Non-cooperative robot servo anti-jamming control method and system to cope with visual sampling delay

Abstract

This invention discloses a non-cooperative robot servo anti-interference control method and system to address visual sampling delay. The method includes determining the relative position measurement values ​​between the UAV and the target mobile platform; determining whether it is a camera sampling moment in each control cycle, if so, inputting the measurement value to the observer; otherwise, generating a relative position prediction value based on the UAV's dynamic model and inputting it; the observer performs state estimation and compensates for time delay, while simultaneously estimating the total UAV disturbance as a state, outputting a full state vector; and determining the control quantity based on the full state vector and a control law including disturbance feedforward compensation to control the UAV to land accurately. This invention resolves the contradiction between "slow vision" and "fast control," overcomes time delay lag, and improves anti-interference capability and stability.

Description

Technical Field

[0001] This invention belongs to the technical field of robotics and automatic control, and particularly relates to a non-cooperative robot servo anti-interference control method and system for dealing with visual sampling delay. Background Technology

[0002] With the widespread application of drone technology in logistics delivery, infrastructure inspection, and military reconnaissance, its mission scenarios are expanding from open, static environments to complex, dynamic environments. One key and challenging application is achieving autonomous takeoff and landing on mobile platforms (such as moving vehicles and undulating ship decks). This capability can greatly expand the operational range and application value of drones, for example, enabling drone-mothership collaboration and delivery vehicles landing alongside them, and is currently one of the cutting-edge research topics in robotics.

[0003] Existing drones face multiple technical challenges when autonomously landing on dynamic mobile platforms. First, the sampling rate of visual sensors (such as ordinary RGB cameras) is typically much lower than the update frequency of the drone's flight controller. This means the controller spends most of its time unable to acquire new visual feedback, relying solely on outdated data for calculations. This creates an inherent contradiction between "slow vision" and "fast control," severely limiting the system's response speed and control performance. The controller's performance is thus constrained by the visual sampling rate, preventing it from reaching its full potential. Second, the acquisition, transmission, and processing of image information introduce non-negligible time delays. These delays cause phase lag in the control loop, worsening its stability margin and potentially leading to system oscillations or even instability. Traditional delay compensation methods are ill-suited to scenarios with delays and interference, where accurate information about the target platform's operating state is unavailable. Finally, the target platform's motion (velocity, acceleration, attitude changes) is an unknown and time-varying strong disturbance for the drone. Furthermore, the drone itself suffers from model uncertainties and external wind disturbances. This makes it difficult for the system to maintain robustness in dynamic, unknown environments. Traditional control methods struggle to provide timely and effective responses, easily leading to target loss or landing failure. Summary of the Invention

[0004] This invention provides a non-cooperative robot servo anti-interference control method and system to address visual sampling delays. It can be used to solve the problem in the prior art that UAVs are difficult to maintain robustness in dynamic and unknown environments, and are difficult to make timely and effective responses, which can easily lead to loss of target tracking or landing failure.

[0005] In a first aspect, the present invention provides a non-cooperative robot servo anti-interference control method to address visual sampling delay, comprising:

[0006] Determine the relative position measurements between the UAV and the target mobile platform;

[0007] Within each control cycle of the airborne observer, determine whether the current moment is the sampling moment of the UAV's airborne camera;

[0008] If the current time is the sampling time, the relative position measurement value of the target moving platform is input to the observer;

[0009] If the current time is not the sampling time, then based on the UAV dynamics model and the state estimate of the previous control cycle, the predicted relative position of the UAV and the target mobile platform is generated and input to the observer.

[0010] Based on the values ​​input to the observer, state estimation is performed and time delay is compensated. At the same time, the total disturbance of the UAV is extended into a state and estimated. The output is a full state vector including the UAV state estimate and the disturbance estimate after time delay compensation.

[0011] The control quantity for the current control cycle is determined based on the full state vector and the preset control law to control the UAV to land on the target mobile platform; wherein, the control law includes feedforward compensation for the disturbance estimate.

