Unmanned aerial vehicle indoor continuous positioning method based on Levenberg-Marquardt in combination with NURBS

The NURBS curve is optimized through the Levenberg-Marquardt algorithm, and the IMU and UWB data are integrated, and the problem of discontinuity in indoor drone positioning is solved, high-precision and stable indoor drone positioning is achieved, and the system adaptability and resilience are enhanced. Great.

CN118913266APending Publication Date: 2024-11-08GUILIN UNIV OF ELECTRONIC TECH +2
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202411240677.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-05
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

In the existing indoor drone positioning technology, traditional IMU, Wi-Fi and Bluetooth methods are difficult to provide high-precision and stable positioning, especially in complex environments, UWB signal attenuation and multipath effect lead to a decrease in positioning accuracy, and IMU and UWB data asynchronousness leads to trajectory discontinuity.

Method used

The NURBS curve is optimized by using the Levenberg-Marquardt algorithm, and by fusing IMU and UWB data, initializing control points and weights, a three-time NURBS trajectory interpolation model is established to realize the continuous trajectory fit of the drone, and reduce the trajectory discontinuity caused by asynchronous data.

Benefits of technology

It significantly improves the positioning accuracy and trajectory continuity of the drone in the indoor environment, enhances the adaptability and robustness of the system, reduces the dependence on hardware synchronization, and alleviates the problem of trajectory discontinuity caused by sensor asynchronousness.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118913266A_ABST
    Figure CN118913266A_ABST
Patent Text Reader

Abstract

The invention discloses a cubic NURBS spline method based on Levenberg-Marquardt algorithm optimization. The cubic NURBS spline method is used for fusing multi-source sensor data to achieve indoor positioning and continuous track fitting of an unmanned aerial vehicle. According to the method, high-frequency motion information and low-frequency accurate position data are obtained by using asynchronous data collected by IMU and UWB sensors. And a control point and a weight are optimized through a Levenberg-Marquardt algorithm so as to meet a convergence condition of trajectory fitting, and a cubic NURBS trajectory interpolation model is constructed. And then, establishing a continuous NURBS trajectory model by using the optimized control points and weights, ensuring that smooth trajectory interpolation is realized in each time window, and further keeping the continuity of the trajectory of the unmanned aerial vehicle in the whole flight process. According to the method, by directly processing the continuous asynchronous data of the multi-source sensor, the dependence on hardware synchronization is reduced, the precision and track continuity of indoor positioning of the unmanned aerial vehicle are improved, and a new solution is provided for continuous track fitting of the unmanned aerial vehicle in a complex environment.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of indoor unmanned aerial vehicle continuous trajectory fitting, and in particular to an LM-NURBS algorithm for performing continuous trajectory fitting using NURBS optimized by a Levenberg-Marquardt algorithm. Background Art

[0002] With the continuous development of indoor drone technology, accurate positioning and navigation have become crucial in various applications. Especially in complex indoor environments, traditional satellite navigation systems such as GPS often cannot provide stable positioning services because the signal has difficulty penetrating buildings. Therefore, in order to ensure the effective operation of drones in various scenarios, the development of indoor high-precision positioning technology has become an inevitable trend. At present, common indoor positioning technologies such as IMU, Wi-Fi and Bluetooth have been widely used, but their respective limitations limit the further improvement of positioning accuracy. For example, IMU is easily affected by cumulative errors, resulting in positioning drift, while the positioning accuracy of Wi-Fi and Bluetooth is easily affected by signal interference and environmental obstacles, and cannot meet the needs of high-precision positioning. Therefore, studying more accurate and stable indoor positioning methods has become an urgent problem to be solved.

