality of the path. Furthermore, for multi-task scenarios, the present invention proposes a series of strategies such as task priority ranking, path connection optimization, task point region clustering, and multi-UAV collaborative scheduling, significantly improving the overall efficiency of complex operations.

[0012] This invention effectively integrates multi-dimensional constraint analysis, rapid response to dynamic environments, and multi-objective collaborative optimization capabilities, providing a safe, efficient, and reliable automatic path planning solution for UAVs in various application scenarios such as inspection, rescue, and logistics, significantly improving the autonomous operation level and mission completion quality of UAV systems. Attached Figure Description

[0013] Figure 1 This is a flowchart illustrating the automatic path planning method based on multi-constraint dynamic optimization provided in the first embodiment of the present invention.

[0014] Figure 2 This is a schematic diagram of the framework of the automatic path planning method based on multi-constraint dynamic optimization provided in the first embodiment of the present invention.

[0015] Figure 3 This is a structural diagram of the FFEM module provided in an embodiment of the present invention.

[0016] Figure 4 This is a structural diagram of the Dueling module provided in an embodiment of the present invention.

[0017] Figure 5 This is a structural diagram of the INEM module provided in an embodiment of the present invention.

[0018] Figure 6 This is a schematic diagram of the automatic path planning device based on multi-constraint dynamic optimization provided in the second embodiment of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0020] refer to Figure 1 As shown, the first embodiment of the present invention discloses an automatic path planning method based on multi-constraint dynamic optimization. The overall framework of this method is as follows: Figure 2 As shown, the structure is divided into four layers from top to bottom: environment perception layer, constraint resolution layer, dynamic optimization layer, and path output layer. It can be executed by an automatic path planning device based on multi-constraint dynamic optimization (hereinafter referred to as the planning device), specifically by one or more processors within the planning device, to achieve the following method: S1, acquire multi-dimensional information collected by the preset environmental perception module, and transform the multi-dimensional information to obtain a set of mathematical constraints; The environmental perception module includes a map module, a GPS positioning module, and a meteorological sensor module connected to a meteorological sensor interface; the multi-dimensional information includes the current latitude, longitude and altitude data of the UAV, real-time meteorological data with wind speed ≤ preset wind speed value (e.g., 5m / s), and map module data.

[0021] The mathematical constraint group includes power constraints, speed constraints, obstacle constraints, and task constraints. Specifically, based on a preset task attribute-priority mapping rule, the received multi-task requirements are transformed to obtain task constraints. The power constraint is the maximum consuming energy (e.g., 200Wh) corresponding to a total planned path energy consumption ≤ remaining power (e.g., 50%) calculated based on the current power level and energy consumption model. The speed constraint is the flight speed ≤ a reserved wind resistance redundancy value (e.g., 6m / s) determined by combining a preset safe speed range for the drone (e.g., 0-8m / s) and wind speed (e.g., 5m / s). The obstacle constraint is the constraint that the distance between the drone's flight trajectory and the obstacle boundary is ≥ a preset safe distance (e.g., 1.5 meters), converted from static obstacle coordinates.

[0022] Specifically, in this embodiment, a preset environmental perception module is first invoked. This module integrates a high-precision GPS positioning module, used to acquire the latitude, longitude, and altitude data of the UAV in real time; a meteorological sensor module connected to an airborne or ground meteorological station interface, responsible for collecting real-time meteorological data of the flight airspace. For example, the system is set to consider only wind speeds ≤ 5 m / s as safe weather conditions for planning, and data exceeding this threshold will trigger an alarm or wait for instructions; and a map module containing static geographic information such as terrain and buildings.

[0023] After acquiring multi-dimensional information, the process moves to the constraint resolution phase. The core task is to transform this raw, physical information into a series of mathematical constraints that the algorithm can directly process; that is, to construct a set of mathematical constraints, which forms the basis for all subsequent optimization decisions. Specifically: For task constraints, this is the core of handling complex operational requirements. When receiving a multi-task instruction such as "inspect three designated locations A, B, and C in sequence, or inspect point A for material delivery + inspect points B and C", the constraint parsing layer will transform it according to the preset task attribute-priority mapping rule. This rule first analyzes various attributes of the task, such as urgency (the urgency of rescue material delivery is higher than that of regular inspection in rescue scenarios), inter-task dependencies (inspection of point B must be started after data upload at point A is completed, and inspection of point C must be carried out after data collection at point B), and time window restrictions (point C must be completed before sunset, and inspection of point B must be completed within 1 hour), etc. Then, based on the preset weight model (urgency weight 0.5, time window weight 0.3, execution dependency weight 0.2), the priority score is calculated, and finally the execution order of material delivery at point A (score 9.2) > inspection of point B (score 7.5) > inspection of point C (score 6.8) is determined, and transformed into a hard constraint that the path must pass through point A first, and the inspection of point B can only be started after the task at point A is completed. This translates into constraints such as: the path must pass through specified coordinates in the order A→B→C, and the dwell time at each inspection point must be ≥30 seconds. This process transforms vague task requirements into precise, calculable path nodes and time sequence requirements, solving the problem of chaotic execution order and low efficiency in multi-task scenarios, and is a prerequisite for achieving efficient task planning.

