Pole transfer unmanned aerial vehicle group cooperative positioning method and system

Abstract

This invention provides a collaborative positioning method and system for a swarm of UAVs transporting power poles, encompassing satellite positioning, radio positioning, and multi-source fusion positioning technologies. The method includes the following steps: when encountering GPS signal degradation, acquiring inter-UAV distance constraints and jointly calculating centimeter-level distributed positioning with carrier phase observations; synchronously collecting airborne inertial data and load data from the suspended power poles, identifying rigid-flexible coupling dynamic parameters through pre-integration and nonlinear estimation, and calculating the dynamic spatial envelope of the power poles to generate a three-dimensional track; when inter-UAV communication quality deteriorates, initiating visual-radio collaborative positioning, fusing relative line-of-sight vectors and distance information, and achieving consistent collaborative positioning with global spatiotemporal reference alignment across multiple UAVs through factor graph optimization. The advantages of this invention are: significantly improved positioning accuracy under electromagnetic degradation environments, enhanced multi-UAV relative constraint positioning capabilities, and guaranteed positioning consistency with global spatiotemporal reference alignment across multiple UAVs in communication degradation scenarios.

Description

Technical Field

[0001] This invention relates to the fields of satellite positioning, radio positioning and multi-source fusion positioning technology, and in particular to a collaborative positioning method and system for a swarm of UAVs used for pole transportation. Background Technology

[0002] In recent years, frequent disasters such as torrential rains and ice storms have led to widespread pole collapses and line breaks in power distribution networks. To improve the efficiency of emergency repairs, it is necessary to utilize drone swarms to transport segmented composite material poles to various repair points. During highly dynamic transport operations, the high-precision and robust collaborative positioning of drone swarms becomes a core prerequisite for ensuring operational safety and efficiency. Existing positioning technologies mainly face the following technical bottlenecks: First, GNSS precise positioning calculations fail under electromagnetic degradation environments. Disasters cause widespread damage to communication base stations, and terrain such as mountainous areas and canyons obstruct satellite signals and create multipath effects. Conventional single-unit positioning architectures that rely on Global Navigation Satellite System (GNSS) and Real-time Dynamic Differential (RTK) can only achieve integer ambiguity in open areas with abundant signal. They cannot maintain centimeter-level positioning accuracy under conditions of satellite signal attenuation, strong electromagnetic interference, and communication blind spots, leading to drift or jumps in positioning results.

[0003] Second, the lack of relative constraints between the machines leads to insufficient distributed positioning accuracy. The hoisting of long utility poles requires multi-machine collaborative operation. Existing positioning schemes do not introduce relative distance constraints between the machines via radio, making it difficult to effectively constrain the integer ambiguity search space when GNSS observation quality is poor, and thus failing to achieve centimeter-level precise positioning for multi-machine distributed systems.

[0004] Third, the lack of a mechanism for multi-aircraft cooperative positioning and global spatiotemporal reference alignment under the condition of degraded inter-aircraft communication. In limited airspace, signals from multiple aircraft are prone to co-channel interference and link interruptions. Existing multi-aircraft positioning technologies are mainly based on the assumption of perfect communication and lack distributed cooperative positioning correction methods for link degradation. Each airborne terminal cannot rely on a single sensor to complete the accurate calculation of the relative pose of multiple aircraft and the alignment with the global spatiotemporal reference, resulting in inconsistent spatiotemporal references for aircraft group positioning and poor positioning consistency during dense formation operations.

[0005] Therefore, how to provide a collaborative positioning method and system for pole-transfer drone swarms to improve positioning accuracy in electromagnetic degradation environments, enhance multi-drone relative constraint positioning capabilities, and ensure positioning consistency of multi-drone global spatiotemporal reference alignment in communication degradation scenarios has become an urgent technical problem to be solved. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a collaborative positioning method and system for pole-transfer unmanned aerial vehicle (UAV) swarms, so as to improve the positioning accuracy in electromagnetic degradation environment, enhance the relative constraint positioning capability of multiple UAVs, and ensure the positioning consistency of multiple UAVs in global spatiotemporal reference alignment under communication degradation scenario.

[0007] In a first aspect, the present invention provides a collaborative positioning method for a swarm of UAVs used for pole transportation, comprising the following steps: Step S1: Obtain the quality index of the Global Navigation Satellite System signal received by each node in the UAV swarm at the current moment, and determine the positioning environment mode of each node; Step S2: Activate the airborne radio ranging module based on the positioning environment mode, and acquire the inter-aircraft radio ranging signal with at least one neighboring node in a broadcast manner as a relative distance constraint; Step S3: In the airborne processing unit of each node, the relative distance constraint is used as a strong constraint condition to construct a joint positioning solution model with the carrier phase observation of the global navigation satellite system, constraining the search space of integer ambiguity, so as to output a centimeter-level distributed precise positioning solution; Step S4: Synchronously acquire the body data collected by the airborne inertial measurement unit and the load data collected by the distributed inertial sensing units deployed on the suspended pole structure, and perform pre-integration processing on the body data to obtain the platform pre-integration result; Step S5: Input the body data, load data and platform pre-integration results into a nonlinear state estimator to identify the rigid-flexible coupling dynamic parameters of the suspension system online, and calculate the dynamic spatial envelope representing the sweep range of the pole in real time. Step S6: Based on the distributed precise positioning solution and dynamic spatial envelope, generate a three-dimensional navigation track online that takes into account both obstacle avoidance and attitude stability margin; Step S7: When the quality of the inter-machine communication link is detected to be lower than the preset quality threshold, the distributed vision-radio cooperative positioning mode is activated, and each node obtains relative measurement information including the relative line-of-sight vector of the neighboring machine through the airborne vision sensor; Step S8: Each node uses the relative measurement information, the distance value extracted from the inter-machine radio ranging signal, and its own distributed precise positioning solution as factors to construct and solve a distributed pose graph optimization problem based on the factor graph, thereby obtaining a consistent collaborative positioning result with global spatiotemporal reference alignment among multiple machines.

[0008] Secondly, the present invention provides a collaborative positioning system for a swarm of UAVs used for pole transportation, comprising the following modules: The positioning environment mode determination module is used to obtain the quality indicators of the global navigation satellite system signals received by each node in the UAV swarm at the current moment, and to determine the positioning environment mode of each node. The inter-aircraft ranging activation and constraint acquisition module is used to activate the airborne radio ranging module based on the positioning environment mode, and acquire the inter-aircraft radio ranging signal with at least one neighboring node in a broadcast manner as a relative distance constraint. The tightly coupled positioning solution module is used in the airborne processing units of each node to construct a joint positioning solution model with the carrier phase observations of the global navigation satellite system, using the relative distance constraint as a strong constraint condition, to constrain the search space of integer ambiguity, so as to output a centimeter-level distributed precise positioning solution; The multi-source inertial data synchronization and pre-integration module is used to synchronously acquire the body data collected by the airborne inertial measurement unit and the load data collected by the inertial sensing units distributed on the suspended pole structure, and to perform pre-integration processing on the body data to obtain the platform pre-integration result. The parameter identification and spatial envelope calculation module is used to input the body data, load data and platform pre-integration results into a nonlinear state estimator, identify the rigid-flexible coupling dynamic parameters of the suspension system online, and calculate the dynamic spatial envelope representing the sweep range of the pole in real time. The dynamic obstacle avoidance trajectory planning module is used to generate a three-dimensional navigation trajectory online that takes into account both obstacle avoidance and attitude stability margin, based on the distributed precise positioning solution and dynamic spatial envelope. The same positioning mode startup module is used to start the distributed vision-radio cooperative positioning mode when the quality of the inter-machine communication link is detected to be lower than the preset quality threshold. Each node obtains relative measurement information including the relative line-of-sight vector of the neighboring machine through the airborne vision sensor. The cooperative localization and state estimation module is used by each node to construct and solve a distributed pose graph optimization problem based on a factor graph, using the relative measurement information, the distance value extracted from the inter-machine radio ranging signal, and its own distributed precise localization solution as factors, thereby obtaining a consistent cooperative localization result with global spatiotemporal reference alignment among multiple machines.

[0009] The advantages of this invention are: 1. First, under electromagnetic degradation, the airborne radio ranging module is automatically activated based on the quality index of the Global Navigation Satellite System (GNSS) signal. The acquired inter-aircraft radio ranging signal is introduced as a strong constraint into the joint positioning solution model, thereby effectively constraining the search space of integer ambiguity. Even under conditions of satellite signal attenuation or multipath propagation, it can still output centimeter-level distributed precise positioning solutions, significantly improving positioning accuracy. At the same time, the relative distance constraint itself enhances the relative constraint positioning capability between multiple aircraft, enabling each node to achieve distributed high-precision positioning even when the GNSS observation quality is poor. In addition, when the quality of the inter-aircraft communication link is lower than the quality threshold, it automatically switches to a distributed vision-radio cooperative positioning mode. The airborne vision sensor is used to obtain the relative line-of-sight vector of neighboring aircraft, and combined with the distance value extracted from the inter-aircraft radio ranging signal and its own distributed precise positioning solution, a distributed pose graph optimization problem based on factor graph is constructed. Thus, even in communication degradation scenarios, global spatiotemporal reference alignment between multiple aircraft can still be achieved, ensuring consistent cooperative positioning results.

[0010] 2. By acquiring the quality indicators of the Global Navigation Satellite System (GNSS) signals from each UAV and using a fuzzy logic inference system to determine the positioning environment mode in real time, the airborne radio ranging module is automatically activated in either a mild or severe electromagnetic degradation environment mode. High-confidence inter-UAV relative distance and angle of arrival information are obtained through a symmetrical bilateral two-way ranging protocol and an anti-multipath mode. This relative distance is then used as a strong constraint and tightly combined with carrier phase observations to construct an integer least squares problem with inequality constraints. A hierarchical ambiguity resolution strategy is adopted, first fixing the integer ambiguity of wide-lane / ultra-wide-lane combinations and then constraining the narrow-lane combination search. This significantly compresses the integer ambiguity search space under satellite signal attenuation and multipath interference, achieving centimeter-level distributed precise positioning solutions. Simultaneously, inter-UAV radio geometric constraints overcome the lack of phase in independent solutions based solely on the GNSS system. To address the issue of insufficient distributed accuracy caused by observations, when the quality of the inter-machine communication link is detected to be below a threshold or the differential correction signal of the global navigation satellite system reference station is interrupted, the system seamlessly switches to a distributed vision-radio cooperative positioning mode. Each node uses the relative line-of-sight vector obtained by the airborne vision sensor and the distance value extracted from the inter-machine radio ranging signal as a factor. Together with its own distributed precise positioning solution and the platform pre-integration results, it constructs and solves a distributed pose graph optimization problem based on the factor graph. It also introduces the load dynamics prediction factor edge of the suspended pole and the dynamic weighted adjustment for occlusion interference. In the process of minimizing the global reprojection error, ranging residual and load dynamics prediction error, it simultaneously completes the accurate estimation of the pose and load status of multiple machines, thereby achieving consistent cooperative positioning between multiple machines with global spatiotemporal reference alignment under conditions of communication degradation or no central link.