[0012] Optionally, determining the relative position measurement between the UAV and the target mobile platform includes:

[0013] The translation matrix of the spatial relationship between the landmark coordinate system M and the camera coordinate system C is calculated using the following formula. :

[0014] ;

[0015] Among them, P C The 3D coordinates of the AprilTag origin in the camera coordinate system C; P is the rotation matrix representing the spatial relationship between the landmark coordinate system M and the camera coordinate system C; m The three-dimensional coordinates of the AprilTag origin in the landmark coordinate system M;

[0016] The relative position measurement value r between the UAV and the target mobile platform is calculated based on the translation matrix of the spatial relationship between the landmark coordinate system M and the camera coordinate system C. pt :

[0017] ;

[0018] Where R is the rotation matrix, used to describe the coordinate transformation from the UAV coordinate system B to the inertial coordinate system W; This is the transformation matrix between the camera coordinate system and the UAV coordinate system.

[0019] Optionally, generating a predicted relative position between the UAV and the target mobile platform based on the UAV dynamics model and the state estimate from the previous control cycle, and inputting it to the observer, includes:

[0020] Constructing the dynamic model expression for the UAV:

[0021] ;

[0022] in, The current control time t k The predicted state; e is the natural constant; T is the sampling time; For the previous time t k-1 The estimated state of the unmanned aerial vehicle; The new state matrix is ​​formed by discretizing the original state matrix B using Euler; u(t) k-1 ) represents the previous time t k-1 The control quantity;

[0023] The predicted relative position y between the UAV and the target mobile platform is calculated using the following formula. p (t k ):

[0024] ;

[0025] Where C is the output state matrix.

[0026] Optionally, the process involves performing state estimation based on the values ​​input to the observer, compensating for time delays, and simultaneously expanding the total UAV disturbance into states and estimating them. The output is a full state vector including the time-delay-compensated UAV state estimate and the disturbance estimate, comprising:

[0027] Calculate the current control time t using the following formula. k Full-state estimation vector :

[0028] ;

[0029] in, This is an estimate of the error between the relative position and the expected relative position; This is an estimate of the derivative of the error between the relative position and the expected relative position; This is the estimated value of the disturbance; The data consists of a mixture of predicted and estimated observations; Q1 is the first state matrix. This represents the new state after expansion; M is a multiple of the sampling time; T is the sampling time. Q1 represents the control input before time 2MT; Q2 represents the second state matrix. Q1 represents the control quantity before time MT; Q2 represents the third state matrix.

[0030] Optionally, determining the control quantity for the current control cycle based on the full state vector and a preset control law to control the UAV to land on the target mobile platform includes:

[0031] Calculate the control quantity u(t) for the current control cycle using the following formula. k ):

[0032] ;

[0033] Where K is the overall controller gain; k1 is the first controller gain; and k2 is the second controller gain. The current control time t k The full-state estimation vector.

[0034] Secondly, the present invention provides a non-cooperative robot servo anti-interference control system to cope with visual sampling delay, comprising:

[0035] The first determining module is used to determine the relative position measurement values ​​between the UAV and the target mobile platform;

[0036] The judgment module is used to determine whether the current moment is the sampling moment of the UAV's airborne camera in each control cycle of the airborne observer;

[0037] The second determining module is used to determine, when the judging module determines that the current time is the sampling time, to input the relative position measurement value of the target moving platform to the observer;

[0038] The third determination module is used to determine, when the judgment module determines that the current time is not the sampling time, to generate a predicted value of the relative position between the UAV and the target mobile platform based on the UAV dynamics model and the state estimate value of the previous control cycle, and input it to the observer.

[0039] The full state vector generation module is used to perform state estimation based on the values ​​input to the observer and compensate for time delay. At the same time, it expands the total disturbance of the UAV into a state and estimates it, and outputs a full state vector including the UAV state estimate and the disturbance estimate after time delay compensation.

[0040] The fourth determination module is used to determine the control quantity of the current control cycle based on the full state vector and the preset control law, so as to control the UAV to land on the target mobile platform; wherein, the control law includes feedforward compensation for the disturbance estimate.

[0041] Optionally, the first determining module includes:

[0042] The first calculation unit is used to calculate the translation matrix of the spatial relationship between the landmark coordinate system M and the camera coordinate system C according to the following formula. :

[0043] ;

[0044] Among them, P C The 3D coordinates of the AprilTag origin in the camera coordinate system C; P is the rotation matrix representing the spatial relationship between the landmark coordinate system M and the camera coordinate system C; m The three-dimensional coordinates of the AprilTag origin in the landmark coordinate system M;

[0045] The second calculation unit is used to calculate the relative position measurement value r between the UAV and the target mobile platform based on the translation matrix of the spatial relationship between the landmark coordinate system M and the camera coordinate system C. pt :

[0046] ;

[0047] Where R is the rotation matrix, used to describe the coordinate transformation from the UAV coordinate system B to the inertial coordinate system W; This is the transformation matrix between the camera coordinate system and the UAV coordinate system.