[0024] For the power constraint, the current remaining power of the drone is first read (e.g., battery capacity is 50%). Combined with the built-in flight energy consumption model (which considers factors such as fuselage weight, aerodynamic characteristics, and motor efficiency), the maximum expendable energy consumption (e.g., 200Wh) corresponding to a total planned path energy consumption ≤ remaining power is calculated. The resulting power constraint is: the total estimated energy consumption of the planned path ≤ 200Wh. This constraint fundamentally avoids the risk of mission interruption or crash due to power depletion, significantly improving the reliability of mission completion.

[0025] For speed constraints, this method integrates two aspects: first, the safe flight speed range defined by the UAV's own performance (e.g., 0-8 m / s); and second, wind speed data (e.g., 5 m / s). To ensure flight stability and wind resistance, a certain safety margin is reserved before actual use. The calculation logic is as follows: for example, in headwind conditions, the UAV's ground speed needs to be adjusted to account for wind speed; the final speed constraint generated by the system may be a flight speed ≤ 6 m / s. This ensures that the planned path not only meets the aircraft's performance limits but also effectively copes with airflow disturbances, enhancing flight safety.

[0026] For obstacle constraints, the static obstacle database in the map module is invoked to obtain the precise coordinates and boundary information of all known obstacles. Each obstacle is then transformed into one or more spatial geometric constraints. A typical expression is: the Euclidean distance from any point on the UAV's flight path to the outer boundary of the obstacle (X) is ≥ 1.5 meters. Here, 1.5 meters is a preset safe distance threshold. By transforming physical obstacles into strict mathematical inequalities, explicit spatial avoidance rules are provided for the path planning algorithm, which is a key guarantee for achieving collision-free flight.

[0027] Through the above steps, heterogeneous information from different sensors and mission commands can be successfully integrated into a well-defined and interconnected set of mathematical constraints. This set of constraints comprehensively describes all the physical limitations and operational requirements faced by the current flight mission, providing comprehensive and accurate input conditions for the subsequent dynamic optimization layer. This ensures that the planned path is not only spatially feasible but also feasible in terms of energy, time, and mission logic, thereby comprehensively improving the intelligence level and mission success rate of UAV autonomous operations.

[0028] Please see Figure 3 This diagram illustrates the internal logic structure of the Flight Energy Modeling (FFEM) module. The left side represents the input unit, the right side the output unit, and the middle the core calculation module. S2: Based on the mathematical constraint set, the UAV's mission constraints are determined. Then, based on preset UAV parameters, UAV flight status, and mission constraints, the FFEM module calculates the energy consumption data for each candidate path. The drone's flight status includes level flight altitude, speed, and attitude angle, while the drone's parameters include fuselage weight and motor power.

[0029] In this embodiment, the FFEM module receives various input data to perform its computational functions. Specifically, UAV parameters include immutable airframe characteristics, such as fuselage weight (e.g., 2kg) and motor rated power (e.g., 500W), which are the fundamental basis for estimating basic power consumption and power response. Flight status includes calculating the UAV's flight profile for each candidate path by combining task constraints and environmental perception data. This includes level flight altitude (e.g., maintained at 100 meters throughout), flight speed (e.g., 5m / s), and attitude angle changes that may occur to avoid obstacles or perform tasks (e.g., roll angle during turns, pitch angle during climbs). Flight status is a dynamic variable that affects energy consumption.

[0030] Based on the above inputs, the FFEM module uses its built-in physical and empirical models to perform energy consumption simulation calculations for each candidate path that meets the spatial and logical requirements of the mathematical constraint set. For example, for a candidate path, the module analyzes it segment by segment. For instance, the energy consumption of path 1 (including 2 sharp turns and 1 10-meter climb) is 80Wh, and the energy consumption of path 2 (smooth flight, no sharp turns / sharp climbs) is 65Wh. The energy consumption data is then output to the Dueling module.