[0003] In order to solve the limitations of traditional indoor positioning technology, UWB technology has attracted widespread attention as an emerging high-precision positioning method. UWB technology uses extremely short pulse radio signal transmission to provide centimeter-level positioning accuracy and has strong anti-interference ability, especially in environments with severe multipath effects. However, despite the advantages of high-precision positioning, UWB also faces certain challenges in practical applications. For example, in long-distance positioning scenarios, UWB signals will be significantly attenuated, and multipath effects may cause positioning accuracy to decrease. In addition, in dynamic environments, UWB positioning results may become unstable due to signal obstruction and reflection. In order to overcome these problems, researchers began to explore positioning technologies that fuse UWB with other sensors to further improve the accuracy and robustness of positioning.

[0004] In multi-sensor fusion positioning technology, the combination of IMU and UWB is a common solution. IMU can provide high-frequency acceleration and angular velocity data, which is suitable for real-time tracking of the target's motion state, but is easily affected by cumulative errors. By combining with UWB's low-frequency, high-precision position data, they can complement each other's shortcomings and improve the overall positioning accuracy and system stability. IMU data can maintain continuous tracking of the target when the UWB signal is temporarily lost or the signal quality decreases, while UWB data can be used to periodically correct the accumulated errors of IMU. However, this fusion method also faces some challenges, such as the asynchrony of IMU and UWB data and the time synchronization problem during data fusion, which may lead to discontinuity of trajectory and reduction of positioning accuracy. Therefore, more advanced fusion algorithms are needed to improve positioning accuracy and trajectory continuity.

[0005] In order to improve the accuracy of fusion positioning and the continuity of trajectory, researchers have recently introduced interpolation methods such as Bezier curves and B-splines into sensor data fusion. Bezier curves generate smooth paths through control points, which can reduce the impact of differences between sensor data on the trajectory. However, they have high computational complexity, and when the control points change, the entire curve will be affected. In contrast, B-spline curves have better local control characteristics and can keep other parts of the curve unchanged when certain control points change, thereby improving the flexibility and accuracy of trajectory fitting. However, B-splines have limitations in representing some complex geometric shapes because these shapes need to be accurately expressed by rational functions. Therefore, in order to more accurately describe complex trajectories and maintain computational efficiency, NURBS curves are introduced as an improvement.

[0006] NURBS curves introduce weights and non-uniform node vectors on the basis of traditional B-splines, greatly enhancing their flexibility and applicability. NURBS curves can accurately represent complex geometric shapes including circles and ellipses, and provide higher accuracy and better local control characteristics while maintaining trajectory smoothness. However, in the construction and application of NURBS curves, the setting of control points and weights is crucial. The control points determine the basic shape of the curve, while the weights affect the stretching and scaling effects of the curve near these control points. In order to optimize these parameters and ensure that NURBS curves can accurately meet the trajectory requirements in practical applications, researchers have introduced various optimization algorithms to adjust control points and weights. However, these traditional optimization methods usually have high computational complexity and are very sensitive to initial parameters, which may lead to unsatisfactory optimization results.

[0007] To solve the above problems, this paper proposes a LM-NURBS algorithm that combines the Levenberg-Marquardt algorithm and NURBS curves. By using the Levenberg-Marquardt algorithm to optimize the NURBS control points and weights, continuous cubic NURBS interpolation is performed within the window time to obtain a continuous indoor UAV estimated trajectory. Experimental results on multiple public datasets show that the LM-NURBS algorithm exhibits higher accuracy than other traditional algorithms, and significantly improves the accuracy and stability of the UAV estimated trajectory in indoor environments. Summary of the invention

[0008] The technical problem to be solved by the present invention is how to effectively use asynchronous data from multi-source sensors to solve the discreteness problem of indoor UAV trajectory estimation.