[0048] Optionally, the third determining module includes:

[0049] Building blocks are used to construct expressions for the dynamic model of the UAV:

[0050] ;

[0051] in, The current control time t k The predicted state; e is the natural constant; T is the sampling time; For the previous time t k-1 The estimated state of the unmanned aerial vehicle; The new state matrix is ​​formed by discretizing the original state matrix B using Euler; u(t) k-1 ) represents the previous time t k-1 The control quantity;

[0052] The third calculation unit is used to calculate the predicted relative position y between the UAV and the target mobile platform according to the following formula. p (t k ):

[0053] ;

[0054] Where C is the output state matrix.

[0055] Optionally, the full-state vector generation module includes:

[0056] The fourth calculation unit is used to calculate the current control time t according to the following formula. k Full-state estimation vector :

[0057] ;

[0058] in, This is an estimate of the error between the relative position and the expected relative position; This is an estimate of the derivative of the error between the relative position and the expected relative position; This is the estimated value of the disturbance; The data consists of a mixture of predicted and estimated observations; Q1 is the first state matrix. This represents the new state after expansion; M is a multiple of the sampling time; T is the sampling time. Q1 represents the control input before time 2MT; Q2 represents the second state matrix. Q1 represents the control quantity before time MT; Q2 represents the third state matrix.

[0059] Optionally, the fourth determining module includes:

[0060] The fifth calculation unit is used to calculate the control quantity u(t) for the current control cycle according to the following formula. k ):

[0061] ;

[0062] Where K is the overall controller gain; k1 is the first controller gain; and k2 is the second controller gain. The current control time t k The full-state estimation vector.

[0063] This invention provides a non-cooperative robot servo anti-interference control method and system to address visual sampling delays. It designs a composite observation architecture that integrates an Alternating Predictive Observer (APO) and a Time Delay Extended State Alternating Predictive Observer (TDESAPO). This architecture directly uses measured values ​​during visual sampling and generates high-frequency virtual relative position predictions based on a high-precision UAV dynamics model during non-sampling times. This successfully increases the control frequency from the visual sampling rate (e.g., 30Hz) to the flight control loop rate (e.g., 100Hz), fundamentally resolving the inherent contradiction between "slow vision" and "fast control." Simultaneously, the observer incorporates an online delay estimation and compensation mechanism, enabling "timestamp alignment" and look-ahead prediction of delayed visual data. This effectively compensates for the phase lag introduced by the image processing link, significantly improving the system's stability margin and avoiding the oscillation or instability problems caused by delays in traditional methods.

[0064] Secondly, this invention extends the total disturbance of the UAV (including unknown maneuvers of the target platform, wind disturbance, and model uncertainty) to the system state and estimates it in real time. Combined with feedforward compensation for the disturbance estimate in the control law, a complete active disturbance rejection control framework is formed. This allows the system to transform the complex tracking problem into a simple integrator cascade system control problem, greatly enhancing its adaptability to dynamic unknown environments and its robustness against disturbances. Simulation experiments show that even under harsh conditions such as rapid maneuvers by the target platform (e.g., 5 m / s linear motion or 0.5 rad / s circular motion) and strong wind disturbance, this invention can still achieve smooth, stable, and high-precision tracking and landing, with performance significantly superior to traditional methods.

[0065] Furthermore, this invention eliminates the reliance on cooperative markers, platform communication, or high-cost, high-precision external sensors. It achieves fully autonomous landing solely through target feature extraction via an onboard vision sensor and processing by a core algorithm. This feature significantly enhances the system's applicability and deployability in non-cooperative scenarios such as emergency rescue and military reconnaissance, while simultaneously reducing the overall system cost and complexity. Attached Figure Description

[0066] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0067] Figure 1 A flowchart illustrating a non-cooperative robot servo anti-interference control method for addressing visual sampling delay, provided in an embodiment of the present invention;

[0068] Figure 2 This is a schematic diagram of the coordinate systems provided in the embodiments of the present invention;