[0011] 3. Achieving highly reliable, centimeter-level distributed precision cooperative positioning in complex electromagnetic environments: By monitoring the quality indicators of the Global Navigation Satellite System (GNSS) signals and using a fuzzy logic reasoning system to determine the positioning environment mode, the inter-machine radio ranging module is activated in either a mild or severe electromagnetic degradation environment mode. The relative distance obtained from ultra-wideband ranging is used as a strong constraint and tightly combined with carrier phase observations to construct an integer least squares problem with inequality constraints. A hierarchical ambiguity resolution strategy is adopted. Partial integer ambiguity is quickly fixed using ultra-wide lane combination / wide lane combination, and then this strong distance constraint is introduced into the search ellipsoid. The integer ambiguity of narrow lane combination is determined by the constrained LAMBDA method. At the same time, the verification weighting matrix is ​​dynamically adjusted according to the measurement confidence of the inter-machine radio ranging signal, thereby effectively compressing the ambiguity search space and significantly improving the success rate of ambiguity fixation and computational efficiency. Even in environments with strong electromagnetic interference, such as power transmission corridors, each node can still independently output centimeter-level real-time distributed precision positioning solutions, ensuring accurate motion references during pole transportation.

[0012] 4. Real-time perception and prediction of the dynamic spatial envelope of the suspended power pole significantly improves flight safety margin in complex environments: By deploying inertial sensing units at the pole's gripping and locking mechanism and remote end, and synchronously collecting body data and load data with the airborne inertial measurement unit, the platform's pre-integration results are obtained through pre-integration processing. These results are then input into a nonlinear state estimator to identify rigid-flexible coupled dynamic parameters such as the swing frequency, damping ratio, and mode shape of the suspended system online. Based on these parameters and the pole's preset geometric model, the time-varying dynamic spatial envelope (i.e., dynamic spatial envelope) of the pole and its swing range in three-dimensional space is calculated in real time. During the trajectory planning stage, the uncertainty of the future time-domain swing phase prediction is superimposed to generate a probabilistic dynamic exclusion zone. This dynamic exclusion zone is fused with a static obstacle map to construct a dynamic cost map, enabling the UAV to actively plan a three-dimensional navigation trajectory that avoids external obstacles while maintaining attitude stability margin under the dual constraints of kinematics and swing dynamics. This effectively prevents the power pole from colliding with surrounding objects and ensures the safety of power construction and transportation.

[0013] 5. Capable of autonomous fault tolerance and seamless switching between cooperative positioning modes under multiple degradation conditions: It not only monitors the quality of inter-drone communication links, but also uses the interruption of differential correction signals from the global navigation satellite system reference station or the failure of the unified command link from the central dispatch console as a mode switching condition, ensuring that the distributed vision-radio cooperative positioning mode can be activated in real time under various centralized degradation scenarios. Each UAV acquires relative measurement information, including the relative line-of-sight vector of neighboring UAVs, through its onboard vision sensor, and together with the distance value extracted from the inter-drone radio ranging signal and its own distributed precise positioning solution, constructs a distributed pose graph optimization problem based on a factor graph, realizing decentralized multi-UAV pose global spatiotemporal reference alignment. This autonomous mode switching mechanism ensures that the system can still output consistent cooperative positioning results under extreme conditions such as loss of ground infrastructure or communication interruption, significantly enhancing the resilience and reliability of UAV swarms performing pole transfer tasks.

[0014] 6. Incorporating the dynamics of the suspended load into the distributed factor graph optimization framework enhances the consistency of multi-machine collaborative positioning and the ability to perceive the pole's state: In the distributed vision-radio collaborative positioning mode, each node not only utilizes the body pose node, platform pre-integration results, ranging factor edges, and visual observation factor edges, but also further introduces load state variable nodes that characterize the relative motion state of the pole. Based on the rigid-flexible coupling dynamic parameters, load dynamics prediction factor edges are constructed to constrain the swing and vibration process of the pole. Simultaneously, the probability of occlusion and non-line-of-sight interference caused by the pole to the inter-machine radio ranging signal and visual observation is evaluated in real time using the dynamic spatial envelope. Based on this, the information matrices of the ranging factor edges and visual observation factor edges are dynamically weighted and adjusted to effectively suppress the measurement degradation error caused by the dynamic occlusion of the suspended structure. Through distributed factor graph iterative optimization, while minimizing the global reprojection error, ranging residual, and load dynamics prediction error, consistent collaborative positioning results with global spatiotemporal reference alignment between multiple machines and the estimation of the pole's motion state are obtained simultaneously, realizing full-state, high-precision collaborative perception of the UAV swarm and the suspended load.

[0015] 7. The environmentally driven multi-mode anti-interference design significantly enhances the robustness of inter-machine relative measurements: By utilizing quality indicators such as signal-to-noise ratio, carrier-to-noise ratio, and multipath error estimates, the fuzzy logic reasoning system quantifies the level of environmental electromagnetic degradation and determines the ranging strategy accordingly. Especially in environments with severe electromagnetic degradation, the ultra-wideband module not only outputs information with confidence and angle of arrival calculated based on the antenna array, but also automatically switches to an anti-multipath mode using pulse position modulation and time-hopping spread spectrum technology to suppress dense multipath interference at the signal system level. This environmentally adaptive perception-ranging mechanism ensures that high-confidence relative distance and direction constraints can still be provided in power facility scenarios with severe multipath effects, providing a reliable measurement basis for subsequent integer ambiguity resolution and factor graph optimization under strong constraints. Attached Figure Description

[0016] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0017] Figure 1 This is a flowchart of a collaborative positioning method for a swarm of UAVs used for pole transportation according to the present invention.

[0018] Figure 2 This is a schematic diagram of the structure of a collaborative positioning system for a pole-transfer unmanned aerial vehicle swarm according to the present invention. Detailed Implementation

[0019] The overall approach of the technical solution in this application is as follows: Based on the quality of the global navigation satellite system signal and the inter-machine communication link, the cooperative positioning mode is adaptively switched; in an electromagnetic degradation environment, the ultra-wideband ranging module is activated to obtain high-confidence inter-machine relative distances, which are then used as strong constraints and tightly combined with carrier phase observations. A hierarchical ambiguity resolution strategy is employed to achieve centimeter-level distributed precise positioning; simultaneously, by synchronously collecting airborne inertial data and distributed inertial sensing data on the suspended pole, rigid-flexible coupling dynamic parameters are identified online, and the dynamic spatial envelope of the pole is calculated in real time to generate a three-dimensional track that balances obstacle avoidance and load stability; when inter-machine communication degrades, a seamless switch to a distributed vision-radio cooperative positioning mode is achieved. Relative line-of-sight vectors, distance information, and precise positioning solutions are used as factors to construct a factor graph optimization problem, and load dynamics prediction factors are introduced. This allows for consistent cooperative positioning and load state estimation with global spatiotemporal reference alignment across multiple drones even under communication constraints, systematically improving the positioning accuracy, operational safety, and cooperative consistency of the pole-transfer drone swarm in complex and harsh environments.

[0020] Please refer to Figures 1 to 2 As shown, a preferred embodiment of the collaborative positioning method for a swarm of UAVs used for pole transportation according to the present invention includes the following steps: Step S1: Obtain the quality index of the Global Navigation Satellite System signal received by each node in the UAV swarm at the current moment, and determine the positioning environment mode of each node; Step S2: Activate the airborne radio ranging module based on the positioning environment mode, and acquire the inter-aircraft radio ranging signal with at least one neighboring node in a broadcast manner as a relative distance constraint; Step S3: In the airborne processing unit of each node, the relative distance constraint is used as a strong constraint condition to construct a joint positioning solution model with the carrier phase observation of the global navigation satellite system, constraining the search space of integer ambiguity, so as to output a centimeter-level distributed precise positioning solution; Step S4: Synchronously acquire the body data collected by the airborne inertial measurement unit and the load data collected by the distributed inertial sensing units deployed on the suspended pole structure, and perform pre-integration processing on the body data to obtain the platform pre-integration result; Step S5: Input the body data, load data and platform pre-integration results into a nonlinear state estimator to identify the rigid-flexible coupling dynamic parameters of the suspension system online, and calculate the dynamic spatial envelope representing the sweep range of the pole in real time. Step S6: Based on the distributed precise positioning solution and dynamic spatial envelope, generate a three-dimensional navigation track online that takes into account both obstacle avoidance and attitude stability margin; Step S7: When the quality of the inter-machine communication link is detected to be lower than the preset quality threshold, the distributed vision-radio cooperative positioning mode is activated, and each node obtains relative measurement information including the relative line-of-sight vector of the neighboring machine through the airborne vision sensor; Simultaneously, each node acquires the angle of arrival information (azimuth and pitch) calculated by the ultra-wideband module, synchronizes it with the relative line-of-sight vector extracted by the airborne vision sensor, and generates a unified relative direction measurement quantity by selecting the best option or by weighted least squares fusion, which is included in the relative measurement information.

[0021] Step S8: Each node uses the relative measurement information, the distance value extracted from the inter-machine radio ranging signal, and its own distributed precise positioning solution as factors to construct and solve a distributed pose graph optimization problem based on the factor graph, thereby obtaining a consistent collaborative positioning result with global spatiotemporal reference alignment among multiple machines.

[0022] Step S1 specifically includes: Step S11: Each node acquires the numerical quality index of the Global Navigation Satellite System signal received at the current moment. The quality index includes at least the observed satellite frequency, signal-to-noise ratio, carrier-to-noise ratio, and multipath error estimate calculated by the receiver's autonomous integrity monitoring. Step S12: Input the quality index into a preset fuzzy logic reasoning system. The fuzzy logic reasoning system outputs a quantitative assessment level that characterizes the degree of degradation of the environmental electromagnetic environment based on the preset membership function and fuzzy rules. An example of the fuzzy logic reasoning system is configured as follows: 1) Fuzzification: Define membership functions for the input variables: satellite frequency, average signal-to-noise ratio (SNR), and maximum multipath error estimate. For example, the fuzzy set for satellite frequency is {few, medium, many}, where the membership function for "few" is a Z-shaped or trapezoidal function (1 for satellites ≤ 4, 0 for satellites ≥ 8). "Many" is the opposite. The fuzzy set for average SNR is {low, medium, high}. The fuzzy set for maximum multipath error estimate is {small, medium, large}.