[0009] In order to solve the above technical problems, the technical solution adopted by the present invention is: an indoor UAV trajectory fitting method based on Levenberg-Marquardt-NURBS, which specifically includes the following steps:

[0010] S1: Get the current time window t i-2 to i+1 The high-frequency motion data of the IMU and the low-frequency precise position data of the UWB are used to initialize the terminal control point P i+1 and weight W i+1 ;

[0011] S2: Combine sensor data and optimize control point P using Levenberg-Marquardt algorithm i+1 and weight W i+1 , when the optimization meets the convergence conditions, the optimized terminal control points and weights are output;

[0012] S3: The optimized control point P i+1 and weight W i+1 Incorporate cubic non-uniform rational B-spline (NURBS), by knowing the current time window t i-2 to i+1 The control points and weights of the set are used to establish a cubic NURBS trajectory interpolation model to achieve continuous trajectory interpolation;

[0013] S4: Output the continuous trajectory and enter the next time window, repeating the above process until the drone reaches the end point.

[0014] The above technical solution is adopted to solve the problem of discontinuous trajectory in traditional positioning systems by fusing IMU and UWB sensor data. The specific method is to use the high-frequency motion data provided by IMU to maintain the real-time motion tracking of the drone, and combine the low-frequency and high-precision position data of UWB for periodic correction to eliminate the cumulative error of IMU. The adaptive strategy is determined by experiments. When the UWB signal is blocked or interfered, only IMU is relied on for short-term positioning to maintain the continuity and stability of the trajectory. Due to the limitations of IMU and UWB, the data is smoothed by NURBS curve, the control points and weights are optimized, and the continuous trajectory interpolation between sensor data is realized, reducing the problem of discontinuous trajectory caused by asynchronous fusion. In the indoor positioning environment, since the UWB signal is easily interfered by the multipath effect and the IMU data is prone to drift in a short time, the Levenberg-Marquardt algorithm is used to optimize the control points and weights of the NURBS curve to ensure the smoothness and accuracy of the trajectory interpolation. This method realizes the continuous positioning of UAVs in complex indoor environments. Through the dynamic fusion of multi-source data, it significantly improves the continuity and accuracy of positioning, ensuring the stability and reliability of the positioning system in various complex environments.

[0015] Preferably, in step S1, the IMU sensor is used to provide high-frequency motion data, the UWB sensor is used to provide low-frequency precise position data, and the IMU and UWB data are used to initialize the end control point P i+1 and weight W i+1 .

[0016] Preferably, the specific steps of step S1 are:

[0017] S11: In time window t i-2 to i+1 The high-frequency motion data is collected by the IMU sensor, and the low-frequency precise position data is collected by the UWB sensor, and the data is asynchronous data;

[0018] S12: Determine the end control point P of the current time window based on the collected high-frequency motion data and low-frequency precise position data i+1 and weight W i+1 , and initialize it.

[0019] Preferably, the specific steps of step S2 are:

[0020] S21: Combine the high-frequency motion data and low-frequency precise position data obtained from the IMU and UWB sensors, and use these data to initialize the end control point P i+1 and weight W i+1 ;

[0021] S22: Apply the Levenberg-Marquardt algorithm to the control point P based on the sensor data i+1 and weight W i+1 Optimize and continuously adjust the control points and weights until the optimization process meets the predetermined convergence conditions. When the optimization process reaches the convergence conditions, the optimized terminal control point P is output. i+1 and weight W i+1 .

[0022] Preferably, in step S22, the Levenberg-Marquardt algorithm is used to iteratively optimize the control point P i+1 and weight W i+1 And output the final optimized control point P i+1 and weight W i+1 ; The specific steps are:

[0023] S221: Based on the control point P to be optimized i+1 and weight W i+1 , combined with the known control points and weights in the current time window, a cubic NURBS curve R(t) is constructed, where N(t) is the spline basis function, P is the control point, W is the weight, and the R(t) formula is:

[0024]

[0025] S222: Define the objective function F(P i+1 ,W i+1 ), which is used to evaluate the error between the sensor data and the trajectory model. j uwb and z k imu are the observed positions calculated from the UWB and IMU data observed at timestamps j and k, respectively. The objective function is defined as:

[0026]