[0069] Figure 3 This is a block diagram of multi-rate (APO) sampling control for interference and measurement delay compensation provided in an embodiment of the present invention;

[0070] Figure 4 A UAV loop diagram based on PBVS-APO provided for embodiments of the present invention;

[0071] Figure 5 A schematic diagram of simulation results for a comparative example provided in an embodiment of the present invention;

[0072] Figure 6 A schematic diagram of the structure of a non-cooperative robot servo anti-interference control system for dealing with visual sampling delay provided in an embodiment of the present invention. Detailed Implementation

[0073] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0074] Example 1

[0075] like Figure 1 As shown, this embodiment provides a non-cooperative robot servo anti-interference control method to address visual sampling delay, including:

[0076] Step 101: Determine the relative position measurement values ​​between the UAV and the target mobile platform.

[0077] For example, in this step, a drone model is constructed:

[0078] .

[0079] Where m represents the mass of the drone, and Its inertial tensor is characterized. Gravitational acceleration is determined by... express, This is a unit vector along the vertical axis. Control inputs include the thrust magnitude f and the torque vector. Both are in the quadcopter body coordinate system The hat operator is defined in the middle. .

[0080] The translation matrix of the spatial relationship between the landmark coordinate system M and the camera coordinate system C is calculated using the following formula. :

[0081] ;

[0082] Among them, P C The 3D coordinates of the AprilTag origin in the camera coordinate system C; P is the rotation matrix representing the spatial relationship between the landmark coordinate system M and the camera coordinate system C; m The three-dimensional coordinates of the AprilTag origin in the landmark coordinate system M;

[0083] Based on the translation matrix of the spatial relationship between the landmark coordinate system M and the camera coordinate system C Calculate the relative position measurement r between the UAV and the target mobile platform. pt :

[0084] ;

[0085] Where R is the rotation matrix, used to describe the coordinate transformation from the UAV coordinate system B to the inertial coordinate system W; This is the transformation matrix between the camera coordinate system and the UAV coordinate system.

[0086] Step 102: Determine whether the current moment is the sampling moment of the UAV's airborne camera within each control cycle of the airborne observer.

[0087] Step 103: If the current time is the sampling time, input the relative position measurement value of the target moving platform into the observer.

[0088] Step 104: If the current time is not the sampling time, generate the predicted relative position of the UAV and the target mobile platform based on the UAV dynamics model and the state estimate of the previous control cycle, and input it into the observer.

[0089] For example, construct the expression for the drone dynamics model:

[0090] .

[0091] in, The current control time t k The predicted state; e is the natural constant; T is the sampling time; For the previous time t k-1 The estimated state of the unmanned aerial vehicle; The new state matrix is ​​formed by discretizing the original state matrix B using Euler; u(t) k-1 ) represents the previous time t k-1 The control quantity.

[0092] The predicted relative position y between the UAV and the target mobile platform is calculated using the following formula. p (t k ):

[0093] .

[0094] Where C is the output state matrix.

[0095] Step 105: Based on the values ​​input to the observer, perform state estimation and compensate for time delay. At the same time, expand the total disturbance of the UAV into a state and estimate it. Output a full state vector including the UAV state estimate and disturbance estimate after time delay compensation.

[0096] For example, the current control time t is calculated according to the following formula. k Full-state estimation vector :

[0097] .

[0098] in, This is an estimate of the error between the relative position and the expected relative position; This is an estimate of the derivative of the error between the relative position and the expected relative position; This is the estimated value of the disturbance; The data consists of a mixture of predicted and estimated observations; Q1 is the first state matrix. This represents the new state after expansion; M is a multiple of the sampling time; T is the sampling time. Q1 represents the control input before time 2MT; Q2 represents the second state matrix. Q1 represents the control quantity before time MT; Q2 represents the third state matrix.

[0099] Step 106: Determine the control quantity for the current control cycle based on the full state vector and the preset control law, so as to control the UAV to land on the target mobile platform; wherein, the control law includes feedforward compensation for the disturbance estimate.

[0100] For example, the control quantity u(t) for the current control cycle is calculated according to the following formula. k ):

[0101] .

[0102] Where K is the overall controller gain; k1 is the first controller gain; and k2 is the second controller gain. The current control time t k The full-state estimation vector.