[0031] Please see Figure 4 This diagram illustrates the optimization logic of the multi-objective dueling module, with multi-objective weight allocation as its core. The left side shows the optimization objective inputs (path length, energy consumption, task completion time, stability), and the right side shows the optimization result outputs (optimal path score and candidate path ranking). In step S3, the Dueling module is invoked to perform weight allocation based on the task scenario and the energy consumption data of each candidate path. A comprehensive score is then calculated for each candidate path, and the path with the highest comprehensive score is selected as the optimal path, while the remaining paths are considered candidate paths. In this embodiment, the method calls the Dueling module to automatically load or dynamically calculate a set of optimized weights from a preset weight configuration library based on the current mission scenario type. The weight allocation reflects the core requirements under different scenarios. After the weights are determined, the Dueling module performs multi-dimensional evaluation and comprehensive scoring on each candidate path. Through the above calculations, the Dueling module sorts all candidate paths from high to low according to their comprehensive scores. The path with the highest comprehensive score is selected as the optimal path in the current scenario and will be output to the flight control system for execution. At the same time, the module retains several candidate paths with scores immediately following the optimal path, along with their complete constraints and status information. These alternative paths are not simply discarded but are cached as quick backup solutions in dynamic environments.

[0032] Specifically, the Dueling module assigns weights based on the task scenario: for example, if the current task is a long-range inspection mission, the weights are set as follows: energy consumption 0.4, path length 0.2, stability 0.3, and time consumption 0.1. Then, the candidate paths are comprehensively scored. Path 1 scores 8 points (out of 10) for length, 5 points for energy consumption, 4 points for stability, and 7 points for time consumption, with a total score of 8 × 0.2 + 5 × 0.4 + 4 × 0.3 + 7 × 0.1 = 5.7 points. Path 2 scores 7 points for length, 9 points for energy consumption, 9 points for stability, and 6 points for time consumption, with a total score of 7 × 0.2 + 9 × 0.4 + 9 × 0.3 + 6 × 0.1 = 8.2 points. At this point, the Dueling module outputs the highest-scoring path 2 as the optimal path, while retaining path 1 and another candidate path (scoring 7.5 points) to ensure rapid switching in case of sudden environmental changes.

[0033] In this embodiment, the dynamic optimization layer combines the energy consumption data from the FFEM module with the weight allocation from the Dueling module to complete the full-link planning of "takeoff point → mission point → return point". Taking "takeoff point → A → B → C → return point" as an example, the FFEM module first calculates the basic energy consumption of each segment: takeoff point to point A (straight-line distance 5km, strong wind area accounts for 20%, energy consumption 120Wh), point A to point B (straight-line distance 3km, no strong wind area, energy consumption 65Wh), point B to point C (straight-line distance 4km, flat terrain, energy consumption 80Wh), point C to return point (straight-line distance 6km, tailwind section accounts for 30%, energy consumption 90Wh); then it optimizes the connection nodes. For example, after the material is dropped at point A, the drone does not need to return to the initial 150-meter altitude, but flies directly to point B at an altitude of 120 meters, reducing climb / descent actions and reducing connection energy consumption by 15%. At the same time, it avoids the dense building area between B and C, controlling the path curvature within 5% and avoiding invalid flight mileage.

[0034] Please see Figure 5 This diagram illustrates the workflow of the Dynamic Environment Interaction (INEM) module, which includes four sub-units: real-time data reception, environmental change recognition, constraint update, and trigger signal output. These units are connected via a data bus. In step S4, while the UAV is flying along the optimal path, the currently acquired multi-dimensional information is compared with the information acquired in the previous moment. When an environmental change is detected, the INEM module updates the mathematical constraint set to achieve dynamic obstacle avoidance.

[0035] Specifically, step S4 further includes: when the UAV is flying according to the optimal path, calculating the change based on the multi-dimensional information currently collected and the multi-dimensional information collected at the previous moment, and determining whether the change is greater than a preset value. When the change is found to be greater than the preset value, it is determined that the current environment has changed. Based on the change, the obstacle constraints in the mathematical constraint group are adjusted and the coordinate range of the obstacle constraints is updated. Based on the updated mathematical constraint group, the FFEM module and Dueling module are called in sequence to perform calculations and replan a new path to achieve dynamic obstacle avoidance.