[0027] S223: Calculate the objective function F(P,W) with respect to P i+1 , and W i+1 The partial derivative of :

[0028]

[0029] S224: According to F(P,W) about P i+1 , and W i+1 Partial derivatives of the objective function F(P,W) are used to calculate the Jacobian matrix J of the control points and weights:

[0030]

[0031] S225: In t i-2 to i+1 In the time window of t uwb and z t imu The position observation data is subtracted from the NURBS interpolation formula R(t) to obtain the residual vector E:

[0032]

[0033] S226: Solving the update amount δ θ , where I is the identity matrix and λ is the damping factor, initialized to λ = 10 -6 . Construct the Levenberg-Marquardt normal equation:

[0034] J T J+λIδθ=-J T E

[0035] S227: Using update amount δ θ Adjust control points and weights:

[0036]

[0037] P i+1new ←P i+1 +δP i+1 ,W i+1new ←W i+1 +δW i+1

[0038] S228: Finally, according to the updated (P i+1 ,W i+1 The error E obtained by new is used to adjust the damping factor λ. If the error decreases, λ is reduced, otherwise it is increased, and it is decided whether to update P i+1 ,W i+1 . Check δ θ The convergence condition is that if δ θ If the convergence condition of is satisfied, the iteration is stopped and the optimized control point P is output. i+1 and weight W i+1 .

[0039] Preferably, the specific steps of step S3 are:

[0040] S31: The optimized control point P i+1 and weight W i+1 Incorporate into a cubic non-uniform rational B-spline (NURBS) curve and update the control points and weight set of the NURBS curve;

[0041] S32: Using the optimized control point Pi+1 and weight W i+1 , and the current time window from t i-2 to i+1 All known control points and weights of , construct the cubic NURBS trajectory interpolation model R(t);

[0042] S33: Through the cubic NURBS trajectory interpolation model R(t), the time window is realized from t i-2 to i+1 The continuous trajectory interpolation within the time window ensures the smoothness and continuity of the trajectory within the time window.

[0043] Preferably, the step S4 includes outputting the continuous trajectory and entering the next time window. Specifically, the steps are:

[0044] S41: Output time window t i-2 to i+1 Continuous trajectory of

[0045] S42: time index i plus the time interval Δt of the control point, entering the next time window;

[0046] Compared with the prior art, the present invention has the following beneficial effects:

[0047] (1) Improve positioning accuracy and trajectory continuity: The LM-NURBS algorithm of the present invention achieves high-precision trajectory fitting and continuous positioning by fusing IMU and UWB data and using the Levenberg-Marquardt algorithm to optimize the control points and weights of the NURBS curve. Compared with the prior art, this method can effectively compensate for the IMU cumulative error and the error caused by the multipath effect of the UWB signal in a complex environment, thereby improving the overall positioning accuracy.

[0048] (2) Enhance the adaptability and robustness of the system: The UAV indoor positioning method of the present invention exhibits excellent adaptability and robustness in various complex indoor environments. By dynamically adjusting the parameters of the fusion algorithm, the present invention can effectively deal with complex situations such as signal interference, occlusion and loss, and maintain the high accuracy and stability of the system. In particular, in the case of severe signal occlusion or drastic environmental changes, the present invention can ensure the continuity and accuracy of positioning by adjusting the fusion strategy of sensor data.

[0049] (3) Reduce the dependence on hardware synchronization: The present invention reduces the strict requirements for hardware synchronization of various sensors by asynchronously fusing multi-source sensor data. By using the interpolation method of NURBS curves, the present invention achieves a smooth transition between sensor data of different frequencies, effectively alleviating the trajectory discontinuity problem caused by sensor asynchrony. This method not only improves the flexibility of data fusion, but also reduces the complexity of system implementation and hardware cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 A schematic diagram of the process framework of the method for fitting the continuous trajectory of an indoor UAV using the LM-NURBS algorithm based on the Levenberg-Marquardt optimized NURBS of the present invention;