[0023] 2) Fuzzy rule base: Define a set of "IF-THEN" rules, for example: Rule 1: IF satellite frequency IS high AND average signal-to-noise ratio IS high AND maximum multipath error IS small THEN environment mode IS open environment mode.

[0024] Rule 2: IF satellite frequency IS AND average signal-to-noise ratio IS THEN environmental mode IS mild electromagnetic degradation environmental mode.

[0025] Rule 3: IF satellite frequency IS low OR maximum multipath error IS large THEN environmental mode IS severe electromagnetic degradation environmental mode.

[0026] 3) Defuzzification: The output of the above rules is defuzzified using the centroid method to obtain a continuous quantitative evaluation level. Based on a preset threshold, this level is mapped to the final open environment mode, mild electromagnetic degradation environment mode, or severe electromagnetic degradation environment mode.

[0027] Step S13: Based on the quantitative evaluation level, determine the positioning environment mode of each node as an open environment mode, a mild electromagnetic degradation environment mode, or a severe electromagnetic degradation environment mode. The drone swarm includes at least one lead drone serving as a mobile reference station. When it is determined that there is a follower node in the swarm that is in the severe electromagnetic degradation environment mode, in step S3, the lead drone directly provides the joint positioning solution model of the follower node in the severe electromagnetic degradation environment mode with an absolute pose reference as an external strong reference constraint through its own high-precision real-time dynamic differential positioning result obtained based on carrier phase differential technology, so as to help it accelerate the convergence of integer ambiguity solution.

[0028] It should be noted that before implementation, the fuzzy logic reasoning system needs to be calibrated based on typical electromagnetic environments (such as open fields, urban canyons, mountainous areas, and areas near power transmission lines) for the expected operation of the UAV swarm. The calibration process includes: collecting a large number of Global Navigation Satellite System (GNSS) signal quality index samples under different typical environments; determining the membership function parameters of each input variable through expert experience or clustering algorithms (e.g., defining specific numerical ranges for "few," "medium," and "many" satellite frequencies); and optimizing the rule conclusion weights in the fuzzy rule base. This calibration process aims to match the quantitative assessment level with the actual degree of electromagnetic environment degradation, ensuring the accuracy and robustness of environmental pattern determination. This calibration is performed offline on the ground system; during online operation, the UAV nodes directly call the pre-calibrated membership functions and fuzzy rule base.

[0029] Step S2 specifically involves: When the positioning environment mode is determined to be a mild electromagnetic degradation environment mode or a severe electromagnetic degradation environment mode, the airborne radio ranging module is activated to acquire the inter-aircraft radio ranging signal with at least one neighboring node in a broadcast manner as a relative distance constraint. The airborne radio ranging module is an ultra-wideband module. The ultra-wideband module adopts a symmetrical bilateral two-way ranging protocol, obtains the distance value with the neighboring aircraft through two-way flight time measurement, and dynamically calculates the confidence level of the distance value based on the ratio of the initial diameter amplitude of the received signal to the total received power. The ultra-wideband module is also equipped with at least three non-collinearly arranged antenna arrays for calculating the angle of arrival information of the relative line-of-sight vector between the computers; when the positioning environment mode is a severely electromagnetically degraded environment mode, the ultra-wideband module automatically switches to anti-multipath mode. In the anti-multipath mode, multipath interference is suppressed by pulse position modulation and time-hopping spread spectrum technology, and the distance value and angle of arrival information with the confidence level are output. When the distance value and its confidence level are used as strong constraints in step S3, they are used to dynamically adjust the weighting matrix in the joint positioning solution model to constrain the search ellipsoid of integer ambiguity. After the angle of arrival information is activated in step S7, it replaces or is tightly coupled and fused with the relative line-of-sight vector output by the airborne visual sensor, and together they serve as components of the relative measurement information in step S8, so as to construct the visual observation factor edge and participate in factor graph optimization.

[0030] Here, we further explain the confidence level of dynamically calculating the distance value based on the ratio of the first path amplitude of the received signal to the total received power. This confidence level index aims to quantify the degree to which the current ranging value is affected by multipath effects. The physical principle is that under line-of-sight dominant channel conditions, the energy of the first path signal arriving at the receiver first accounts for a higher proportion of the total received energy; while under non-line-of-sight or multipath-dense conditions, the first path signal may be attenuated, while the energy proportion of subsequent reflection paths increases, leading to a decrease in this ratio. Therefore, this ratio can serve as a real-time, dynamic quality assessment factor; the higher the value, the more reliable the current distance measurement. In practical implementation, channel impulse response analysis can be integrated into the baseband processing unit of the ultra-wideband receiver to extract the peak amplitude of the first path and the total received power within a certain time window, calculate their normalized ratio, and map it to a confidence level range of 0 to 1.

[0031] Step S3 specifically involves: In the airborne processing unit of each node, the received signals from the Global Navigation Satellite System are used to construct the inter-satellite-inter-station double-difference carrier phase observation equation; The relative distance constraint from at least one neighboring node is used as a strong constraint and tightly combined with the carrier phase observation equation to form an integer least squares problem with inequality constraints. When this node is a follower node in a severely electromagnetically degraded environment mode, and receives a high-precision RTK absolute pose reference obtained based on carrier phase differential technology broadcast by the lead UAV via the inter-machine link, the absolute pose reference is constructed into a form like... The equality constraint equations are then extended to the integer least squares problem with inequality constraints as additional strong baseline constraints to further compress the search ellipsoid of integer ambiguities and accelerate ambiguity convergence. The absolute position vector provided for the navigation drone. This is the position vector to be estimated for this machine.

[0032] Initiate a hierarchical ambiguity resolution strategy, which specifically includes: First-level calculation: Based on the received signals from the Global Navigation Satellite System, a combined carrier phase observation with different wavelengths is constructed. The combined carrier phase observation includes a wide-lane combination and an ultra-wide-lane combination with wavelengths longer than the narrow-lane combination. The common-view satellite with the highest elevation angle between itself and the neighboring satellite that transmits the inter-satellite radio ranging signal is selected. The carrier phase observation of the ultra-wide-lane combination or the wide-lane combination is used to perform three-difference processing to directly calculate and fix the integer ambiguity of the corresponding combination. Specifically, the triple-difference processing refers to selecting a common-view satellite with the highest elevation angle, observed jointly by both the local unit A and its neighboring unit B (which transmits the inter-unit radio ranging signal), as the reference satellite. First, using carrier phase observations of ultra-wide lane combinations or wide lane combinations, double-difference observations are constructed between units (A and B) and between satellites (reference satellite and non-reference satellite). Then, the current epoch is further differiated from the value of these double-difference observations at previous epochs to obtain a triple-difference observation that eliminates the influence of receiver clock bias residuals, initial phase deviations, and other error terms, excluding the integer characteristics of integer ambiguity. Based on this triple-difference observation, the floating-point ambiguity of the corresponding ultra-wide lane or wide lane combination can be directly solved with high precision using least squares or Kalman filtering, and quickly rounded and fixed due to its extremely long equivalent wavelength.

[0033] Second-level solution: The integer ambiguity of the fixed wide lane combination or ultra-wide lane combination is taken as a known quantity, transformed and substituted back into the carrier phase observation equation. Using the constrained LAMBDA method, the candidate set of integer ambiguity of the original frequency point or narrow lane combination is searched and determined within the search ellipsoid constructed by the relative distance constraint. Verification and confirmation: For the integer ambiguity candidate set of the narrow alley combination, calculate the weighted sum of squares of the relative distance constraint residuals, and dynamically adjust the weighting matrix according to the measurement confidence of the inter-machine radio ranging signal. From this, select candidate combinations that meet the preset threshold Ratio test value and have an ambiguity fixing success rate higher than the preset confidence level, and use them as the final fixed original frequency point integer ambiguity. Based on the final fixed integer ambiguity of the original frequency point, a centimeter-level distributed precise positioning solution is output.

[0034] Specifically, the weighted sum of squares of the fused relative distance constraint residuals is used as the test statistic for the Ratio test, and its calculation method is defined by the following formula: ; in, The weighted sum of squares that incorporates the relative distance constraint residuals is the test statistic used for the Ratio test; The carrier phase double-difference observation residual vector is calculated by back-calculation after fixing the candidate integer ambiguity set; Let f(x) be the covariance matrix of the carrier phase double-difference observations, and let f(x) be its inverse matrix. This is the weight matrix of the carrier phase observations; Let the relative distance constraint residual vector be defined as follows: ,in It is a candidate location solution based on local machine A and neighboring machine B. , Calculated geometric distance, It is the distance value extracted from the inter-machine radio ranging signal; The fundamental weight matrix for the distance measurement is typically set to the variance of the distance measurement noise. The reciprocal of, that is γ∈[0,1] represents the dynamic confidence level of the inter-machine radio ranging signal, used to dynamically adjust the weights. In this embodiment, γ is preferably selected as γ. This makes the adjusted distance constraint weight matrix as When the confidence level γ is high, the weight of the constraint increases; when the confidence level is low, the weight is weakened to avoid introducing incorrect constraints.

[0035] When screening candidate combinations, calculate the second smallest... Value and minimum The ratio of the values ​​is used as the Ratio test value. The fixed solution is accepted only when the value is greater than a preset threshold (e.g., 3.0) and the ambiguity fixation success rate based on the covariance matrix estimation of the candidate set is higher than a preset confidence level (e.g., 99%).

[0036] Step S4 specifically involves: The system synchronously acquires body data, including body angular velocity and body acceleration, collected by the airborne inertial measurement unit, and load data, including load angular velocity and load acceleration, collected by the distributed inertial sensing units deployed on the suspended pole structure. The system performs time synchronization and spatial alignment on the body data and load data, and performs pre-integration processing on the body angular velocity and body acceleration to obtain the relative pose increment between the current frame and the previous frame as the platform pre-integration result. The inertial sensing unit includes at least one first inertial measurement subunit installed near the pole gripping and locking mechanism, and at least one second inertial measurement subunit installed at the far end or geometric center of the pole.

[0037] The time synchronization refers to using the system time of the onboard processing unit as a reference, utilizing the hardware timestamps attached to each sensor data packet, and employing interpolation or Kalman filter array alignment to precisely synchronize the data of the onboard IMU and each suspended load IMU to the same sampling time, thereby eliminating time deviations caused by sensor clock drift and data transmission delays.

