Unmanned aerial vehicle path planning method based on Bayesian channel prediction and conflict resolution

Abstract

The invention discloses an unmanned aerial vehicle path planning method based on Bayesian channel prediction and conflict resolution, and belongs to the technical field of unmanned aerial vehicle path planning. According to the method, a multi-objective optimization model including communication constraint, energy consumption limitation and task timeliness requirements is constructed, and the complexity and challenge of an actual application scene are comprehensively reflected. In the first stage of the method, robust trajectory generation of a single unmanned aerial vehicle under dynamic wireless channel and complex environment disturbance is focused, real-time rolling optimization and dynamic adjustment of a flight path of the unmanned aerial vehicle are realized by fusing channel prediction of Bayesian filtering and robust model prediction control, and communication link stability and flight safety are guaranteed. In the second stage, on the basis of the initial trajectory generated in the previous stage, a local-cluster-global three-level cooperation mechanism is constructed based on an asynchronous layered ADMM distributed cooperation optimization framework, and safe cooperation work and task allocation of a large-scale unmanned aerial vehicle group are realized through efficient path conflict detection and scheduling coordination; and the overall scheduling efficiency and robustness of the system are obviously improved.

Description

Technical Field

[0001] This invention belongs to the field of UAV path planning technology. Specifically, it is a method that integrates Bayesian channel prediction, robust model predictive control and asynchronous hierarchical distributed optimization to solve the problems of multi-UAV collaborative scheduling, dynamic path optimization and communication link stability in complex urban environments. Background Technology

[0002] With the continuous development of emerging applications such as smart cities, public safety, and emergency management, urban sensing systems are placing higher demands on "last-mile" communication capabilities. In real-world scenarios, a large number of Internet of Things (IoT) terminal devices (such as environmental sensors, cameras, and wearable devices) distributed throughout urban areas are responsible for collecting critical information during emergencies, including emergency data such as fire alarms, explosion detection, and toxic gas leaks. This information needs to be transmitted to urban management platforms, emergency command centers, or edge servers as quickly as possible to assist in rapid decision-making and response.

[0003] However, due to limitations in the deployment density of terrestrial cellular base stations, complex terrain, and obstruction by tall buildings, signal blind spots and areas of communication discontinuity are widespread in urban environments. This is especially true in areas such as road corners, under overpasses, older residential areas, and underground passages, where ground equipment often cannot report data in a timely manner via traditional 4G / 5G communication links, severely impacting the timeliness and accuracy of information. In emergencies, cellular networks may also become completely unavailable due to congestion or damage, further exacerbating data upload problems.

[0004] To address the problem of urban communication blind spots, mobile relay platform technologies (such as vehicle-mounted communication, portable signal base stations, and unmanned aerial vehicle (UAV) communication systems) have become a research and application hotspot in recent years. In particular, unmanned aerial vehicle (UAV) platforms, due to their advantages such as rapid deployment, high-altitude command and control, flexible flight paths, and aerial loitering capabilities, are widely used in disaster relief, security patrols, and temporary network restoration. Multiple UAV systems, by constructing an aerial self-organizing network, can cover ground blind spots, enabling data collection and relay uploads from IoT terminals, thus becoming an important supplement to ground infrastructure. Summary of the Invention

[0005] This invention addresses the signal blind spots and unstable links encountered by Internet of Things (IoT) devices in complex urban environments when uploading emergency data. It proposes a two-stage solution based on multi-UAV collaborative scheduling and dynamic path optimization. The system centers on multiple UAVs and IoT devices deployed in the urban environment. It fully considers the impact of obstacles in the city on flight paths and communication links, constructing a multi-objective optimization model that includes communication constraints, energy consumption limitations, and task timeliness requirements. This model comprehensively reflects the complexity and challenges of real-world application scenarios, significantly improving the efficiency and reliability of UAV-assisted emergency information transmission for IoT terminals in urban environments, and meeting the real-time, efficient communication needs of smart cities and emergency management.

[0006] In terms of scheme design, the first phase focuses on robust trajectory generation for a single UAV under dynamic wireless channels and complex environmental disturbances. By integrating Bayesian filtering for channel prediction with robust model predictive control, real-time rolling optimization and dynamic adjustment of the UAV's flight path are achieved, ensuring communication link stability and flight safety. The second phase, based on the initial trajectory generated in the first phase, constructs a three-level cooperation mechanism—local, cluster, and global—based on the asynchronous hierarchical ADMM distributed collaborative optimization framework. Through efficient path conflict detection and scheduling coordination, safe collaborative operation and task allocation for large-scale UAV swarms are achieved, significantly improving the overall scheduling efficiency and robustness of the system.