[0036] In this embodiment, the operating environment perception module collects multi-dimensional information such as GPS positioning, weather sensor data, and map updates at a fixed frequency (e.g., 3 times per second). It continuously compares and calculates the multi-dimensional information collected at the current moment with the multi-dimensional information collected and stored at the previous moment to quantitatively determine whether the environment has changed in a way that is sufficient to affect flight safety. Specifically, the module calculates the amount of change in specific information, such as the coordinate movement distance of dynamic obstacles (e.g., other drones, birds), instantaneous changes in wind speed and direction, or coordinate updates of temporarily appearing no-fly zones. Preset trigger thresholds for various changes; for example, when the detected coordinate change of a moving target is greater than 0.5 meters (preset value), it is determined that a meaningful dynamic environmental change has occurred. Once the INEM module determines that the change exceeds the preset threshold, it immediately initiates a constraint update process, performing efficient local processing on the change point. For example, when a new dynamic drone is detected intruding into the flight corridor, the INEM module predicts its trajectory range within a future period (e.g., the next 5 seconds) based on the intruding target's current speed and heading. Subsequently, the module updates the "obstacle constraints" in the mathematical constraint group in real time: based on the original set of static obstacle constraints, a new constraint condition is added regarding the predicted trajectory range of the dynamic obstacle, such as "the flight trajectory of the aircraft must maintain a distance of at least 2 meters from the predicted position area of ​​the dynamic target D". At the same time, if the wind speed suddenly increases and exceeds the safety redundancy value, the INEM module will also update the "speed constraints" to further limit the flight speed and ensure stability.

[0037] After the constraint group is updated, the INEM module immediately sends a "fast replanning trigger signal" to the dynamic optimization layer. This signal carries the updated constraint information and the location of the change. Upon receiving this signal, the dynamic optimization layer does not perform a time-consuming global path recalculation. Instead, it fully utilizes cached high-scoring candidate paths and, in conjunction with new local constraints, performs rapid local path adjustments around the changed area. For example, the system might select a path from the candidate path library that already slightly bypasses the area as a new base, or it might only smoothly adjust the affected segment of the current optimal path to avoid the predicted collision area. Because it eliminates the need for repeated evaluation of the unchanged environment and a large amount of redundant computation, this local replanning can be completed in a very short time (e.g., within 0.3 seconds).

[0038] Specifically, the INEM module receives environmental perception layer data in real time. For example, when another drone (dynamic obstacle) is detected entering the flight area and its position change is greater than 0.5 meters (triggering the environmental change threshold), the INEM module immediately initiates constraint update. The constraint update unit updates the coordinate range of the obstacle constraint to the real-time motion trajectory prediction range of the dynamic drone (based on its current speed and direction, predicting the position interval within the next 5 seconds). At the same time, the INEM module sends a "fast replanning trigger signal" to the dynamic optimization layer. The dynamic optimization layer skips some redundant calculation steps (such as simplifying historical path backtracking) and only performs local path adjustment for the changed area, reducing the replanning time to less than 0.3 seconds, ensuring that the drone avoids the dynamic obstacle in time.

[0039] Preferably, for scenarios where multiple task points are distributed in a dispersed manner, a task clustering sub-unit is also included; wherein, the K-means algorithm is used to divide task points with similar geographical locations into the same region, and priority is given to planning paths within the same region before planning connecting paths between regions; at the same time, combined with the collected regional meteorological characteristics, the execution time of regional tasks is adjusted, and high-energy-consuming cross-regional flights are scheduled during tailwind periods.

[0040] Specifically, in this embodiment, for scenarios where multiple task points are scattered (such as cross-administrative region inspections), a task clustering sub-unit is added to the constraint parsing layer. The K-means algorithm is used to divide task points with similar geographical locations (distance ≤ 3km) into the same region. For example, the four inspection points D, E, F, and G (D is 2.5km away from E, F is 2.8km away from G, and D is 8km away from F) are clustered into the DE region and the FG region. The dynamic optimization layer prioritizes planning the path within the same region (D→E), and then plans the connecting path between regions (E→F) to avoid repeated cross-regional flights. At the same time, combined with regional meteorological characteristics (such as wind speed ≤ 3m / s in the morning in the DE region and wind speed ≤ 3m / s in the afternoon in the FG region), the execution time of regional tasks is automatically adjusted, and high-energy-consuming cross-regional flights are arranged during tailwind periods, further reducing the total energy consumption by 18%-22%.