[0038] The aforementioned spatial alignment refers to the process of converting and representing the load data, such as acceleration and angular velocity, collected by each inertial sensing unit distributed on the suspended pole structure in a unified coordinate system, such as the UAV's body coordinate system, using pre-calibrated extrinsic parameters (including the translation vector and rotation matrix of each load IMU relative to the center of mass of the UAV platform), so as to facilitate subsequent pre-integration and dynamic modeling.

[0039] Step S5 specifically involves: Establish a nonlinear state estimator that describes the coupling between the rigid body motion of the UAV platform and the oscillation motion of the pole; Using the aforementioned body data, load data, and platform pre-integration results as observations, an extended Kalman filter or unscented Kalman filter algorithm is employed to estimate the swing frequency, damping ratio, and mode shape of the pole relative to the UAV platform online, which serve as the rigid-flexible coupling dynamic parameters of the suspension system. Based on the rigid-flexible coupling dynamic parameters and the preset geometric model of the pole, the time-varying envelope occupied by the pole and its swing range in three-dimensional space is calculated in real time and used as the dynamic spatial envelope.

[0040] The rigid-flexible coupling dynamic parameters of the suspension system identified online by the nonlinear state estimator are further defined as follows: Oscillation frequency The natural frequency of the pole as it undergoes pendulum motion in the coordinate system of the UAV platform, measured in Hertz (Hz).

[0041] Damping ratio The ratio of the damping coefficient to the critical damping of the oscillating motion of the pole is a dimensionless quantity.

[0042] Modal vibration mode : A function describing the distribution of elastic deformation of the pole along its length direction r. In this embodiment, to balance computational efficiency and accuracy, the mode shape is... The online identification is a linear combination of a finite-order cantilever beam modes, mathematically expressed as: .in, The i-th order mode shape basis functions are preset. The coordinates of the i-th mode shape are obtained through online estimation using the nonlinear state estimator. Therefore, subsequent calculations can all be based on a defined set of parameters. To proceed.

[0043] Step S6 specifically includes: Step S61: Using the distributed precise positioning solution as the global pose reference at the current moment, receive the dynamic spatial envelope, and calculate the swing phase prediction uncertainty of the dynamic spatial envelope in the future preset time domain based on the swing frequency and damping ratio in the rigid-flexible coupling dynamic parameters. Generate a dynamic repulsion region with time-varying and probabilistic characteristics by superimposing the dynamic spatial envelope and the prediction uncertainty. Specifically, the current swing state estimate and its covariance matrix output by the nonlinear state estimator in step S5 are used. The linearized state transition matrix of the rigid-flexible coupling dynamic model Perform covariance propagation to obtain the predicted covariance of the swing angle at future time t. ,in Let be the process noise covariance. Radial expansion of the dynamic space envelope using the 3σ boundary of this covariance generates a dynamic repulsion region with probabilistic boundary properties.

[0044] Step S62: Merge the dynamic rejection zone with the pre-stored static environmental obstacle map to construct a dynamic cost map; Step S63: Under the conditions of satisfying the kinematic constraints of the UAV and the swing dynamics constraints of the suspension system, starting from the global pose reference, a set of candidate trajectory sequences is searched in the dynamic cost map by graph search or sampling search algorithm; Step S64: Construct a composite cost function, which includes at least: an obstacle avoidance cost term that maximizes the distance between the dynamic spatial envelope and the obstacle, a load stabilization cost term that minimizes the kinetic energy or amplitude of the pole swing, and a formation coordination cost term that maintains a preset geometric relationship between the dynamic spatial envelope of the machine and at least one neighboring machine in the formation. The load stabilization cost term directly corresponds to and implements the active oscillation suppression function. As an example, this cost term... It can be specifically modeled as a prediction time domain arrive Inside, the angular velocity of the pole relative to the suspension point The weighted quadratic integral of the angular displacement θ: ; Where α and β are positive weighting coefficients, and θ and The trajectory is predicted during tracking using the rigid-flexible coupled dynamic model identified in step S5. By minimizing this cost term, a three-dimensional navigation track with active pole sway suppression characteristics can be generated, ensuring a smooth and accurate transfer process.

[0045] Step S65: By minimizing the composite cost function, select an optimal trajectory point sequence from the candidate trajectory sequence and output it in real time to generate a three-dimensional navigation track that takes into account obstacle avoidance, attitude stability margin and active sway suppression online.

[0046] Specifically, the implementation of the formation coordination cost term in step S64 depends on the formation configuration definition of the UAVs. In practical applications, this configuration can be pre-set according to the geometry of the utility pole and the layout of the hoisting points. For example, in a typical scenario where a single utility pole is hoisted by two UAVs in the front and rear, this cost term can be defined as a potential function. When the dynamic spatial envelope of the two UAVs tends to disrupt the preset safety interval in the vertical and horizontal directions (e.g., the front-to-back distance is less than the length of the utility pole plus a safety margin), the value of this potential function increases sharply. By minimizing this potential function, the two UAVs can actively coordinate their speed and attitude during maneuvers such as turning, acceleration, and deceleration to maintain the geometric stability of the entire hoisting formation and prevent pole tearing or collision.

[0047] In step S7, the activation conditions for the distributed visual-radio cooperative positioning mode also include detecting the interruption of the differential correction signal of the global navigation satellite system reference station or the failure of the unified command link of the central dispatching station.

[0048] Step S8 specifically includes: Step S81: Each node uses its own distributed precise localization solution as the absolute pose constraint and the platform pre-integration result as the inter-frame relative motion constraint to construct the ontology pose graph nodes and edges. Step S82: Each node extracts the distance value from the inter-machine radio ranging signal and constructs a ranging factor edge connecting its own pose node and the corresponding neighboring pose node. Step S83: Each node performs feature association on the relative line-of-sight vector contained in the relative measurement information to construct a visual observation factor edge connecting its own pose node and the corresponding neighboring pose node; If this node also obtains the angle of arrival information corresponding to the neighboring machine, then in addition to the image reprojection error, an additional angle of arrival residual term is introduced into the visual observation factor edge to achieve tight-coupled pose constraint between vision and radio.

[0049] Step S84: Each node retrieves the rigid-flexible coupling dynamic parameters and dynamic spatial envelope, defines a load state variable node for each UAV to characterize the relative motion state of the pole, and constructs a kinematic prediction model describing the relative swing and vibration process of the pole based on the rigid-flexible coupling dynamic parameters. The kinematic prediction model is used as a load dynamic prediction factor edge connecting the body pose node and the load state variable node. The kinematic prediction model describing the relative swaying and vibration process of the pole is the core of constructing the load dynamics prediction factor. As an example, one specific implementation of this model is a sway prediction model based on a constant velocity model, whose discretized state transition equation is expressed as: ; in, for The load state variable node at any given time is specifically defined as a state vector, which at least includes the relative position of the pole's center of mass in the UAV's body coordinate system. ) and speed ( ),Right now ; It is the state transition matrix, which describes the free evolution of the load state over time; Is The motion excitation of the UAV platform, including platform acceleration and angular velocity, is derived from the pre-integration result by the pose node of the body at any time. It is a control input matrix that transforms the platform motion excitation into an influence on the load motion state. For time step.

[0050] In factor graph optimization, the above equation serves as the basis for the error function of the load dynamics prediction factor edge, and its prediction error is defined as follows: By minimizing the weighted norm of this error, the dynamic history of the load can be incorporated as a constraint into the global optimization.

[0051] Step S85: Each node calculates the probability of occlusion or non-line-of-sight interference to the inter-machine radio ranging signal and visual observation under the current estimated load pose based on the dynamic spatial envelope, and dynamically weights and adjusts the information matrix of the ranging factor edge and the visual observation factor edge accordingly. As one implementation method, for ranging signals, if the line segment connecting the phase centers of the local and neighboring antennas intersects with the dynamic spatial envelope of the pole, then the obstruction probability is determined. Otherwise, it is 0. For visual observation, the proportion of occluded pixels is calculated based on the relationship between the projection point of the neighboring camera on the image plane and the envelope projection area of ​​the pole. The weighting factor of the information matrix is ​​taken as... And multiply the corresponding factor side information matrix by This reduces the impact of occluded measurements during optimization.

[0052] Step S86: Each node uses a distributed iterative optimization algorithm, such as the consensus-based augmented Lagrange method or the distributed Gauss-Newton method, to jointly solve the factor graph composed of the ontology pose graph nodes and edges, ranging factor edges, visual observation factor edges, load state variable nodes, and load dynamics prediction factor edges. While minimizing the global reprojection error and ranging residual, the load dynamics prediction error is also minimized, thereby synchronously obtaining the consistent collaborative positioning results of global spatiotemporal reference alignment among multiple machines, as well as the motion state estimation of the suspended poles of each machine.

[0053] It needs further clarification that during the distributed iterative optimization process in step S86, each node only exchanges locally linearized factor graph information (e.g., the gradient of the local cost function or the marginal probability distribution constructed by the Shure complement) with adjacent nodes in the communication topology, rather than sharing the complete state vector. This distributed implementation not only reduces communication bandwidth requirements but also avoids system failures caused by single points of failure. When a new node joins or leaves the inter-machine communication network, the topology of the factor graph is dynamically adjusted, and the optimization algorithm can adaptively reconstruct the dependencies between adjacent nodes, achieving plug-and-play cooperative localization capabilities. This is crucial for situations where UAVs may withdraw or substitutes may enter during the mission.

[0054] The method further includes: Step S9: Based on the consistent cooperative positioning results and rigid-flexible coupling dynamic parameters, each node performs the following operations for any two machines in the formation that have a collision risk: Step S91: Call the dynamic spatial envelope of the local machine and the neighboring machine, and construct their respective load swing state prediction models based on the swing frequency and damping ratio in the rigid-flexible coupling dynamic parameters, predict a time domain ahead, and generate time-varying prediction envelopes of the local machine and the neighboring machine with probability boundary attributes in the future time period. Step S92: Superimpose the time-varying predicted envelopes of the local machine and the neighboring machine in a common spatiotemporal coordinate system, and dynamically calculate the shortest predicted intrusion distance between the two envelopes at each future time and its probability confidence. The probability confidence is obtained by joint propagation of the covariance matrix of the distributed precise positioning solution involved in the calculation, as well as the information matrices of the visual observation factor edge and the ranging factor edge. The joint propagation process of the probability confidence is specifically implemented as follows: First, obtain the covariance matrix or information matrix corresponding to each error source affecting the uncertainty of the relative pose estimation between the two machines. This includes the covariance matrix of the distributed precise positioning solution of the local machine and the neighboring machine. and These can be obtained from the solution process in step S3; and the information matrix of the visual observation factor side connecting the two machines. Information matrix of ranging factor sides .