[0007] The technical solution of the present invention will be described in detail below.

[0008] A method for UAV path planning based on Bayesian channel prediction and conflict resolution includes the following steps:

[0009] Step 1: Establish a multi-UAV collaborative scheduling and trajectory planning system model in an urban environment; the system includes a set of IoT devices, a set of UAVs, a set of obstacles in the urban environment, and base stations. There are communication links between IoT devices, UAVs, and base stations. Establish an air-to-ground communication model.

[0010] Step 2: Establish a joint optimization problem of task scheduling and flight path, determine the task allocation variable, UAV trajectory position sequence, and transmission power allocation variable as decision variables, and set constraints, including communication, time, energy, and area constraints;

[0011] Step 3: Dynamic robust trajectory generation; Construct a dynamic channel state prediction model based on Bayesian filtering to predict the signal-to-interference-plus-noise ratio (SINR), then estimate the upper bound of the channel prediction error using the ellipsoidal uncertainty set modeling method, and finally use a robust model predictive control architecture to achieve rolling optimization of the UAV trajectory while satisfying the constraints.

[0012] Step 4: Distributed collaborative path optimization; The system is divided into local layer, cluster layer and global layer; The local layer includes a single UAV, and the optimized single-UAV trajectory is obtained through the above steps; The cluster layer is divided into multiple UAV clusters based on spatial proximity, and the trajectories are coordinated within the cluster; The global layer periodically collects all cluster boundary trajectories to detect whether there are conflicts between cross-cluster trajectories and coordinates the trajectories.

[0013] Furthermore, step 1 is detailed as follows:

[0014] Step 1.1: Construct the system model; the system consists of a set of M IoT devices. N drones gathered A set of m obstacles in an urban environment It consists of a base station; each IoT device i has three-dimensional position coordinates (x, y, y). i ,y i ,z i The amount of data to be uploaded is D. i and upload deadline δ i Each U drone j Having an initial position (x) u ,y u ,z u Maximum flight time Maximum energy capacity and available transmission power P u Power control range Each obstacle m Define its spatial location and shape. m = {(x,y,z)(x,y,z)∈the space occupied by obstacles}.

[0015] Step 1.2 Communication Model; Based on the air-to-ground communication model, the communication channel between the UAV u and the IoT device i meets the minimum signal-to-interference-plus-noise ratio (SINR) threshold requirement to ensure reliable data transmission. SINR is defined as:

[0016]

[0017] Among them, P u，i (t) represents the transmission power allocated by the drone u to the IoT device i at time t; h u，i (t) represents the unmanned aerial vehicle u j Channel gain to IoT device i; σ 2 I represents the noise power at the receiving end. u，i (t) represents the interference power; the transmission rate is determined by the bandwidth B and SINR, satisfying the Shannon formula:

[0018] R u，i (t)=B log2(1+SINRu，i (t))

[0019] And ensure that the SINR of all communication links is greater than or equal to the threshold γ. min .

[0020] Furthermore, the optimization objective of the joint optimization problem of task scheduling and flight path is to minimize the maximum time for all IoT devices to complete data upload through task allocation and path planning.

[0021] Furthermore, step 3 is detailed as follows:

[0022] Step 3.1: Dynamic Channel Modeling and SINR Prediction; The system collects historical communication data between the UAV and IoT devices in real time and constructs a dynamic channel state prediction model based on Bayesian filtering:

[0023]

[0024] in This represents the signal-to-interference-plus-noise ratio (SIR) at time t, predicted based on existing data at time t+1. SINR represents expectations. u，i (t+1) represents the actual SINR value between UAV u and device i at time t+1; y 1：t The historical channel observation dataset, from time 1 to time t, is used as the input for prediction;

[0025] Step 3.2: Ellipsoidal uncertainty set construction and robust error modeling; determining the range of channel prediction error based on the ellipsoidal uncertainty set modeling method:

[0026]

[0027] Where ε is the uncertainty set, representing the set of all possible channel prediction errors δ. The error vector δ represents the difference between the actual channel and the predicted channel in n-dimensional space, where δ is the difference between the actual channel and the predicted channel. -1 It is the inverse of the covariance matrix, χ 2 Chi-square value;

[0028] Step 3.3: Robust model predictive control trajectory rolling optimization; Based on the above channel prediction and uncertainty set construction results, within each control period k, the system considers the future planning period T. p The flight trajectory of the UAV is optimized by minimizing the costs of trajectory smoothing and energy consumption, taking into account the impact of mid-channel fluctuations and environmental disturbances. A sequence of UAV trajectory points is generated, and the constraints are continuously satisfied throughout the planning period.