[0041] Through a closed-loop mechanism of real-time monitoring, quantitative judgment, constraint-focused updates, and triggering local replanning, this invention endows UAVs with real-time situational awareness and rapid response capabilities. It fundamentally overcomes the response lag defects caused by the reliance on static environmental data and the long replanning time of traditional algorithms. In practical applications, this means that when a sudden dynamic obstacle appears, the UAV can autonomously and smoothly adjust its course in less than a second to avoid collision risks. Tests have shown that it can reduce collision hazards in such scenarios. Simultaneously, the local optimization strategy greatly saves onboard computing resources, ensuring the high speed and stability of the system response. This enables the UAV to operate safely and reliably in dynamic environments with multi-aircraft collaboration and complex airspace traffic, significantly expanding the applicable boundaries and mission reliability of the automatic flight system.

[0042] Preferably, for scenarios where the workload of multiple tasks exceeds the load of a single drone, a collaborative task allocation subunit is also included; wherein, based on the remaining battery power, load capacity, and current location of each drone, a Hungarian algorithm is used to allocate tasks, and material delivery points that are less than a preset range away from each drone are assigned to the corresponding drones; at the same time, the flight altitudes of each drone are staggered.

[0043] When a drone's battery level drops below a preset value during a mission, the task is reassigned using a Hungarian algorithm based on the remaining battery level, load capacity, and current location of each drone. The unfinished task is then transferred to another drone with sufficient battery power.

[0044] Specifically, in this embodiment, when faced with scenarios where the workload of multiple tasks exceeds the load of a single drone (e.g., 10 material delivery points + 8 inspection points), the dynamic optimization layer adds a collaborative task allocation subunit. Based on the remaining battery power, load capacity (e.g., drone 1 can carry 3 sets of materials, drone 2 can carry 2 sets of materials), and current location of each drone, tasks are allocated using the Hungarian algorithm: material delivery points A, B, and C, which are closer to drone 1, are assigned to drone 1; material delivery points D and E, plus inspection points F and G, which are closer to drone 2, are assigned to drone 2. Simultaneously, a "non-interference flight corridor" is planned (e.g., drone 1 flies at an altitude of 120-150 meters, drone 2 flies at an altitude of 80-110 meters) to avoid collision risks. If drone 1 has less than 30% battery remaining after completing delivery at point A, the subunit can transfer the unfinished task at point C to drone 2, which has sufficient battery power (60% remaining), ensuring uninterrupted task completion.

[0045] By introducing a collaborative task allocation and dynamic scheduling mechanism, this invention achieves a leap from single-machine intelligence to swarm intelligence. It not only significantly improves the overall efficiency (reducing total flight mileage) and inherent safety of multi-machine operations through initial optimized allocation and hierarchical airspace management; more importantly, its dynamic task handover function based on real-time status constitutes a powerful system resilience. When a single drone experiences insufficient power, sudden malfunction, or encounters insurmountable obstacles, the system can self-repair and reorganize tasks like an organic whole, ensuring that the overall mission is not interrupted by a single point of failure, achieving high robustness and high completion rate in mission execution. This enables UAV swarms to be reliably applied to large-scale, complex, and continuous operation-required emergency response or infrastructure inspection scenarios, elevating the practical value of automated systems to a new level.

[0046] Preferably, in response to sudden changes during the task, a dynamic task adjustment interface is also included. This interface is configured to: when a temporary obstacle is detected within a preset distance, a task execution anomaly signal is fed back to facilitate constraint updates.

[0047] Specifically, in this embodiment, to cope with unexpected changes during the mission (such as the inability to land temporarily at point B), the constraint resolution layer reserves a dynamic mission adjustment interface: when the UAV detects a temporary obstacle 1 kilometer before reaching point B, it immediately reports a mission execution anomaly; the constraint resolution layer quickly recalculates priorities, and if the inspection time window at point C is tight (30 minutes remaining), the mission order is adjusted to A→C→B, and the path constraints are updated synchronously (such as shortening the dwell time at point C from 20 seconds to 15 seconds). The dynamic optimization layer completes path replanning within 0.5 seconds based on the new constraints, ensuring that the total time is controllable. Meanwhile, the constraint resolution layer supports user-defined optimization target interfaces, allowing users to adjust weights according to the scenario: in emergency rescue scenarios, the mission time weight can be increased to 0.6 and the energy consumption weight reduced to 0.1, prioritizing the shortest time path; in long-range inspection scenarios, the energy consumption weight can be increased to 0.5 and the stability weight increased to 0.3, prioritizing gentler routes; the interface also supports saving scenario templates for emergency rescue, regular inspection, etc., for easy subsequent direct use, improving operational efficiency.