[0103] like Figure 2 , Figure 3 and Figure 4 As shown, the landmark coordinate system and camera coordinate system The rotation and translation between them are respectively determined by and express. The origin is from This indicates that the origin is fixed on the moving platform and in the inertial coordinate system. The text indicates that due to the dynamic characteristics of the landing platform, its speed is... This indicates that, based on these estimates, it is possible to... Calculate the position of the quadcopter relative to the center of the landmark. . This represents a predetermined and fixed rotation transformation matrix, which characterizes... Compared to The direction.

[0104]

[0105] For ease of explanation, the X-axis is chosen as the representative case (the analysis process for the Y-axis and Z-axis is the same), resulting in the simplified model above. Direct measurement is not possible. It can only obtain information about relative positions. Information regarding the error between the actual and expected relative positions. and its derivative Total disturbance information (Including known gravitational effects), are combined into a new state vector. ,in yes The The first derivative. A, B, and C are system matrices obtained from the model. Indicates the state on the X-axis. The time delay variable is treated as a constant in this paper. The time delay step size is determined based on the camera sampling period and the time required by the sensing module. Set as .

[0106]

[0107] in, express The estimated value, Defined as observer gain, Depend on Given. At the sampling time The actual output is obtained by measuring with sensors. Conversely, for non-sampling times Using virtual output measurement Virtual output measurement It was generated through a dynamic system model. The following results were obtained by integrating the system model:

[0108]

[0109]

[0110] in, , According to the above formula, It can be represented as

[0111]

[0112]

[0113] The above is an introduction to the APO formula. Next, we need to combine it with time delay to design an extended state alternating predictive observer (TDESAPO).

[0114]

[0115]

[0116] The above formulas are obtained through the corresponding discrete model and the forward Euler method. With the help of these formulas, the time delay prediction module can be designed as follows:

[0117]

[0118]

[0119]

[0120] in, , and They are , and The estimated value, , , Due to the state Due to the unmeasurability of the output, the discrete-time APO is considered sufficient for designing output control laws. Therefore, the fast-update output feedback control law is expressed as:

[0121]

[0122]

[0123] Where F equals , equal , equal , The feedback control gain is to be determined. Represented as .

[0124] This embodiment included a simulation comparison experiment:

[0125] The mobile platform moves along a straight trajectory at constant speeds of 1 m / s and 5 m / s, while the quadcopter drone starts from a position offset relative to the platform, with offsets of -10 m in the X-axis direction, +30 m in the Y-axis direction, and +1 m in the Z-axis direction (assuming the quadcopter drone's vision system has successfully located the platform). A sampling period of 0.05 seconds and a time delay of 0.1 seconds are used. To evaluate anti-interference performance, an arbitrary reverse wind disturbance d = 2 m / s is applied relative to the platform's direction of motion. The control cycles of PBVS-APO and PBVS-SRC are set to 0.01 seconds and 0.05 seconds, respectively. In Simulation 2, the mobile platform is set to move in a circle with a radius of 15 meters, with angular velocities of 0.25 radians / s and 0.5 radians / s, respectively, while the initial position of the quadcopter drone is the same as in Simulation 1. During these simulations, a wind disturbance of 2 m / s is applied along the negative X-axis. According to established landing rules, landing is only permitted when the X-axis position error is kept below 0.4 meters and the Y-axis error is also below 0.4 meters.

[0126] Comparison with ordinary PBVS solution

[0127] Control group: A traditional PBVS with the same controller and vision rate (20Hz) was used, without dedicated anti-interference and delay compensation design.

[0128] Experimental Results: In a simulated environment with a moving platform and wind disturbance, this solution quickly failed to track. Its control commands were severely delayed, the UAV trajectory diverged, and it was unable to complete the landing mission. This result directly confirms the inevitable failure caused by "defects 1, 2, and 3" in the existing technology. Due to the direct divergence, in reality, the UAV would have already crashed, so there is no curve for this control group in the simulation.

[0129] Cause and effect principle: Its failure directly stemmed from the technical solution of "low frequency control + no time delay processing + no interference rejection", while the present invention directly targets and solves these problems through the technical solution of "high speed + time delay compensation + self-interference rejection".

[0130] Compared with the optimized PBVS scheme (with disturbance rejection but no delay resistance)

[0131] Control group: The controller still operates at the same rate as vision (20Hz), but introduces noise immunity design (such as ADRC), but does not handle latency.