[0029]

[0030] in, E represents the two-dimensional position of the UAV u at time t; u (t) represents the energy consumption of UAV u at time t; λ>0 represents the weighting coefficient;

[0031] Step 3.4: Design of Bézier curve interpolator for obstacle avoidance;

[0032] The system introduces a Bézier interpolator to interpolate the trajectory point sequence generated by RMPC, producing a smooth trajectory:

[0033]

[0034] Where b k It is a control point, B k,n (t) is a Bézier basis function.

[0035] Furthermore, step 4 is detailed as follows:

[0036] Step 4.1: Design of the asynchronous layered ADMM optimization framework;

[0037] Local layer: Optimized single-drone trajectory {q} for each drone u u (t)} u∈U ;

[0038] Cluster layer: Drones are divided into multiple clusters based on spatial proximity. in This represents the k-th subset of drones, where drones within the cluster share the boundary trajectory variable {q}. c (t)};Trajectory coordination is achieved through asynchronous ADMM mechanism, and trajectory optimization within the cluster is performed to ensure the trajectory of each UAV {q u (t)} and cluster boundary trajectory {q c (t)} is consistent: The drones in the cluster simultaneously optimize trajectory smoothness and energy consumption, and work together to avoid local trajectory conflicts;

[0039] Global layer: Periodically collect all cluster boundary trajectories {q c (t)}, detect whether there is a conflict in the cross-cluster trajectory, specifically determined by: the existence of a cluster drones in and cluster C m drones in And at time t, satisfying |q u (t)-q v (t)|<d safe Where, d safeThe safe distance threshold for trajectory conflict detection; if a cross-cluster conflict is detected, the system coordinates the relevant clusters to adjust the trajectory through global broadcast to ensure the overall path safety;

[0040] Step 4.2: Maintaining safe distances and determining convex hull conflicts among multiple UAVs; Each UAV generates a dynamic convex hull representation of its flight path based on its own trajectory. Through an O(1) complexity intersection detection mechanism, it quickly determines whether there is an intersection or potential collision risk between adjacent paths. The determination conditions are as follows:

[0041] and

[0042] Where Conv u Conv v Let U and V be the convex hulls of the drone's trajectories, respectively. The convex hull is the smallest convex polygon that encloses the flight path, representing the safe activity area, and is denoted as Conv. u Conv v Any point within the cluster; when a potential conflict is detected, the system triggers a boundary variable update mechanism within the cluster to adjust a portion of the trajectory of the conflicting drone to restore a safe distance;

[0043] Step 4.3: Cross-cluster conflict detection and scheduling coordination; The system deploys a cross-cluster conflict detection module at the global layer, periodically collecting boundary path segment information between clusters, and coordinating the scheduling of each pair of cluster boundary trajectories q. c and q c′ Detect whether there exists a time point t such that the distance between two trajectories is less than a threshold d. safe :

[0044]

[0045] If a potential cross-cluster path conflict is detected, the system prioritizes dynamically adjusting the cluster that has the least impact on the boundary path. Specifically:

[0046]

[0047] Among them, C k This represents the k-th drone cluster, where all drones in the cluster participate in the adjustment. This represents the new trajectory coordinates of the UAV u after time t adjustment; simultaneously, the scheduling and coordination mechanism re-plans the equipment task allocation variables involved in the conflict area.

[0048] Furthermore, the specific constraints are as follows:

[0049] Communication constraints: Ensure that the link between the drone and the equipment meets the minimum SINR threshold;

[0050] Energy Constraint: The total energy consumption of the drone during the mission shall not exceed its maximum energy capacity;

[0051] Time constraint: Device data upload time must not exceed its deadline;

[0052] Flight area constraints: The drone's trajectory must remain within the legal flight area;

[0053] Obstacle avoidance constraints: Drones must not enter areas of space occupied by obstacles at any time.

[0054] This invention significantly improves the adaptability of UAV path planning to changes in the communication environment by introducing dynamic link quality prediction based on Bayesian filtering, solving the problem that traditional static path planning struggles to cope with link fluctuations and environmental disturbances. Furthermore, robust model predictive control is used to achieve dynamic rolling optimization of the trajectory, ensuring the stability of the flight path and communication reliability in complex urban environments. In terms of multi-UAV cooperative scheduling, this invention innovatively constructs an asynchronous hierarchical ADMM distributed optimization framework, effectively overcoming the computational and communication bottlenecks in large-scale UAV systems, achieving efficient path conflict detection and rapid resolution, and improving the system's real-time performance and scalability. Combined with dynamic transmission power adjustment and energy management, the overall solution maximizes energy efficiency and system task coverage while ensuring mission timeliness, significantly outperforming existing technologies. Attached Figure Description

[0055] Figure 1 This is a schematic diagram of the system physical architecture and application scenario in an example of the present invention.