[0055] Then, the covariance intersection (CI) algorithm is used to uniformly fuse these multi-source uncertainties to obtain a fused conservative covariance matrix. : ; in, Based on the covariance matrix of the local machine and its neighboring machines and The relative pose covariance matrix obtained from propagation; These are weighting coefficients used to adjust the level of trust in each information source. Finally, the fused covariance matrix is... The predicted intrusion distance is obtained by applying this probability confidence level during the calculation of the shortest predicted intrusion distance (e.g., through an unscented transformation). This probability characterizes the likelihood that the predicted intrusion distance is less than a certain security threshold.

[0056] Step S10: When the shortest predicted intrusion distance is considered unsafe under a preset probability-time joint threshold, a cooperative avoidance strategy optimizer is activated. The cooperative avoidance strategy optimizer generates collision avoidance commands by solving a constrained multi-objective optimization problem, which includes: Step S101: The main objective function minimizes the weighted sum of the predicted intrusion probability and a load disturbance suppression cost; the load disturbance suppression cost quantifies the excitation degree of the candidate obstacle avoidance maneuver route on the swing amplitude of the hoisting pole based on the load state variable node and the load dynamics prediction factor edge. Step S102: Hard constraints, including: the candidate maneuver trajectory must ensure that the time-varying prediction envelope of the two aircraft remains greater than a dynamic safety margin at any time, the dynamic safety margin being jointly determined by the confidence of the inter-aircraft radio ranging signal, the probability of visual observation occlusion, and the quality of the communication link. Step S103: Assign decision weights according to formation roles and communication status. If the drone is the lead drone, the collision avoidance command it generates is broadcast to the relevant follower nodes through the inter-drone link to perform cooperative path replanning. If it is a follower node in a communication degradation scenario, it will prioritize the generation of avoidance maneuver commands that move the drone away from the predicted intrusion area based on the relative measurement information obtained by itself through the airborne vision sensor and ultra-wideband module.

[0057] A preferred embodiment of the UAV swarm cooperative positioning system for pole transport according to the present invention includes the following modules: The positioning environment mode determination module is used to obtain the quality indicators of the global navigation satellite system signals received by each node in the UAV swarm at the current moment, and to determine the positioning environment mode of each node. The inter-aircraft ranging activation and constraint acquisition module is used to activate the airborne radio ranging module based on the positioning environment mode, and acquire the inter-aircraft radio ranging signal with at least one neighboring node in a broadcast manner as a relative distance constraint. The tightly coupled positioning solution module is used in the airborne processing units of each node to construct a joint positioning solution model with the carrier phase observations of the global navigation satellite system, using the relative distance constraint as a strong constraint condition, to constrain the search space of integer ambiguity, so as to output a centimeter-level distributed precise positioning solution; The multi-source inertial data synchronization and pre-integration module is used to synchronously acquire the body data collected by the airborne inertial measurement unit and the load data collected by the inertial sensing units distributed on the suspended pole structure, and to perform pre-integration processing on the body data to obtain the platform pre-integration result. The parameter identification and spatial envelope calculation module is used to input the body data, load data and platform pre-integration results into a nonlinear state estimator, identify the rigid-flexible coupling dynamic parameters of the suspension system online, and calculate the dynamic spatial envelope representing the sweep range of the pole in real time. The dynamic obstacle avoidance trajectory planning module is used to generate a three-dimensional navigation trajectory online that takes into account both obstacle avoidance and attitude stability margin, based on the distributed precise positioning solution and dynamic spatial envelope. The same positioning mode startup module is used to start the distributed vision-radio cooperative positioning mode when the quality of the inter-machine communication link is detected to be lower than the preset quality threshold. Each node obtains relative measurement information including the relative line-of-sight vector of the neighboring machine through the airborne vision sensor. Simultaneously, each node acquires the angle of arrival information (azimuth and pitch) calculated by the ultra-wideband module, synchronizes it with the relative line-of-sight vector extracted by the airborne vision sensor, and generates a unified relative direction measurement quantity by selecting the best option or by weighted least squares fusion, which is included in the relative measurement information.

[0058] The cooperative localization and state estimation module is used by each node to construct and solve a distributed pose graph optimization problem based on a factor graph, using the relative measurement information, the distance value extracted from the inter-machine radio ranging signal, and its own distributed precise localization solution as factors, thereby obtaining a consistent cooperative localization result with global spatiotemporal reference alignment among multiple machines.

[0059] The positioning environment mode determination module specifically includes: The GNSS quality index acquisition and quantification unit is used by each node to acquire the numerical quality index of the global navigation satellite system signal received at the current moment. The quality index includes at least the observed satellite frequency, signal-to-noise ratio, carrier-to-noise ratio, and multipath error estimate calculated by the receiver's autonomous integrity monitoring. The fuzzy logic reasoning evaluation unit is used to input the quality index into a preset fuzzy logic reasoning system. The fuzzy logic reasoning system outputs a quantitative evaluation level that characterizes the degree of degradation of the environmental electromagnetic environment based on the preset membership function and fuzzy rules. An example of the fuzzy logic reasoning system is configured as follows: 1) Fuzzification: Define membership functions for the input variables: satellite frequency, average signal-to-noise ratio (SNR), and maximum multipath error estimate. For example, the fuzzy set for satellite frequency is {few, medium, many}, where the membership function for "few" is a Z-shaped or trapezoidal function (1 for satellites ≤ 4, 0 for satellites ≥ 8). "Many" is the opposite. The fuzzy set for average SNR is {low, medium, high}. The fuzzy set for maximum multipath error estimate is {small, medium, large}.

[0060] 2) Fuzzy rule base: Define a set of "IF-THEN" rules, for example: Rule 1: IF satellite frequency IS high AND average signal-to-noise ratio IS high AND maximum multipath error IS small THEN environment mode IS open environment mode.

[0061] Rule 2: IF satellite frequency IS AND average signal-to-noise ratio IS THEN environmental mode IS mild electromagnetic degradation environmental mode.

[0062] Rule 3: IF satellite frequency IS low OR maximum multipath error IS large THEN environmental mode IS severe electromagnetic degradation environmental mode.

[0063] 3) Defuzzification: The output of the above rules is defuzzified using the centroid method to obtain a continuous quantitative evaluation level. Based on a preset threshold, this level is mapped to the final open environment mode, mild electromagnetic degradation environment mode, or severe electromagnetic degradation environment mode.

[0064] The environment mode determination output unit is used to determine the positioning environment mode of each node as an open environment mode, a mild electromagnetic degradation environment mode, or a severe electromagnetic degradation environment mode based on the quantitative evaluation level. The drone swarm includes at least one lead drone serving as a mobile reference station. When it is determined that there is a follower node in the swarm that is in the severe electromagnetic degradation environment mode, in step S3, the lead drone directly provides the joint positioning solution model of the follower node in the severe electromagnetic degradation environment mode with an absolute pose reference as an external strong reference constraint through its own high-precision real-time dynamic differential positioning result obtained based on carrier phase differential technology, so as to help it accelerate the convergence of integer ambiguity solution.

[0065] It should be noted that before implementation, the fuzzy logic reasoning system needs to be calibrated based on typical electromagnetic environments (such as open fields, urban canyons, mountainous areas, and areas near power transmission lines) for the expected operation of the UAV swarm. The calibration process includes: collecting a large number of Global Navigation Satellite System (GNSS) signal quality index samples under different typical environments; determining the membership function parameters of each input variable through expert experience or clustering algorithms (e.g., defining specific numerical ranges for "few," "medium," and "many" satellite frequencies); and optimizing the rule conclusion weights in the fuzzy rule base. This calibration process aims to match the quantitative assessment level with the actual degree of electromagnetic environment degradation, ensuring the accuracy and robustness of environmental pattern determination. This calibration is performed offline on the ground system; during online operation, the UAV nodes directly call the pre-calibrated membership functions and fuzzy rule base.

[0066] The inter-machine ranging activation and constraint acquisition module is specifically used for: When the positioning environment mode is determined to be a mild electromagnetic degradation environment mode or a severe electromagnetic degradation environment mode, the airborne radio ranging module is activated to acquire the inter-aircraft radio ranging signal with at least one neighboring node in a broadcast manner as a relative distance constraint. The airborne radio ranging module is an ultra-wideband module. The ultra-wideband module adopts a symmetrical bilateral two-way ranging protocol, obtains the distance value with the neighboring aircraft through two-way flight time measurement, and dynamically calculates the confidence level of the distance value based on the ratio of the initial diameter amplitude of the received signal to the total received power. The ultra-wideband module is also equipped with at least three non-collinearly arranged antenna arrays for calculating the angle of arrival information of the relative line-of-sight vector between the computers; when the positioning environment mode is a severely electromagnetically degraded environment mode, the ultra-wideband module automatically switches to anti-multipath mode. In the anti-multipath mode, multipath interference is suppressed by pulse position modulation and time-hopping spread spectrum technology, and the distance value and angle of arrival information with the confidence level are output. When the distance value and its confidence level are used as strong constraints in the compact combination positioning solution module, they are used to dynamically adjust the weighting matrix in the joint positioning solution model to constrain the search ellipsoid of integer ambiguity. After the distributed vision-radio cooperative positioning mode is activated by the same positioning mode activation module, the angle of arrival information replaces or is tightly coupled and fused with the relative line-of-sight vector output by the airborne vision sensor, and together they serve as components of the relative measurement information in the cooperative positioning and state estimation module, so as to construct the visual observation factor edge and participate in factor graph optimization.

[0067] Here, we further explain the confidence level of dynamically calculating the distance value based on the ratio of the first path amplitude of the received signal to the total received power. This confidence level index aims to quantify the degree to which the current ranging value is affected by multipath effects. The physical principle is that under line-of-sight dominant channel conditions, the energy of the first path signal arriving at the receiver first accounts for a higher proportion of the total received energy; while under non-line-of-sight or multipath-dense conditions, the first path signal may be attenuated, while the energy proportion of subsequent reflection paths increases, leading to a decrease in this ratio. Therefore, this ratio can serve as a real-time, dynamic quality assessment factor; the higher the value, the more reliable the current distance measurement. In practical implementation, channel impulse response analysis can be integrated into the baseband processing unit of the ultra-wideband receiver to extract the peak amplitude of the first path and the total received power within a certain time window, calculate their normalized ratio, and map it to a confidence level range of 0 to 1.