[0051] Figure 2 It is a schematic diagram of the process framework of optimizing control points and weights based on the Levenberg-Marquardt algorithm of the present invention;

[0052] FIG3( a ) is a comparison diagram of the indoor UAV continuous trajectory fitting method using the LM-NURBS algorithm of the present invention and other trajectory fitting methods in Experiment 1;

[0053] FIG3( b ) is a comparison diagram of the indoor UAV continuous trajectory fitting method using the LM-NURBS algorithm of the present invention and other trajectory fitting methods in Experiment 2;

[0054] FIG3( c ) is a comparison diagram of the indoor UAV continuous trajectory fitting method using the LM-NURBS algorithm of the present invention in Experiment 3 and other trajectory fitting methods. DETAILED DESCRIPTION

[0055] The embodiments of the present invention are described in detail below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and are not intended to limit the protection scope of the present invention.

[0056] Example: Figures 1-2 As shown, the indoor and outdoor seamless positioning method based on UWB / GNSS / IMU specifically includes the following steps:

[0057] S1: Get the current time window t i-2 to i+1 The high-frequency motion data of the IMU and the low-frequency precise position data of the UWB are used to initialize the terminal control point P i+1 and weight W i+1 .

[0058] The specific steps of step S1 are:

[0059] S11: In time window t i-2 to i+1 Collect high-frequency motion data through IMU sensors and collect low-frequency precise position data through UWB sensors;

[0060] S12: Determine the end control point P of the current time window based on the collected high-frequency motion data and low-frequency precise position data i+1 and weight Wi+1 , and initialize it.

[0061] S2: Combine sensor data and optimize control point P using Levenberg-Marquardt algorithm i+1 and weight W i+1 , when the optimization meets the convergence conditions, the optimized terminal control points and weights are output.

[0062] The specific steps of step S2 are:

[0063] S21: Combine the high-frequency motion data and low-frequency precise position data obtained from the IMU and UWB sensors, and use these data to initialize the end control point P i+1 and weight W i+1 ;

[0064] S22: Apply the Levenberg-Marquardt algorithm to the control point P based on the sensor data i+1 and weight W i+1 Optimize and continuously adjust the control points and weights until the optimization process meets the predetermined convergence conditions. When the optimization process reaches the convergence conditions, the optimized terminal control point P is output. i+1 and weight W i+1 .

[0065] In step S22, the Levenberg-Marquardt algorithm is used to optimize the control point P i+1 and weight W i+1 ; The specific steps are:

[0066] S221: Based on the control point P to be optimized i+1 and weight W i+1 , combined with the known control points and weights in the current time window, a cubic NURBS curve R(t) is constructed, where N(t) is the spline basis function, P is the control point, W is the weight, and the R(t) formula is:

[0067]

[0068] S222: Define the objective function F(P i+1 ,W i+1 ), which is used to evaluate the error between the sensor data and the trajectory model. j uwb and z k imu are the observed positions calculated from the UWB and IMU data observed at timestamps j and k, respectively. The objective function is defined as:

[0069]

[0070] S223: Calculate the objective function F(P,W) with respect to P i+1 , and W i+1 The partial derivative of :

[0071]

[0072] S224: According to F(P,W) about P i+1 , and W i+1 The partial derivative of the objective function F(P i+1 ,P i+1 )The Jacobian matrix J of the control points and weights:

[0073]

[0074] S225: In t i-2 to i+1 In the time window of t uwb and z t imu The position observation data is subtracted from the NURBS interpolation formula R(t) to obtain the residual vector E:

[0075]

[0076] S226: Solving the update amount δ θ , where I is the identity matrix and λ is the damping factor, initialized to λ = 10 -6 . Construct the Levenberg-Marquardt normal equation:

[0077] J T J+λIδθ=-J T E

[0078] S227: Using update amount δ θ Adjust control points and weights:

[0079]

[0080] P i+1new ←P i+1 +δP i+1 ,W i+1new ←W i+1 +δW i+1

[0081] S228: Finally, according to the updated (P i+1 ,W i+1 The error E obtained by new is used to adjust the damping factor λ. If the error decreases, λ is reduced, otherwise it is increased, and it is decided whether to update P i+1 ,W i+1. Check δ θ The convergence condition is that if δ θ If the convergence condition is met, the iteration is stopped and the optimized control point P i+1 and weight W i+1 .