[0056] Figure 2 This is an overall flowchart of the multi-UAV dynamic scheduling system in an example of the present invention.

[0057] Figure 3 This is a schematic diagram of the asynchronous hierarchical ADMM collaborative optimization framework in an example of the present invention. Detailed Implementation

[0058] A method for UAV path planning based on Bayesian channel prediction and conflict resolution, such as Figure 2 As shown, it includes the following steps:

[0059] Step 1: Establish a model for a multi-UAV collaborative scheduling and trajectory planning system in an urban environment;

[0060] Step 1.1: Construct the system model;

[0061] like Figure 1 As shown, the system consists of a set of M IoT devices. N drones gathered A set of m obstacles in an urban environment It consists of one or more base stations. Each IoT device i has three-dimensional location coordinates (x, y, y). i y i , z i The amount of data to be uploaded is D. i and upload deadline δ i Each drone U j Having an initial position (x) u ,y u ,z u Maximum flight time Maximum energy capacity and available transmission power This refers to the power control range of the drone's communication module. Each obstacle... m Define its spatial location and shape. m = {(x,y,z)(x,y,z)∈the space occupied by obstacles}.

[0062] Step 1.2 Communication Model

[0063] Based on the air-to-ground (A2G) communication model, the communication channel between the UAV u and the IoT device i meets the minimum signal-to-interference-plus-noise ratio (SINR) threshold requirement to ensure reliable data transmission. SINR is defined as:

[0064]

[0065] Among them, P u,i (t) represents the transmission power (in watts) allocated by the drone u to the IoT device i at time t, which is a decision variable and is dynamically adjusted by the system; h u，i (t) represents the unmanned aerial vehicle u j The channel gain to IoT device i reflects signal propagation loss, including path loss, shadowing fading, and multipath fading. It is determined by environmental factors and location, and is obtained through wireless channel measurement or modeling; σ 2 The receiver noise power is typically a statistical estimate of the ambient noise and is a system constant; I u，i (t) represents the interference power, indicating signal interference from other drones or devices, which changes dynamically with the system scheduling status. The transmission rate is determined by the bandwidth B and SINR, satisfying the Shannon formula:

[0066] R u，i (t)=B log2(1+SINR u，i (t))

[0067] In practice, bandwidth B is the frequency band width specified by the wireless communication protocol. Ensure that the SINR of all communication links meets the threshold γ. min ,Right now

[0068]

[0069] Where γ min The minimum SINR threshold specified for the communication system ensures the reliability and bit error rate of data transmission.

[0070] Step 2: Establish a joint optimization problem for task scheduling and flight path;

[0071] The optimization goal is to minimize the maximum time for all IoT devices to complete data uploads through task allocation and path planning.

[0072]

[0073] Among them, T i This represents the data upload completion time (in seconds) for device i. This variable is jointly determined by drone path planning, task allocation, and transmission scheduling. For example, if device A generates 1MB of emergency data and uploads it via drone relay, T... A This refers to the time span from the start of data transmission to its completion.

[0074] Step 2.1 Determine the decision variables;

[0075] Task allocation variable x u，i (t)∈{0,1}, indicating whether UAV u is responsible for receiving data from device i at time t, determined by the scheduling algorithm, and the UAV trajectory position sequence q. u (t)=(x j (t), y j (t) represents the two-dimensional spatial position of UAV u at time t, output by the trajectory planning module; transmission power allocation P u，i (t) represents the power allocated by UAV u to device i at time t, which is a dynamically adjusted variable that satisfies the upper and lower power limits.

[0076] Step 2.2: Constraints to be satisfied;

[0077] 2.2.1 Communication constraints to ensure that the link between the UAV and the equipment meets the minimum SINR threshold:

[0078]

[0079] This constraint ensures the reliability of data transmission and prevents signal interruption.

[0080] 2.2.2 Energy Constraint: The total energy consumption of the UAV during the mission shall not exceed its maximum energy capacity.

[0081]

[0082] Where E u(t) represents the energy consumption of the UAV u at time t, including flight energy consumption and communication energy consumption, which is calculated by the energy model.

[0083] 2.2.3 Time Constraints, Device I i Data upload time must not exceed its deadline:

[0084]

[0085] Ensure that urgent data is uploaded in a timely manner within the specified time.

[0086] 2.2.4 Flight area constraints: The drone's trajectory must remain within the legal flight area.

[0087]

[0088] Where A valid This refers to the airspace areas within a city where flying is permitted, designated by the relevant management authorities.

[0089] The trajectory position sequence q of the drone u (t)=(x u (t), y u (t), z u (t) must satisfy the obstacle avoidance constraint:

[0090]

[0091] This means that drones must not enter areas occupied by obstacles at any time.