[0048] Through a standardized and interface-based dynamic task adjustment mechanism, this invention endows the automatic path planning system with ad-hoc decision-making capabilities. It transforms the drone from a rigid machine rigidly executing a preset route into an intelligent agent capable of sensing unexpected events, assessing their impact, and autonomously adjusting its strategy with minimal human intervention. This flexible optimization capability significantly enhances the system's practicality and reliability in real, open, and unstructured environments. When faced with emergencies, the system can proactively avoid risks and reorganize tasks, preventing the entire task chain from being interrupted or the drone from being trapped in dangerous waiting due to temporary blockages at a single point. This significantly enhances the success rate and robustness of complex task execution, making it particularly suitable for highly uncertain application scenarios such as urban emergency response and wilderness search and rescue.

[0049] In summary, this invention provides an automatic path planning method based on multi-constraint dynamic optimization, aiming to address the shortcomings of existing technologies in multi-constraint adaptation, dynamic environment response, multi-objective optimization, path repetition, and excessive time consumption. Through a four-layer architecture and three core modules, it achieves real-time environmental perception, accurate analysis of multiple constraints, and rapid optimization in dynamic scenarios. It can adapt to various practical scenarios such as complex obstacles, multi-tasking, and dynamic interference, meeting the efficient and safe flight requirements of UAVs in tasks such as inspection, rescue, and material delivery. It is applicable to various types of UAVs, including multi-rotor and fixed-wing UAVs.

[0050] Specifically, this method first collects multi-dimensional information in real time through an environmental perception module, and transforms multiple constraints such as power consumption, speed, obstacles, and multi-task priorities into a computable set of mathematical constraints. This overcomes the problems of traditional algorithms having single constraints and poor adaptability, laying a foundational framework for path planning that balances safety and mission requirements. Building on this, the system introduces a flight energy consumption modeling module, which performs refined energy consumption predictions for candidate paths based on specific UAV parameters and flight status, achieving a quantitative assessment of path economics and providing key input for multi-objective optimization. Through the multi-objective optimization module, the system dynamically allocates weights according to different mission scenarios, comprehensively scoring and selecting the best path length, energy consumption, stability, and time consumption. This represents a leap from single-index optimization to multi-objective balanced decision-making, effectively solving the problem of insufficient practicality caused by imbalanced optimization objectives. It can significantly reduce energy consumption and extend endurance in long-range inspections, or improve flight stability and equipment accuracy in precision operations.

[0051] Most importantly, this invention achieves real-time monitoring and rapid response to sudden environmental changes through a dynamic environment interaction module. This module continuously compares environmental information, determines whether changes exceed thresholds, and immediately triggers constraint updates and local path replanning upon detecting dynamic obstacles, significantly reducing replanning time and substantially improving the obstacle avoidance capability and flight reliability of UAVs in dynamic airspace. For multi-task and multi-UAV collaborative scenarios, the system further proposes strategies such as task priority ranking, regional clustering planning, multi-UAV collaborative task allocation, and dynamic scheduling. This not only optimizes the order and efficiency of single-UAV multi-task execution but also achieves load balancing and high robustness in UAV swarm operations through dynamic task handover and hierarchical airspace management, comprehensively improving the overall efficiency of large-scale, distributed operations.

[0052] Furthermore, the system's reserved dynamic task adjustment interface enables the UAV to cope with unexpected anomalies during mission execution. Through a feedback mechanism, it triggers a flexible reorganization of constraints and task logic, further enhancing the system's adaptability and mission completion rate in open and uncertain environments. In summary, this invention, through multi-layered and modular technological integration, systematically solves the core deficiencies of existing UAV path planning technologies in multi-constraint adaptation, dynamic response, multi-objective optimization, and multi-task collaboration. It provides a safe, efficient, and reliable autonomous flight solution for UAVs in diverse application scenarios such as inspection, rescue, and logistics, demonstrating significant practical value and broad applicability.