[0082] S3: The optimized control point P i+1 and weight W i+1 Incorporate cubic non-uniform rational B-spline (NURBS), by knowing the current time window t i-2 to i+1 A cubic NURBS trajectory interpolation model is established based on the control points and weight sets to achieve continuous trajectory interpolation.

[0083] The specific steps of step S3 are:

[0084] S31: The optimized control point P i+1 and weight W i+1 Incorporate into a cubic non-uniform rational B-spline (NURBS) curve and update the control points and weight set of the NURBS curve;

[0085] S32: Using the optimized control point P i+1 and weight W i+1 , and the current time window from t i-2 to i+1 All known control points and weights of , construct the cubic NURBS trajectory interpolation model R(t);

[0086] S33: Through the cubic NURBS trajectory interpolation model R(t), the time window is realized from t i-2 to i+1 The continuous trajectory interpolation within the time window ensures the smoothness and continuity of the trajectory within the time window.

[0087] S4: Output the continuous trajectory and enter the next time window, repeating the above process until the drone reaches the end point.

[0088] The specific steps of step S4 are:

[0089] S41: Output time window t i-2 to i+1 Continuous trajectory of

[0090] S42: The time index i plus the time interval Δt of the control point enters the next time window.

[0091] In order to better illustrate the technical effect of the present invention, other trajectory fitting methods are used to conduct experimental verification with the trajectory fitting method of the present invention. The verification results are shown in Figures 3(a), 3(b), and 3(c). Among them, Figure 3(a) shows the motion trajectory of the drone in an indoor environment using the trajectory fitting method of the present invention; Figures 3(b) and 3(c) show the trajectory diagrams of the drone randomly moving around the central obstacle in an indoor environment using the method of the present invention.

[0092] For ordinary technicians in this field, the specific embodiments are only illustrative descriptions of the present invention. It is obvious that the specific implementation of the present invention is not limited to the above-mentioned methods. As long as various non-substantial improvements are made using the method concepts and technical solutions of the present invention, or the concepts and technical solutions of the present invention are directly applied to other occasions without improvement, they are all within the protection scope of the present invention.

Claims

1. A LM-NURBS algorithm based on Levenberg-Marquardt optimized NURBS is used to fit the indoor continuous trajectory of UAV, characterized in that: The specific steps include: S1: Get the current time window t i-2 to i+1 The high-frequency motion data of the IMU and the low-frequency precise position data of the UWB are used to initialize the terminal control point P i+1 and weight W i+1 ; S2: Combine sensor data and optimize control point P using Levenberg-Marquardt algorithm i+1 and weight W i+1 , when the optimization meets the convergence conditions, the optimized terminal control points and weights are output; S3: The optimized control point P i+1 and weight W i+1 Incorporate cubic non-uniform rational B-spline (NURBS), by knowing the current time window t i-2 to i+1 The control points and weights of the set are used to establish a cubic NURBS trajectory interpolation model to achieve continuous trajectory interpolation; S4: Output the continuous trajectory and enter the next time window, repeating the above process until the drone reaches the end point.

2. The method for estimating and positioning a UAV continuous trajectory according to claim 1, characterized in that: The specific steps of step S1 are: S11: In time window t i-2 to i+1 The high-frequency motion data is collected by the IMU sensor, and the low-frequency precise position data is collected by the UWB sensor, and the data is asynchronous data; S12: Determine the end control point P of the current time window based on the collected high-frequency motion data and low-frequency precise position data i+1 and weight W i+1 , and initialize it.