[0092] Step 3: Dynamic Robust Trajectory Generation (RCATC)

[0093] Step 3.1: Dynamic channel modeling and SINR prediction;

[0094] Considering the issues of fluctuating communication channel quality (such as signal-to-interference-plus-noise ratio, SINR) caused by high-rise buildings and frequent multipath effects in urban environments, the system collects historical communication data between the UAV and IoT devices in real time and constructs a dynamic channel state prediction model based on Bayesian filtering:

[0095]

[0096] in This represents the signal-to-interference-plus-noise ratio (SINR) at time t, predicted based on existing data for time t+1. Expectation, or average in mathematics, is the value of SINR. Here, the expectation symbol indicates that we are estimating the SINR for future times based on existing historical observation data, and that this estimation considers all possible scenarios. u，i(t+1) represents the actual SINR value between UAV u and device i at time t+1. This is a measure of the actual channel quality, which depends on factors such as signal strength, interference, and noise; y 1：t The historical channel observation dataset, from time 1 to time t, is used as the input for prediction.

[0097] Based on observations over a period of time, this model can predict the SINR change trend over several future time slots, thereby anticipating potential link degradation areas and providing prior support for trajectory adjustment.

[0098] Step 3.2: Construction of ellipsoidal uncertainty set and robust error modeling;

[0099] Because channel prediction inherently involves errors, this scheme introduces an ellipsoidal uncertainty set modeling method to estimate the upper bound of the channel prediction error. This method describes the possible range of error variation during channel prediction. The uncertainty set of the channel gain is defined as follows:

[0100]

[0101] Where ε is the uncertainty set, representing the set of all possible channel prediction errors δ. That is, this set contains all possible error values ​​and describes the range of channel prediction errors; Let δ represent the error vector in n-dimensional space. Here, δ is a vector representing the error in channel prediction, typically the difference between the actual channel and the predicted channel. It is a quadratic expression representing the magnitude of the error. Specifically, It is the transpose of the error vector, P -1 χ is the inverse of the covariance matrix, representing the distribution of the error and the channel prediction uncertainty. This expression quantifies the "magnitude" of the error, i.e., the uncertainty of the error relative to the channel prediction. 2 χ² is a constant called the chi-square value, which represents the tolerance range of error and is usually related to the confidence level. A smaller χ² value... 2 A value of χ² means we require a small prediction error, while a large χ² value indicates a higher prediction error. 2 A higher value allows for a larger prediction error.

[0102] The ellipsoidal uncertainty set defines the possible range of channel gain fluctuations, providing a safety boundary for robust trajectory planning. By considering the worst-case channel conditions, the system can generate paths that remain feasible and stable in real-world environments, improving the robustness and reliability of task scheduling.

[0103] Step 3.3: Robust Model Predictive Control (RMPC) Trajectory Rolling Optimization

[0104] Based on the aforementioned channel prediction and uncertainty set construction results, the trajectory optimization module employs a robust model predictive control (RMPC) architecture to achieve rolling optimization of the UAV trajectory. Within each control period k, the system considers future planning periods T. p The flight trajectory of the UAV is optimized by minimizing the costs of trajectory smoothing and energy consumption in order to mitigate the impact of mid-channel fluctuations and environmental disturbances, and to generate a sequence of UAV trajectory points that continuously meet communication quality, flight constraints, and energy constraints throughout the planning period.

[0105]

[0106] in, E represents the two-dimensional position of the UAV u at time t; u (t) represents the energy consumption of UAV u at time t; λ > 0 represents the weighting coefficient, used to balance trajectory smoothness and energy consumption. Subject to the following constraints:

[0107] (1) The total energy consumption of the UAV during the mission shall not exceed its maximum energy capacity:

[0108]

[0109] (2) Flight area constraints

[0110] The drone's trajectory must remain within the legal flight area:

[0111]

[0112] Where A valid This indicates the airspace areas where flying is permitted within the city, and the flight path and location sequence q of the drones are determined by the management department. u (t)=(x u (t), y u (t), z u (t) must satisfy the obstacle avoidance constraint:

[0113]

[0114] This means that drones must not enter areas occupied by obstacles at any time.

[0115] Step 3.4: Design of a Bézier Curve Interpolator for Obstacle Avoidance

[0116] To further improve trajectory smoothness and obstacle avoidance capabilities, the system introduces a Bézier interpolator to interpolate the trajectory point sequence generated by RMPC. This generates a smooth trajectory.