[0053] Compared with existing technologies, this method has the following advantages: 1. Multi-constraint collaborative adaptation enhances path safety and task completion rate. This invention transforms multi-dimensional constraints such as battery power, speed, obstacles, and task priority into mathematical conditions through a constraint parsing layer, ensuring that the planned path simultaneously meets multiple safety and task requirements. For example, when battery power is low, the algorithm prioritizes planning paths with lower energy consumption to avoid task failure due to battery depletion; in scenarios with obstacles and speed limitations, the path meets both safe distance requirements and does not exceed speed thresholds, significantly reducing the risk of drone collisions and equipment failures, and improving the task completion rate by more than 25% compared to existing technologies.

[0054] 2. Rapid response to dynamic environments, enhancing flight reliability. Leveraging the real-time environment recognition and rapid constraint update capabilities of the INEM module, this invention can complete path replanning after the appearance of dynamic obstacles within 0.3 seconds, an 80% speedup compared to existing algorithms (average replanning time 1.5 seconds). Simultaneously, by adjusting local paths rather than global replanning, computational resource consumption is reduced, ensuring stable flight of the UAV in complex dynamic environments (such as multi-UAV collaborative operations and sudden airflow), reducing collision risk by more than 60%.

[0055] 3. Multi-objective optimization balances practicality and economy. The collaborative work of the FFEM and Dueling modules achieves multi-objective balanced optimization of path length, energy consumption, stability, and task time. For example, in long-range missions, the planned path energy consumption is reduced by 30% compared to existing shortest path algorithms, and the drone's endurance is extended by 20%. In high-precision operations (such as aerial photography and inspection), the smoothness of the path improves flight attitude stability by 40%, and significantly enhances the working accuracy of onboard equipment (such as the clarity of aerial images and the accuracy of inspection data), balancing operational effectiveness and equipment usage costs.

[0056] 4. Strong adaptability to multi-task scenarios, improving operational efficiency. For multi-objective tasks, this invention can rationally plan the execution order and path connection based on task priority, avoiding path duplication and task delays. For example, in multi-location inspection tasks, the algorithm-planned path reduces flight mileage by 20% compared to existing technologies, shortens the total task time by 15%-20%, and ensures that high-priority tasks are completed first, meeting the high-efficiency operation needs of practical scenarios such as emergency rescue and engineering inspection, and has a wider range of applications.

[0057] Please see Figure 6 A second embodiment of the present invention provides an automatic path planning device based on multi-constraint dynamic optimization, comprising: The conversion unit 101 is used to acquire multi-dimensional information collected by the preset environmental perception module, and to convert the multi-dimensional information to obtain a set of mathematical constraints. The energy consumption calculation unit 102 is used to determine the mission constraints of the UAV according to the mathematical constraint group, and to calculate the energy consumption data of each candidate path based on the preset UAV parameters, UAV flight status and mission constraints using the FFEM module. The path planning unit 103 is used to call the Dueling module to perform weight allocation processing based on the task scenario and the energy consumption data of each candidate path, to give a comprehensive score to each candidate path after allocation, and to select the path with the highest comprehensive score as the optimal path, and to select the remaining paths as candidate paths. The dynamic obstacle avoidance unit 104 is used to compare the multi-dimensional information currently collected with the multi-dimensional information collected at the previous moment when the UAV is flying according to the optimal path. When it is determined that there is an environmental change, the INEM module is used to update the mathematical constraint group to achieve dynamic obstacle avoidance.

[0058] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. An automatic path planning method based on multi-constraint dynamic optimization, characterized in that, include: The system acquires multi-dimensional information collected by a preset environmental perception module, and transforms the multi-dimensional information to obtain a set of mathematical constraints. The task constraints of the UAV are determined based on the mathematical constraint set, and the energy consumption data of each candidate path is calculated using the FFEM module based on the preset UAV parameters, UAV flight status and task constraints. The Dueling module is called to perform weight allocation based on the task scenario and energy consumption data of each candidate path. The allocated candidate paths are then comprehensively scored, and the path with the highest comprehensive score is selected as the optimal path, while the remaining paths are selected as candidate paths. When the drone flies along the optimal path, the multi-dimensional information collected at the moment is compared with the multi-dimensional information collected at the previous moment. When an environmental change is detected, the INEM module is used to update the mathematical constraint group to achieve dynamic obstacle avoidance.

2. The automatic path planning method based on multi-constraint dynamic optimization according to claim 1, characterized in that, The environmental perception module includes a map module, a GPS positioning module, and a meteorological sensor module connected to a meteorological sensor interface; the multi-dimensional information includes the current latitude, longitude, and altitude data of the UAV, real-time meteorological data with wind speed ≤ preset wind speed value, and map module data.