3. The method for estimating and positioning a UAV continuous trajectory according to claim 1, characterized in that: The specific steps of step S2 are: S21: Combine the high-frequency motion data and low-frequency precise position data obtained from the IMU and UWB sensors, and use these data to initialize the end control point P i+1 and weight W i+1 ; S22: Apply the Levenberg-Marquardt algorithm to the control point P based on the sensor data i+1 and weight W i+1 Optimize and continuously adjust the control points and weights until the optimization process meets the predetermined convergence conditions. When the optimization process reaches the convergence conditions, the optimized terminal control point P is output. i+1 and weight W i+1 .

4. The Levenberg-Marquardt algorithm according to claim 3 optimizes the control point P i+1 and weight W i+1 , characterized in that, The specific steps of step S22 are: S221: Based on the control point P to be optimized i+1 and weight W i+1 , combined with the known control points and weights in the current time window, a cubic NURBS curve R(t) is constructed, where N(t) is the spline basis function, P is the control point, W is the weight, and the R(t) formula is: S222: Define the objective function F(P i+1 ,W i+1 ), which is used to evaluate the error between the sensor data and the trajectory model. j uwb and z k imu are the observed positions calculated from the UWB and IMU data observed at timestamps j and k, respectively. The objective function is defined as: S223: Calculate the objective function F(P,W) with respect to P i+1 , and W i+1 The partial derivative of : S224: According to F(P,W) about P i+1 , and W i+1 The partial derivative of the objective function F(P i+1 ,W i+1 )The Jacobian matrix J of the control points and weights: S225: In t i-2 to i+1 In the time window of t uwb and z t imu The position observation data is subtracted from the NURBS interpolation formula R(t) to obtain the residual vector E: S226: Solving the update amount δ θ , where I is the identity matrix and λ is the damping factor, initialized to λ = 10 -6 . Construct the Levenberg-Marquardt normal equation: J T J+λIδθ=-J T E S227: Using update amount δ θ Adjust control points and weights: P i+1new ←P i+1 +δP i+1 ,W i+1new ←W i+1 +δW i+1 S228: Finally, according to the updated (P i+1 ,W i+1 The error E obtained by )new is used to adjust the damping factor λ. If the error decreases, λ is reduced, otherwise it is increased, and it is decided whether to update P i+1 ,W i+1 . Check δ θ The convergence condition is that if δ θ If the convergence condition is met, the iteration is stopped and the optimized control point P i+1 and weight W i+1 .

5. The method for fitting indoor continuous trajectories of unmanned aerial vehicles based on Levenberg-Marquardt optimized NURBS according to claim 1, characterized in that: The specific steps of step S3 are: S31: The optimized control point P i+1 and weight W i+1 Incorporate into a cubic non-uniform rational B-spline (NURBS) curve and update the control points and weight set of the NURBS curve; S32: Using the optimized control point P i+1 and weight W i+1 , and the current time window from t i-2 to i+1 All known control points and weights of , construct the cubic NURBS trajectory interpolation model R(t); S33: Through the cubic NURBS trajectory interpolation model R(t), the time window is realized from t i-2 to i+1 The continuous trajectory interpolation within the time window ensures the smoothness and continuity of the trajectory within the time window.

6. The method for fitting indoor continuous trajectories of unmanned aerial vehicles based on Levenberg-Marquardt optimized NURBS according to claim 1, characterized in that: The specific steps of step S4 are: S41: Output time window t i-2 to i+1 Continuous trajectory of S42: The time index i plus the time interval Δt of the control point enters the next time window.

Citation Information

Cited By

  • Adaptive optimization-based NURBS curve robot trajectory offset point focusing fitting method

    CN119511734A

  • Adaptive optimization based nurbs curve robot trajectory offset point focused fitting method

    CN119511734B