[0117]

[0118] Where qu (y) represents the position of the UAV u at time t. It is a weighted sum of Bézier curves. This is achieved through control point b. k and basis function B k,n (t), weighted summation gives the specific location of the drone at time t, b k These are control points, which determine the shape of the curve. Each control point corresponds to a point on the trajectory, and the location and number of control points will affect the direction and curvature of the curve. k，n (t) is a Bézier basis function that acts as a weight in curve generation, controlling the influence of each control point on the final trajectory position. The shape of the basis function determines the range and degree of influence of the control points on the trajectory.

[0119] Step 4: Distributed Cooperative Path Optimization (AH-ADMM);

[0120] Step 4.1: Design of the Asynchronous Layered ADMM Optimization Framework

[0121] The first phase ensured smooth trajectory and energy optimization for each drone in complex environments. However, in multi-drone collaborative scenarios, in addition to individual trajectory optimization, collective path coordination and conflict avoidance are also required. Therefore, as... Figure 3 As shown, the system is designed with the following framework to handle path conflict issues among multiple UAVs:

[0122] (1) Local layer: Each UAV u generates an optimized single-machine trajectory {q} after the first stage. u (t)} u∈U ;

[0123] (2) Cluster layer: UAVs are divided into multiple clusters based on spatial proximity. in This represents the k-th subset of drones, where drones within the cluster share the boundary trajectory variable q. c (t), used to coordinate drone trajectories and avoid conflicts. To ensure that the trajectories q of each drone within the cluster are consistent... u (t) and cluster boundary trajectory q c (t) remains consistent: Within a drone swarm, drones simultaneously optimize trajectory smoothness and energy consumption, collaboratively avoiding local trajectory conflicts. The trajectory optimization problem within a swarm can be expressed as:

[0124]

[0125] Where, q u (t) is the trajectory of U drone u in the cluster at time t, q c (t) is the boundary trajectory shared by the cluster. λ s and λ eThese are weighting coefficients, used to balance trajectory smoothness and energy consumption, respectively. |q u (t+1)-2q u (t)+q u (t-1)| 2 This represents the trajectory smoothness term, while It is the propulsion energy consumption of the UAV at time t.

[0126] (3) Global layer: Periodically collect all cluster boundary trajectories {q c (t)}, detect whether there is a conflict in the cross-cluster trajectory, specifically determined by: the presence of drones. and And at time t, satisfying |q u (t)-q v (t)|<d safe Where, d safe This is the safe distance threshold for trajectory conflict detection. If a cross-cluster conflict is detected, the system coordinates with relevant clusters to adjust the trajectory via global broadcast to ensure overall path safety.

[0127] Step 4.2: Maintaining safe distances between multiple drones and determining convex hull conflicts;

[0128] To ensure that cooperative paths do not interfere with each other, a path conflict determination algorithm based on convex hull envelopes is designed. Each UAV generates a dynamic convex hull representation of its flight path based on its own trajectory. An O(1) complexity intersection detection mechanism is used to quickly determine whether there is an intersection or potential collision risk between adjacent paths. Determination criteria:

[0129] and

[0130] Where Conv u Conv v Let p and q be the convex hulls of the drone's trajectories u and v, respectively (the smallest convex polygon enclosing the flight path, representing the "safe activity area"), and let p and q be the convex hulls Conv. u Conv v An arbitrary point within the range is used to calculate the minimum distance between two drones, d. safe The system determines the safe distance threshold by judging whether the "safe activity areas" of two drones overlap and are too close, and detects the risk of collision in real time (similar to how autonomous vehicles use LiDAR to detect the position of surrounding vehicles).

[0131] When a potential conflict is detected, the system triggers a boundary variable update mechanism within the cluster, adjusting certain trajectory segments of the conflicting drones to restore a safe distance. Compared to traditional point-to-point conflict detection methods, this algorithm significantly reduces computational complexity and system latency, making it suitable for rapid conflict identification and response in large-scale drone systems.

[0132] Step 4.3: Cross-cluster conflict detection and scheduling coordination;

[0133] Based on the cluster path optimization of region division, to solve the problem of inconsistent or mutually occluded path boundaries between clusters, the system deploys a cross-cluster conflict detection module at the global layer. This module periodically collects boundary path segment information between each cluster and performs a cross-cluster conflict detection on each pair of cluster boundary trajectories q. c and q c′ Detect whether there exists a time point t such that the distance between two trajectories is less than a threshold d. safe :

[0134]

[0135] If a potential cross-cluster path conflict is detected, the system prioritizes dynamically adjusting the cluster that has the least impact on the boundary path. Specifically:

[0136]

[0137] Among them, C k This indicates the k-th drone cluster, where all drones in the cluster will participate in the adjustment. This indicates that drone u belongs to this cluster; t is the time index, representing a discrete time point on the trajectory. q u (t) represents the original trajectory coordinates of the UAV u at time t; This represents the new trajectory coordinates of the drone u after time t adjustment. This represents the square of the trajectory adjustment distance of the UAV u at time t, reflecting the magnitude of the trajectory change. Simultaneously, the scheduling and coordination mechanism re-plans the equipment task allocation variable x involved in the conflict area. u,i (t) ensures that each task is uniquely assigned, while spatially distributed processing ensures path security and task coverage.