[0068] The compact combination positioning solution module is specifically used for: In the airborne processing unit of each node, the received signals from the Global Navigation Satellite System are used to construct the inter-satellite-inter-station double-difference carrier phase observation equation; The relative distance constraint from at least one neighboring node is used as a strong constraint and tightly combined with the carrier phase observation equation to form an integer least squares problem with inequality constraints. When this node is a follower node in a severely electromagnetically degraded environment mode, and receives a high-precision RTK absolute pose reference obtained based on carrier phase differential technology broadcast by the lead UAV via the inter-machine link, the absolute pose reference is constructed into a form like... The equality constraint equations are then extended to the integer least squares problem with inequality constraints as additional strong baseline constraints to further compress the search ellipsoid of integer ambiguities and accelerate ambiguity convergence. The absolute position vector provided for the navigation drone. This is the position vector to be estimated for this machine.

[0069] Initiate a hierarchical ambiguity resolution strategy, which specifically includes: First-level calculation: Based on the received signals from the Global Navigation Satellite System, a combined carrier phase observation with different wavelengths is constructed. The combined carrier phase observation includes a wide-lane combination and an ultra-wide-lane combination with wavelengths longer than the narrow-lane combination. The common-view satellite with the highest elevation angle between itself and the neighboring satellite that transmits the inter-satellite radio ranging signal is selected. The carrier phase observation of the ultra-wide-lane combination or the wide-lane combination is used to perform three-difference processing to directly calculate and fix the integer ambiguity of the corresponding combination. Specifically, the triple-difference processing refers to selecting a common-view satellite with the highest elevation angle, observed jointly by both the local unit A and its neighboring unit B (which transmits the inter-unit radio ranging signal), as the reference satellite. First, using carrier phase observations of ultra-wide lane combinations or wide lane combinations, double-difference observations are constructed between units (A and B) and between satellites (reference satellite and non-reference satellite). Then, the current epoch is further differiated from the value of these double-difference observations at previous epochs to obtain a triple-difference observation that eliminates the influence of receiver clock bias residuals, initial phase deviations, and other error terms, excluding the integer characteristics of integer ambiguity. Based on this triple-difference observation, the floating-point ambiguity of the corresponding ultra-wide lane or wide lane combination can be directly solved with high precision using least squares or Kalman filtering, and quickly rounded and fixed due to its extremely long equivalent wavelength.

[0070] Second-level solution: The integer ambiguity of the fixed wide lane combination or ultra-wide lane combination is taken as a known quantity, transformed and substituted back into the carrier phase observation equation. Using the constrained LAMBDA method, the candidate set of integer ambiguity of the original frequency point or narrow lane combination is searched and determined within the search ellipsoid constructed by the relative distance constraint. Verification and confirmation: For the integer ambiguity candidate set of the narrow alley combination, calculate the weighted sum of squares of the relative distance constraint residuals, and dynamically adjust the weighting matrix according to the measurement confidence of the inter-machine radio ranging signal. From this, select candidate combinations that meet the preset threshold Ratio test value and have an ambiguity fixing success rate higher than the preset confidence level, and use them as the final fixed original frequency point integer ambiguity. Based on the final fixed integer ambiguity of the original frequency point, a centimeter-level distributed precise positioning solution is output.

[0071] Specifically, the weighted sum of squares of the fused relative distance constraint residuals is used as the test statistic for the Ratio test, and its calculation method is defined by the following formula: ; in, The weighted sum of squares that incorporates the relative distance constraint residuals is the test statistic used for the Ratio test; The carrier phase double-difference observation residual vector is calculated by back-calculation after fixing the candidate integer ambiguity set; Let f(x) be the covariance matrix of the carrier phase double-difference observations, and let f(x) be its inverse matrix. This is the weight matrix of the carrier phase observations; Let the relative distance constraint residual vector be defined as follows: ,in It is a candidate location solution based on local machine A and neighboring machine B. , Calculated geometric distance, It is the distance value extracted from the inter-machine radio ranging signal; The fundamental weight matrix for the distance measurement is typically set to the variance of the distance measurement noise. The reciprocal of, that is γ∈[0,1] represents the dynamic confidence level of the inter-machine radio ranging signal, used to dynamically adjust the weights. In this embodiment, γ is preferably selected as γ. This makes the adjusted distance constraint weight matrix as When the confidence level γ is high, the weight of the constraint increases; when the confidence level is low, the weight is weakened to avoid introducing incorrect constraints.

[0072] When screening candidate combinations, calculate the second smallest... Value and minimum The ratio of the values ​​is used as the Ratio test value. The fixed solution is accepted only when the value is greater than a preset threshold (e.g., 3.0) and the ambiguity fixation success rate based on the covariance matrix estimation of the candidate set is higher than a preset confidence level (e.g., 99%).

[0073] The multi-source inertial data synchronization and pre-integration module is specifically used for: The system synchronously acquires body data, including body angular velocity and body acceleration, collected by the airborne inertial measurement unit, and load data, including load angular velocity and load acceleration, collected by the distributed inertial sensing units deployed on the suspended pole structure. The system performs time synchronization and spatial alignment on the body data and load data, and performs pre-integration processing on the body angular velocity and body acceleration to obtain the relative pose increment between the current frame and the previous frame as the platform pre-integration result. The inertial sensing unit includes at least one first inertial measurement subunit installed near the pole gripping and locking mechanism, and at least one second inertial measurement subunit installed at the far end or geometric center of the pole.

[0074] The time synchronization refers to using the system time of the onboard processing unit as a reference, utilizing the hardware timestamps attached to each sensor data packet, and employing interpolation or Kalman filter array alignment to precisely synchronize the data of the onboard IMU and each suspended load IMU to the same sampling time, thereby eliminating time deviations caused by sensor clock drift and data transmission delays.

[0075] The aforementioned spatial alignment refers to the process of converting and representing the load data, such as acceleration and angular velocity, collected by each inertial sensing unit distributed on the suspended pole structure in a unified coordinate system, such as the UAV's body coordinate system, using pre-calibrated extrinsic parameters (including the translation vector and rotation matrix of each load IMU relative to the center of mass of the UAV platform), so as to facilitate subsequent pre-integration and dynamic modeling.

[0076] The parameter identification and spatial envelope calculation module is specifically used for: Establish a nonlinear state estimator that describes the coupling between the rigid body motion of the UAV platform and the oscillation motion of the pole; Using the aforementioned body data, load data, and platform pre-integration results as observations, an extended Kalman filter or unscented Kalman filter algorithm is employed to estimate the swing frequency, damping ratio, and mode shape of the pole relative to the UAV platform online, which serve as the rigid-flexible coupling dynamic parameters of the suspension system. Based on the rigid-flexible coupling dynamic parameters and the preset geometric model of the pole, the time-varying envelope occupied by the pole and its swing range in three-dimensional space is calculated in real time and used as the dynamic spatial envelope.

[0077] The rigid-flexible coupling dynamic parameters of the suspension system identified online by the nonlinear state estimator are further defined as follows: Oscillation frequency The natural frequency of the pole as it undergoes pendulum motion in the coordinate system of the UAV platform, measured in Hertz (Hz).

[0078] Damping ratio The ratio of the damping coefficient to the critical damping of the oscillating motion of the pole is a dimensionless quantity.

[0079] Modal vibration mode : A function describing the distribution of elastic deformation of the pole along its length direction r. In this embodiment, to balance computational efficiency and accuracy, the mode shape is... The online identification is a linear combination of a finite-order cantilever beam modes, mathematically expressed as: .in, The i-th order mode shape basis functions are preset. The coordinates of the i-th mode shape are obtained through online estimation using the nonlinear state estimator. Therefore, subsequent calculations can all be based on a defined set of parameters. To proceed.

[0080] The dynamic obstacle avoidance trajectory planning module specifically includes: The dynamic rejection region generation unit is used to receive the dynamic spatial envelope, using the distributed precise positioning solution as the global pose reference at the current moment, and to calculate the swing phase prediction uncertainty of the dynamic spatial envelope in the future preset time domain based on the swing frequency and damping ratio in the rigid-flexible coupling dynamic parameters. The result of superimposing the dynamic spatial envelope and the prediction uncertainty generates a dynamic rejection region with time-varying and probabilistic characteristics. Specifically, the current oscillation state estimate and its covariance matrix are obtained from the output of the nonlinear state estimator in the parameter identification and spatial envelope calculation module. The linearized state transition matrix of the rigid-flexible coupling dynamic model Perform covariance propagation to obtain the predicted covariance of the swing angle at future time t. ,in Let be the process noise covariance. Radial expansion of the dynamic space envelope using the 3σ boundary of this covariance generates a dynamic repulsion region with probabilistic boundary properties.

[0081] The dynamic cost map construction unit is used to merge the dynamic exclusion zone with a pre-stored static environmental obstacle map to construct a dynamic cost map. The candidate trajectory sequence search unit is used to find a set of candidate trajectory sequences in the dynamic cost map, starting from the global pose reference, under the conditions of satisfying the kinematic constraints of the UAV and the swing dynamics constraints of the suspension system, through graph search or sampling search algorithms. A composite cost function construction unit is used to construct a composite cost function, which includes at least: an obstacle avoidance cost term that maximizes the distance between the dynamic spatial envelope and the obstacle, a load stability cost term that minimizes the kinetic energy or amplitude of the pole swing, and a formation coordination cost term that maintains a preset geometric relationship between the dynamic spatial envelope of the machine and at least one neighboring machine in the formation. The load stabilization cost term directly corresponds to and implements the active oscillation suppression function. As an example, this cost term... It can be specifically modeled as a prediction time domain arrive Inside, the angular velocity of the pole relative to the suspension point The weighted quadratic integral of the angular displacement θ: ; Where α and β are positive weighting coefficients, and θ and The trajectory is predicted during tracking by the rigid-flexible coupled dynamic model identified in the parameter identification and spatial envelope calculation module. By minimizing this cost term, a three-dimensional navigation track with active pole sway suppression characteristics can be generated, ensuring the smoothness and accuracy of the transfer process.

[0082] The optimal trajectory selection and output unit is used to select an optimal trajectory point sequence from the candidate trajectory sequence by minimizing the composite cost function, and output it in real time to generate a three-dimensional navigation trajectory that takes into account obstacle avoidance, attitude stability margin and active sway suppression online.

[0083] Specifically, the implementation of the formation coordination cost term in the composite cost function construction unit depends on the UAV formation configuration definition. In practical applications, this configuration can be pre-set according to the geometry of the pole and the layout of the hoisting points. For example, in a typical scenario where a single pole is hoisted by two UAVs in the front and rear, this cost term can be defined as a potential function. When the dynamic spatial envelope of the two UAVs tends to violate the preset safety interval in the vertical and horizontal directions (e.g., the front-to-back distance is less than the pole length plus the safety margin), the value of this potential function increases sharply. By minimizing this potential function, the two UAVs can actively coordinate their speed and attitude during maneuvers such as turning, acceleration, and deceleration to maintain the geometric stability of the entire hoisting formation and prevent pole tearing or collision.