3. The automatic path planning method based on multi-constraint dynamic optimization according to claim 1, characterized in that, The mathematical constraint group includes power constraints, speed constraints, obstacle constraints, and task constraints. Among them, the received multi-task requirements are transformed based on the preset task attribute-priority mapping rules to obtain task constraints; the power constraint is calculated based on the current power and energy consumption model, where the total energy consumption of the planned path is less than or equal to the maximum consuming energy corresponding to the remaining power; the speed constraint is determined by combining the preset safe speed range of the UAV and the wind speed, where the flight speed is less than or equal to the reserved wind resistance redundancy value; the obstacle constraint is the constraint that the distance between the static obstacle coordinates and the obstacle boundary is greater than or equal to the preset safe distance, which is the distance between the UAV flight trajectory and the obstacle boundary.

4. The automatic path planning method based on multi-constraint dynamic optimization according to claim 1, characterized in that, The drone's flight status includes level flight altitude, speed, and attitude angle, while the drone's parameters include fuselage weight and motor power.

5. The automatic path planning method based on multi-constraint dynamic optimization according to claim 1, characterized in that, While the drone is flying along the optimal path, the currently collected multi-dimensional information is compared with the multi-dimensional information collected at the previous moment. When an environmental change is detected, the INEM module is used to update the mathematical constraint group, specifically: When the drone flies along the optimal path, the change is calculated based on the multi-dimensional information collected at the current moment and the multi-dimensional information collected at the previous moment, and it is determined whether the change is greater than the preset value. When the change is found to be greater than the preset value, it is determined that the current environment has changed. Based on the change, the obstacle constraints in the mathematical constraint group are adjusted and the coordinate range of the obstacle constraints is updated. Based on the updated mathematical constraint group, the FFEM module and Dueling module are called in sequence to perform calculations and replan a new path to achieve dynamic obstacle avoidance.

6. The automatic path planning method based on multi-constraint dynamic optimization according to claim 1, characterized in that, For scenarios where multiple task points are scattered, a task clustering subunit is also included. The K-means algorithm is used to divide geographically close task points into the same region, and priority is given to planning paths within the same region before planning connecting paths between regions. At the same time, the regional task execution time is adjusted by combining the collected regional meteorological characteristics, and high-energy-consuming cross-regional flights are scheduled during tailwind periods.

7. The automatic path planning method based on multi-constraint dynamic optimization according to claim 1, characterized in that, For scenarios where the workload of multiple tasks exceeds the load of a single drone, a collaborative task allocation subunit is also included. Based on the remaining battery power, load capacity, and current location of each drone, tasks are allocated using a Hungarian algorithm to assign material delivery points that are less than a preset range away from each drone to the corresponding drone. At the same time, the flight altitudes of each drone are staggered.

8. The automatic path planning method based on multi-constraint dynamic optimization according to claim 7, characterized in that, When a drone's battery level drops below a preset value during a mission, the task is reassigned using a Hungarian algorithm based on the remaining battery level, load capacity, and current location of each drone. The unfinished task is then transferred to another drone with sufficient battery power.

9. The automatic path planning method based on multi-constraint dynamic optimization according to claim 1, characterized in that, In response to unexpected changes during the task, a dynamic task adjustment interface is also included. This interface is configured to: when a temporary obstacle is detected within a preset distance, a task execution anomaly signal is fed back to facilitate constraint updates.

10. An automatic path planning device based on multi-constraint dynamic optimization, characterized in that, include: The conversion unit is used to acquire multi-dimensional information collected by the preset environmental perception module, and to convert the multi-dimensional information to obtain a set of mathematical constraints. The energy consumption calculation unit is used to determine the mission constraints of the UAV based on the mathematical constraint set, and to calculate the energy consumption data of each candidate path based on the preset UAV parameters, UAV flight status and mission constraints using the FFEM module. The path planning unit calls the Dueling module to perform weight allocation processing based on the task scenario and the energy consumption data of each candidate path, performs a comprehensive score on each candidate path after allocation, selects the path with the highest comprehensive score as the optimal path, and selects the remaining paths as candidate paths. The dynamic obstacle avoidance unit is used to compare the multi-dimensional information collected at the current moment with the multi-dimensional information collected at the previous moment when the UAV is flying according to the optimal path. When it is determined that there is an environmental change, the INEM module is used to update the mathematical constraint group to achieve dynamic obstacle avoidance.

Images