[0138] It is understood that the present invention has been described through some embodiments, and those skilled in the art will recognize that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of the invention. Furthermore, under the teachings of the present invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the invention. Therefore, the present invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the protection scope of the present invention.

Claims

1. A method for UAV path planning based on Bayesian channel prediction and conflict resolution, characterized in that, Includes the following steps: Step 1: Establish a multi-UAV collaborative scheduling and trajectory planning system model in an urban environment; the system includes a set of IoT devices, a set of UAVs, a set of obstacles in the urban environment, and base stations. There are communication links between IoT devices, UAVs, and base stations. Establish an air-to-ground communication model. Step 2: Establish a joint optimization problem of task scheduling and flight path, determine the task allocation variable, UAV trajectory position sequence, and transmission power allocation variable as decision variables, and set constraints, including communication, time, energy, and area constraints; Step 3: Dynamic robust trajectory generation; Construct a dynamic channel state prediction model based on Bayesian filtering to predict the signal-to-interference-plus-noise ratio (SINR), then estimate the upper bound of the channel prediction error using the ellipsoidal uncertainty set modeling method, and finally use a robust model predictive control architecture to achieve rolling optimization of the UAV trajectory while satisfying the constraints. Step 4: Distributed collaborative path optimization; The system is divided into local layer, cluster layer and global layer; The local layer includes a single UAV, and the optimized single-UAV trajectory is obtained through the above steps; The cluster layer is divided into multiple drone clusters based on spatial proximity, and the trajectories are coordinated within the clusters; The global layer periodically collects all cluster boundary trajectories to detect and coordinate cross-cluster trajectories to prevent conflicts.

2. The UAV path planning method based on Bayesian channel prediction and conflict resolution according to claim 1, characterized in that, Step 1 is described in detail as follows: Step 1.1: Construct the system model; the system consists of a set of M IoT devices. N drones gathered A set of m obstacles in an urban environment It consists of a base station; each IoT device i has three-dimensional position coordinates (x, y, y). i y i , z i The amount of data to be uploaded is D. i and upload deadline δ i Each U drone j Having an initial position (x) u y u , z u Maximum flight time Maximum energy capacity and available transmission power P u Power control range Each obstacle m Define its spatial location and shape. m = {(x, y, z) | (x, y, z) ∈ the space occupied by obstacles}. Step 1.2 Communication Model; Based on the air-to-ground communication model, the communication channel between UAV u and IoT device i meets the minimum signal-to-interference-plus-noise ratio (SINR) threshold requirement. SINR is defined as: Among them, P u，i (t) represents the transmission power allocated by the drone u to the IoT device i at time t; h u，i (t) represents the unmanned aerial vehicle u j Channel gain to IoT device i; σ 2 I represents the noise power at the receiving end. u，i (t) represents the interference power; the transmission rate is determined by the bandwidth B and SINR, satisfying the Shannon formula: R u，i (t)=Blog2(1+SINR u，i (t)) and ensure that the SINR of all communication links is greater than or equal to the threshold γ. min .

3. The UAV path planning method based on Bayesian channel prediction and conflict resolution according to claim 2, characterized in that, The optimization objective of the joint optimization problem of task scheduling and flight path is to minimize the maximum time for all IoT devices to complete data upload through task allocation and path planning.

4. The UAV path planning method based on Bayesian channel prediction and conflict resolution according to claim 3, characterized in that, Step 3 is as follows: Step 3.1: Dynamic Channel Modeling and SINR Prediction; The system collects historical communication data between the UAV and IoT devices in real time and constructs a dynamic channel state prediction model based on Bayesian filtering: in This represents the signal-to-interference-plus-noise ratio (SIR) at time t, predicted based on existing data at time t+1. SINR represents expectations. u,i (t+1) represents the actual SINR value between UAV u and device i at time t+1; y 1：t The historical channel observation dataset, from time 1 to time t, is used as the input for prediction; Step 3.2: Ellipsoidal uncertainty set construction and robust error modeling; determining the range of channel prediction error based on the ellipsoidal uncertainty set modeling method: Where ε is the uncertainty set, representing the set of all possible channel prediction errors δ. The error vector δ represents the difference between the actual channel and the predicted channel in n-dimensional space, where δ is the difference between the actual channel and the predicted channel. -1 It is the inverse of the covariance matrix, χ 2 Chi-square value; Step 3.3: Robust model predictive control trajectory rolling optimization; Based on the above channel prediction and uncertainty set construction results, within each control period k, the system considers the future planning period T. p The flight trajectory of the UAV is optimized by minimizing the costs of trajectory smoothing and energy consumption, taking into account the impact of mid-channel fluctuations and environmental disturbances. A sequence of UAV trajectory points is generated, and the constraints are continuously satisfied throughout the planning period. in, E represents the two-dimensional position of the UAV u at time t; u (t) represents the energy consumption of UAV u at time t; λ>0 represents the weighting coefficient; Step 3.4: Design of Bézier curve interpolator for obstacle avoidance; The system introduces a Bézier interpolator to interpolate the trajectory point sequence generated by RMPC, producing a smooth trajectory: Where b k It is a control point, B k，n (t) is a Bézier basis function.