[0084] In the co-positioning mode activation module, the activation conditions for the distributed visual-radio cooperative positioning mode also include detecting the interruption of the differential correction signal of the global navigation satellite system reference station or the failure of the unified command link of the central dispatch console.

[0085] The cooperative localization and state estimation module specifically includes: The ontology pose graph construction unit is used to construct ontology pose graph nodes and edges by using each node as its own distributed precise localization solution as absolute pose constraint and the platform pre-integration result as inter-frame relative motion constraint. The ranging factor edge construction unit is used by each node to construct a ranging factor edge connecting its own pose node and the corresponding neighboring pose node by the distance value extracted from the inter-machine radio ranging signal. The visual observation factor edge construction unit is used for each node to perform feature association on the relative line-of-sight vector contained in the relative measurement information, and construct a visual observation factor edge connecting its own pose node and the corresponding neighbor pose node. If this node also obtains the angle of arrival information corresponding to the neighboring machine, then in addition to the image reprojection error, an additional angle of arrival residual term is introduced into the visual observation factor edge to achieve tight-coupled pose constraint between vision and radio.

[0086] The load dynamics prediction factor edge construction unit is used for each node to retrieve the rigid-flexible coupling dynamic parameters and dynamic space envelope, define a load state variable node for each UAV to characterize the relative motion state of the pole, and construct a kinematic prediction model describing the relative swing and vibration process of the pole based on the rigid-flexible coupling dynamic parameters. The kinematic prediction model is used as the load dynamics prediction factor edge connecting the body pose node and the load state variable node. The kinematic prediction model describing the relative swaying and vibration process of the pole is the core of constructing the load dynamics prediction factor. As an example, one specific implementation of this model is a sway prediction model based on a constant velocity model, whose discretized state transition equation is expressed as: ; in, for The load state variable node at any given time is specifically defined as a state vector, which at least includes the relative position of the pole's center of mass in the UAV's body coordinate system. ) and speed ( ),Right now ; It is the state transition matrix, which describes the free evolution of the load state over time; Is The motion excitation of the UAV platform, including platform acceleration and angular velocity, is derived from the pre-integration result by the pose node of the body at any time. It is a control input matrix that transforms the platform motion excitation into an influence on the load motion state. For time step.

[0087] In factor graph optimization, the above equation serves as the basis for the error function of the load dynamics prediction factor edge, and its prediction error is defined as follows: By minimizing the weighted norm of this error, the dynamic history of the load can be incorporated as a constraint into the global optimization.

[0088] The non-line-of-sight interference adaptive weighting unit is used by each node to calculate the probability of occlusion or non-line-of-sight interference to the inter-machine radio ranging signal and visual observation under the current estimated load pose based on the dynamic spatial envelope, and to dynamically adjust the information matrix of the ranging factor side and the visual observation factor side accordingly. As one implementation method, for ranging signals, if the line segment connecting the phase centers of the local and neighboring antennas intersects with the dynamic spatial envelope of the pole, then the obstruction probability is determined. Otherwise, it is 0. For visual observation, the proportion of occluded pixels is calculated based on the relationship between the projection point of the neighboring camera on the image plane and the envelope projection area of ​​the pole. The weighting factor of the information matrix is ​​taken as... And multiply the corresponding factor side information matrix by This reduces the impact of occluded measurements during optimization.

[0089] The distributed factor graph iterative optimization and result output unit is used by each node to jointly solve the factor graph composed of the ontology pose graph nodes and edges, ranging factor edges, visual observation factor edges, load state variable nodes and load dynamics prediction factor edges through distributed iterative optimization algorithms, such as consensus-based augmented Lagrange method or distributed Gauss-Newton method. This minimizes the global reprojection error and ranging residual while minimizing the load dynamics prediction error, thereby synchronously obtaining consistent collaborative positioning results of global spatiotemporal reference alignment among multiple machines, as well as motion state estimation of the suspended poles of each machine.

[0090] It needs further clarification that during the distributed iterative optimization of the distributed factor graph and the distributed iterative optimization of the result output unit, each node only exchanges locally linearized factor graph information (e.g., the gradient of the local cost function or the marginal probability distribution constructed by the Shure complement) with adjacent nodes in the communication topology, rather than sharing the complete state vector. This distributed implementation not only reduces communication bandwidth requirements but also avoids system failures caused by single points of failure. When a new node joins or leaves the inter-machine communication network, the topology of the factor graph is dynamically adjusted, and the optimization algorithm can adaptively reconstruct the dependencies between adjacent nodes, achieving plug-and-play cooperative localization capabilities. This is crucial for situations where UAVs may withdraw or substitutes may enter during the mission.

[0091] The system also includes: The collision risk prediction and assessment module is used by each node to perform the following operations for any two machines in the formation that have a collision risk, based on the consistent collaborative positioning results and rigid-flexible coupling dynamic parameters: The time-varying prediction envelope generation unit is used to call the dynamic spatial envelope of the local machine and the neighboring machine, and construct their respective load swing state prediction models based on the swing frequency and damping ratio in the rigid-flexible coupling dynamic parameters, predict a time domain ahead, and generate time-varying prediction envelopes of the local machine and the neighboring machine with probability boundary attributes in the future time period. The intrusion distance and confidence calculation unit is used to superimpose the time-varying predicted envelopes of the local machine and the neighboring machine in a common spatiotemporal coordinate system, and dynamically calculate the shortest predicted intrusion distance and its probability confidence between the two envelopes at future times. The probability confidence is obtained by joint propagation of the covariance matrix of the distributed precise positioning solution involved in the calculation, as well as the information matrices of the visual observation factor edge and the ranging factor edge. The joint propagation process of the probability confidence is specifically implemented as follows: First, obtain the covariance matrix or information matrix corresponding to each error source affecting the uncertainty of the relative pose estimation between the two machines. This includes the covariance matrix of the distributed precise positioning solution of the local machine and the neighboring machine. and These can be obtained from the solution process in step S3; and the information matrix of the visual observation factor side connecting the two machines. Information matrix of ranging factor sides .

[0092] Then, the covariance intersection (CI) algorithm is used to uniformly fuse these multi-source uncertainties to obtain a fused conservative covariance matrix. : ; in, Based on the covariance matrix of the local machine and its neighboring machines and The relative pose covariance matrix obtained from propagation; These are weighting coefficients used to adjust the level of trust in each information source. Finally, the fused covariance matrix is... The predicted intrusion distance is obtained by applying this probability confidence level during the calculation of the shortest predicted intrusion distance (e.g., through an unscented transformation). This probability characterizes the likelihood that the predicted intrusion distance is less than a certain security threshold.

[0093] A collaborative avoidance strategy optimization module is used to activate a collaborative avoidance strategy optimizer when the shortest predicted intrusion distance is considered unsafe under a preset probability-time joint threshold. The collaborative avoidance strategy optimizer generates collision avoidance commands by solving a constrained multi-objective optimization problem, which includes: The main objective function construction unit is used for the main objective function, which minimizes the weighted sum of the predicted intrusion probability and a load disturbance suppression cost; the load disturbance suppression cost quantifies the excitation degree of the candidate obstacle avoidance maneuver route on the swing amplitude of the hoisting pole based on the load state variable node and the load dynamics prediction factor edge. The hard constraint definition unit is used for hard constraint conditions, including: the candidate maneuver trajectory must ensure that the time-varying prediction envelope of the two aircraft remains greater than a dynamic safety margin at any time, and the dynamic safety margin is jointly determined by the confidence of the inter-aircraft radio ranging signal, the probability of visual observation occlusion and the quality of the communication link. The decision weight allocation unit is used to allocate decision weights according to formation roles and communication status. If the drone is the lead drone, the collision avoidance command it generates is broadcast to the relevant follower nodes through the inter-drone link to perform cooperative path replanning. If it is a follower node in a communication degradation scenario, it will prioritize the relative measurement information obtained by itself through the airborne vision sensor and ultra-wideband module to independently generate an avoidance maneuver command that moves the drone away from the predicted intrusion area.

[0094] While specific embodiments of the present invention have been described above, those skilled in the art should understand that the specific embodiments described are merely illustrative and not intended to limit the scope of the present invention. Equivalent modifications and variations made by those skilled in the art in accordance with the spirit of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A collaborative positioning method for a swarm of UAVs used for pole transportation, characterized in that: Includes the following steps: Step S1: Obtain the quality index of the Global Navigation Satellite System signal received by each node in the UAV swarm at the current moment, and determine the positioning environment mode of each node; Step S2: Activate the airborne radio ranging module based on the positioning environment mode, and acquire the inter-aircraft radio ranging signal with at least one neighboring node in a broadcast manner as a relative distance constraint; Step S3: In the airborne processing unit of each node, the relative distance constraint is used as a strong constraint condition to construct a joint positioning solution model with the carrier phase observation of the global navigation satellite system, constraining the search space of integer ambiguity, so as to output a centimeter-level distributed precise positioning solution; Step S4: Synchronously acquire the body data collected by the airborne inertial measurement unit and the load data collected by the distributed inertial sensing units deployed on the suspended pole structure, and perform pre-integration processing on the body data to obtain the platform pre-integration result; Step S5: Input the body data, load data and platform pre-integration results into a nonlinear state estimator to identify the rigid-flexible coupling dynamic parameters of the suspension system online, and calculate the dynamic spatial envelope representing the sweep range of the pole in real time. Step S6: Based on the distributed precise positioning solution and dynamic spatial envelope, generate a three-dimensional navigation track online that takes into account both obstacle avoidance and attitude stability margin; Step S7: When the quality of the inter-machine communication link is detected to be lower than the preset quality threshold, the distributed vision-radio cooperative positioning mode is activated, and each node obtains relative measurement information including the relative line-of-sight vector of the neighboring machine through the airborne vision sensor; Step S8: Each node uses the relative measurement information, the distance value extracted from the inter-machine radio ranging signal, and its own distributed precise positioning solution as factors to construct and solve a distributed pose graph optimization problem based on the factor graph, thereby obtaining a consistent collaborative positioning result with global spatiotemporal reference alignment among multiple machines.

2. The method for collaborative positioning of a swarm of UAVs for pole transportation as described in claim 1, characterized in that: Step S1 specifically includes: Step S11: Each node acquires the numerical quality index of the Global Navigation Satellite System signal received at the current moment. The quality index includes at least the observed satellite frequency, signal-to-noise ratio, carrier-to-noise ratio, and multipath error estimate calculated by the receiver's autonomous integrity monitoring. Step S12: Input the quality index into a preset fuzzy logic reasoning system. The fuzzy logic reasoning system outputs a quantitative assessment level that characterizes the degree of degradation of the environmental electromagnetic environment based on the preset membership function and fuzzy rules. Step S13: Based on the quantitative evaluation level, determine the positioning environment mode of each node as an open environment mode, a mild electromagnetic degradation environment mode, or a severe electromagnetic degradation environment mode.

3. The method for collaborative positioning of a swarm of UAVs for pole transportation as described in claim 1, characterized in that: Step S2 specifically involves: When the positioning environment mode is determined to be a mild electromagnetic degradation environment mode or a severe electromagnetic degradation environment mode, the airborne radio ranging module is activated to acquire the inter-aircraft radio ranging signal with at least one neighboring node in a broadcast manner as a relative distance constraint. The airborne radio ranging module is an ultra-wideband module. The ultra-wideband module adopts a symmetrical bilateral two-way ranging protocol, obtains the distance value with the neighboring aircraft through two-way flight time measurement, and dynamically calculates the confidence level of the distance value based on the ratio of the initial diameter amplitude of the received signal to the total received power. The ultra-wideband module is also equipped with at least three non-collinearly arranged antenna arrays for solving the angle of arrival information of the relative line-of-sight vector between the computers; When the positioning environment mode is a severely electromagnetically degraded environment mode, the ultra-wideband module automatically switches to anti-multipath mode; In the anti-multipath mode, multipath interference is suppressed by pulse position modulation and time-hopping spread spectrum techniques, and distance value and angle of arrival information with the confidence level are output.

4. The method for collaborative positioning of a swarm of UAVs for pole transportation as described in claim 1, characterized in that: Step S3 specifically involves: In the airborne processing unit of each node, the received signals from the Global Navigation Satellite System are used to construct the inter-satellite-inter-station double-difference carrier phase observation equation; The relative distance constraint from at least one neighboring node is used as a strong constraint and tightly combined with the carrier phase observation equation to form an integer least squares problem with inequality constraints. Initiate a hierarchical ambiguity resolution strategy, which specifically includes: First-level calculation: Based on the received signals from the Global Navigation Satellite System, a combined carrier phase observation with different wavelengths is constructed. The combined carrier phase observation includes a wide-lane combination and an ultra-wide-lane combination with wavelengths longer than the narrow-lane combination. The common-view satellite with the highest elevation angle between itself and the neighboring satellite that transmits the inter-satellite radio ranging signal is selected. The carrier phase observation of the ultra-wide-lane combination or the wide-lane combination is used to perform three-difference processing to directly calculate and fix the integer ambiguity of the corresponding combination. Second-level solution: The integer ambiguity of the fixed wide lane combination or ultra-wide lane combination is taken as a known quantity, transformed and substituted back into the carrier phase observation equation. Using the constrained LAMBDA method, the candidate set of integer ambiguity of the original frequency point or narrow lane combination is searched and determined within the search ellipsoid constructed by the relative distance constraint. Verification and confirmation: For the integer ambiguity candidate set of the narrow alley combination, calculate the weighted sum of squares of the relative distance constraint residuals, and dynamically adjust the weighting matrix according to the measurement confidence of the inter-machine radio ranging signal. From this, select candidate combinations that meet the preset threshold Ratio test value and have an ambiguity fixing success rate higher than the preset confidence level, and use them as the final fixed original frequency point integer ambiguity. Based on the final fixed integer ambiguity of the original frequency point, a centimeter-level distributed precise positioning solution is output.

5. The method for collaborative positioning of a swarm of UAVs for pole transportation as described in claim 1, characterized in that: Step S4 specifically involves: The system synchronously acquires body data, including body angular velocity and body acceleration, collected by the airborne inertial measurement unit, and load data, including load angular velocity and load acceleration, collected by the distributed inertial sensing units deployed on the suspended pole structure. The system performs time synchronization and spatial alignment on the body data and load data, and performs pre-integration processing on the body angular velocity and body acceleration to obtain the relative pose increment between the current frame and the previous frame as the platform pre-integration result. The inertial sensing unit includes at least one first inertial measurement subunit installed near the pole gripping and locking mechanism, and at least one second inertial measurement subunit installed at the far end or geometric center of the pole.

6. The method for collaborative positioning of a swarm of UAVs for pole transportation as described in claim 1, characterized in that: Step S5 specifically involves: Establish a nonlinear state estimator that describes the coupling between the rigid body motion of the UAV platform and the oscillation motion of the pole; Using the aforementioned body data, load data, and platform pre-integration results as observations, an extended Kalman filter or unscented Kalman filter algorithm is employed to estimate the swing frequency, damping ratio, and mode shape of the pole relative to the UAV platform online, which serve as the rigid-flexible coupling dynamic parameters of the suspension system. Based on the rigid-flexible coupling dynamic parameters and the preset geometric model of the pole, the time-varying envelope occupied by the pole and its swing range in three-dimensional space is calculated in real time and used as the dynamic spatial envelope.

7. The method for collaborative positioning of a swarm of UAVs for pole transportation as described in claim 1, characterized in that: Step S6 specifically includes: Step S61: Using the distributed precise positioning solution as the global pose reference at the current moment, receive the dynamic spatial envelope, and calculate the swing phase prediction uncertainty of the dynamic spatial envelope in the future preset time domain based on the swing frequency and damping ratio in the rigid-flexible coupling dynamic parameters. Generate a dynamic repulsion region with time-varying and probabilistic characteristics by superimposing the dynamic spatial envelope and the prediction uncertainty. Step S62: Merge the dynamic rejection zone with the pre-stored static environmental obstacle map to construct a dynamic cost map; Step S63: Under the conditions of satisfying the kinematic constraints of the UAV and the swing dynamics constraints of the suspension system, starting from the global pose reference, a set of candidate trajectory sequences is searched in the dynamic cost map by graph search or sampling search algorithm; Step S64: Construct a composite cost function, which includes at least: an obstacle avoidance cost term that maximizes the distance between the dynamic spatial envelope and the obstacle, a load stabilization cost term that minimizes the kinetic energy or amplitude of the pole swing, and a formation coordination cost term that maintains a preset geometric relationship between the dynamic spatial envelope of the machine and at least one neighboring machine in the formation. Step S65: By minimizing the composite cost function, select an optimal trajectory point sequence from the candidate trajectory sequence and output it in real time to generate a three-dimensional navigation track that takes into account obstacle avoidance, attitude stability margin and active sway suppression online.

8. The method for collaborative positioning of a swarm of UAVs for pole transportation as described in claim 1, characterized in that: In step S7, the activation conditions for the distributed visual-radio cooperative positioning mode also include detecting the interruption of the differential correction signal of the global navigation satellite system reference station or the failure of the unified command link of the central dispatching station.

9. The method for collaborative positioning of a swarm of UAVs for pole transportation as described in claim 1, characterized in that: Step S8 specifically includes: Step S81: Each node uses its own distributed precise localization solution as the absolute pose constraint and the platform pre-integration result as the inter-frame relative motion constraint to construct the ontology pose graph nodes and edges. Step S82: Each node extracts the distance value from the inter-machine radio ranging signal and constructs a ranging factor edge connecting its own pose node and the corresponding neighboring pose node. Step S83: Each node performs feature association on the relative line-of-sight vector contained in the relative measurement information to construct a visual observation factor edge connecting its own pose node and the corresponding neighboring pose node; Step S84: Each node retrieves the rigid-flexible coupling dynamic parameters and dynamic spatial envelope, defines a load state variable node for each UAV to characterize the relative motion state of the pole, and constructs a kinematic prediction model describing the relative swing and vibration process of the pole based on the rigid-flexible coupling dynamic parameters. The kinematic prediction model is used as a load dynamic prediction factor edge connecting the body pose node and the load state variable node. Step S85: Each node calculates the probability of occlusion or non-line-of-sight interference to the inter-machine radio ranging signal and visual observation under the current estimated load pose based on the dynamic spatial envelope, and dynamically weights and adjusts the information matrix of the ranging factor edge and the visual observation factor edge accordingly. Step S86: Each node uses a distributed iterative optimization algorithm to jointly solve the factor graph composed of the ontology pose graph nodes and edges, ranging factor edges, visual observation factor edges, load state variable nodes and load dynamics prediction factor edges. While minimizing the global reprojection error and ranging residual, the load dynamics prediction error is also minimized, thereby synchronously obtaining the consistent collaborative positioning results of global spatiotemporal reference alignment among multiple machines, as well as the motion state estimation of the suspended poles of each machine.

10. A collaborative positioning system for a swarm of unmanned aerial vehicles (UAVs) used for pole transportation, characterized in that: Includes the following modules: The positioning environment mode determination module is used to obtain the quality indicators of the global navigation satellite system signals received by each node in the UAV swarm at the current moment, and to determine the positioning environment mode of each node. The inter-aircraft ranging activation and constraint acquisition module is used to activate the airborne radio ranging module based on the positioning environment mode, and acquire the inter-aircraft radio ranging signal with at least one neighboring node in a broadcast manner as a relative distance constraint. The tightly coupled positioning solution module is used in the airborne processing units of each node to construct a joint positioning solution model with the carrier phase observations of the global navigation satellite system, using the relative distance constraint as a strong constraint condition, to constrain the search space of integer ambiguity, so as to output a centimeter-level distributed precise positioning solution; The multi-source inertial data synchronization and pre-integration module is used to synchronously acquire the body data collected by the airborne inertial measurement unit and the load data collected by the inertial sensing units distributed on the suspended pole structure, and to perform pre-integration processing on the body data to obtain the platform pre-integration result. The parameter identification and spatial envelope calculation module is used to input the body data, load data and platform pre-integration results into a nonlinear state estimator, identify the rigid-flexible coupling dynamic parameters of the suspension system online, and calculate the dynamic spatial envelope representing the sweep range of the pole in real time. The dynamic obstacle avoidance trajectory planning module is used to generate a three-dimensional navigation trajectory online that takes into account both obstacle avoidance and attitude stability margin, based on the distributed precise positioning solution and dynamic spatial envelope. The same positioning mode startup module is used to start the distributed vision-radio cooperative positioning mode when the quality of the inter-machine communication link is detected to be lower than the preset quality threshold. Each node obtains relative measurement information including the relative line-of-sight vector of the neighboring machine through the airborne vision sensor. The cooperative localization and state estimation module is used by each node to construct and solve a distributed pose graph optimization problem based on a factor graph, using the relative measurement information, the distance value extracted from the inter-machine radio ranging signal, and its own distributed precise localization solution as factors, thereby obtaining a consistent cooperative localization result with global spatiotemporal reference alignment among multiple machines.

Images