5. The UAV path planning method based on Bayesian channel prediction and conflict resolution according to claim 4, characterized in that, Step 4 is described in detail below: Step 4.1: Design of the asynchronous layered ADMM optimization framework; Local layer: Optimized single-drone trajectory q for each drone u u (y); Cluster layer: Drones are divided into multiple clusters based on spatial proximity. in This represents the k-th subset of drones, where drones within the cluster share the boundary trajectory variable q. c (t); By coordinating drones within a cluster, trajectory smoothness and energy consumption are simultaneously optimized, and trajectory optimization is performed to ensure that the trajectories of each drone q are guaranteed. u (t) and cluster boundary trajectory qc ( t) Consistent: q u (t)≈q c (t), Avoid local trajectory conflicts; Global layer: Periodically collect all cluster boundary trajectories q c (t), detect whether there is a conflict in the cross-cluster trajectory, specifically determined by: the existence of a cluster drones in and cluster C m drones in And at time t, satisfying |q u (t)-q v (t)|<d safe Where, d safe The safe distance threshold for trajectory conflict detection; if a cross-cluster conflict is detected, the system coordinates the relevant clusters to adjust the trajectory through global broadcast to ensure the overall path safety; Step 4.2: Maintaining safe distances and determining convex hull conflicts among multiple UAVs; Each UAV generates a dynamic convex hull representation of its flight path based on its own trajectory. Through an O(1) complexity intersection detection mechanism, it quickly determines whether there is an intersection or potential collision risk between adjacent paths. The determination conditions are as follows: and Where Conv u Conv v Let U and V be the convex hulls of the drone's trajectories, respectively. The convex hull is the smallest convex polygon that encloses the flight path, representing the safe activity area, and is denoted as Conv. u Conv v Any point within the cluster; when a potential conflict is detected, the system triggers a boundary variable update mechanism within the cluster to adjust part of the trajectory segment of the conflicting drone to restore a safe distance; Step 4.3: Cross-cluster conflict detection and scheduling coordination; The system deploys a cross-cluster conflict detection module at the global layer, periodically collecting boundary path segment information between clusters, and coordinating the scheduling of each pair of cluster boundary trajectories q. c and q c′ Detect whether there exists a time point t such that the distance between two trajectories is less than a threshold d. safe : If a potential cross-cluster path conflict is detected, the system prioritizes dynamically adjusting the cluster that has the least impact on the boundary path. Specifically: Among them, C k This represents the k-th drone cluster, where all drones in the cluster participate in the adjustment. This represents the new trajectory coordinates of the UAV u after time t adjustment; simultaneously, the scheduling and coordination mechanism re-plans the equipment task allocation variables involved in the conflict area.

6. The UAV path planning method based on Bayesian channel prediction and conflict resolution according to claim 5, characterized in that, The specific constraints are as follows: Communication constraints: Ensure that the link between the drone and the equipment meets the minimum SINR threshold; Energy Constraint: The total energy consumption of the drone during the mission shall not exceed its maximum energy capacity; Time constraint: Device data upload time must not exceed its deadline; Flight area constraints: The drone's trajectory must remain within the legal flight area; Obstacle avoidance constraints: Drones must not enter areas of space occupied by obstacles at any time.

7. The UAV path planning method based on Bayesian channel prediction and conflict resolution according to claim 6, characterized in that, The intra-cluster trajectory optimization problem is represented as: Where, q u (t) is the trajectory of U drone u in the cluster at time t, q c (t) is the shared boundary trajectory of the cluster; λ s and λ e These are weighting coefficients, which balance trajectory smoothness and energy consumption respectively; |q u (t+1)-2q u (t)+q u (t-1)| 2 This represents the trajectory smoothness term, while It is the propulsion energy consumption of the UAV at time t.

Images