[0132] Experimental results: The proposed scheme can achieve tracking when the target platform moves at low speeds, demonstrating the effectiveness of the disturbance rejection design. However, once the platform performs rapid maneuvers, the system also experiences tracking loss or a significant performance degradation. The root cause is the phase lag resulting from both low frequency and time delay, which renders the state information received by the disturbance rejection controller severely outdated, preventing it from making proper compensation.

[0133] Causal principle: This comparative experiment convincingly demonstrates that disturbance rejection design alone cannot overcome the coupling defects of "low frequency + time delay". The inventive contribution of this invention lies in its pioneering combination of "disturbance rejection" with "high speed / time delay compensation," forming a complete solution. Only when the control speed is sufficiently high and the state estimation is timely and accurate enough (through time delay compensation) can the disturbance rejection controller fully realize its effectiveness.

[0134] The present invention's solution

[0135] Experimental conditions: Under the same simulation environment, the platform was subjected to various speeds of maneuvering (including rapid maneuvering) and strong wind disturbance. The controller cycle of PBVS-APO was 100Hz.

[0136] Experimental results: The proposed solution successfully, stably, and accurately completed the tracking and landing tasks. The UAV trajectory was smooth, the control commands were continuous, and there were no oscillations or divergences.

[0137] Conclusion: The experimental data strongly demonstrate that the technical solution proposed in this invention directly and effectively solves all the core defects in the closest prior art, achieving a leapfrog improvement in technical performance. Its effect cannot be easily achieved by conventional optimization by those skilled in the art, highlighting the outstanding substantive features and significant progress of this invention.

[0138] like Figure 5 As shown, when the relative position error on the X-axis remains below 0.2 meters and the error on the Y-axis remains below 0.1 meters, the quadcopter drone begins to descend, with the height allowed to drop to 0.2 meters along the Z-axis. For Figure 5 The unknown platform shown in (a) is moving at a speed of 1 m / s. Both PBVS-APO and PBVS-SRC adhere to landing rules, but PBVS-APO achieves a faster descent rate. Figure 5 In (b), descent is permitted only if the X-axis position error remains below 0.4 meters and the Y-axis error does not exceed 0.1 meters. When the platform speed increased to 5 m / s, the PBVS-SRC failed to meet the landing requirements (primarily due to the X-axis position error exceeding 0.4 meters), resulting in a failed landing attempt. In contrast, the PBVS-APO strictly adhered to the landing rules and successfully completed the landing. Figure 5 As shown in (c), when tracking a platform moving at 0.25 radians / second, the PBVS-SRC system exhibits reduced stability and a time delay in meeting landing requirements. Conversely, the PBVS-APO system demonstrates superior stability and a faster descent rate. At a higher angular velocity of 0.5 radians / second, as... Figure 5As shown in (d), the PBVS-SRC failed to meet the landing criteria, resulting in a landing failure. In contrast, the PBVS-APO system consistently met the landing rules and successfully completed the landing operation.

[0139] In summary, the non-cooperative robot servo anti-interference control method for addressing visual sampling delay provided in this embodiment deeply integrates "alternating prediction" to solve the "slow vision" problem and "online estimation and compensation" to solve the "time delay" problem into a unified observer framework. When no new visual data is available, the observer generates high-frequency virtual relative pose predictions based on a high-precision UAV dynamics model, increasing the control update frequency from the camera sampling rate (e.g., 30Hz) to the flight control loop rate. The observer simultaneously predicts the state changes during the time delay process of the visual channel. It compares the state predicted by the model with the actual measured values ​​delivered after the time delay. Subsequently, this estimate is used to perform "timestamp alignment" and look-ahead prediction on the delayed visual data, compensating for the time delay effect and outputting the state estimate for the current moment. This mechanism overcomes the two core challenges of low sampling rate and high time delay. It breaks the bottleneck of control performance being limited by the hardware sampling rate, achieving high-frequency closed-loop control; simultaneously, through online time delay compensation, it significantly improves the system's phase margin and stability. This is the technological cornerstone for achieving high-performance, highly robust autonomous landing.

[0140] This embodiment provides a position-based visual servo (PBVS) control framework optimized for dynamic landing. Its core is an integrated high-speed, highly disturbance-resistant time-delay estimation active disturbance rejection position observer (TDESAPO) as the control core. TDESAPO takes the high-frequency state estimate output by the aforementioned observer as input, calculates and outputs control commands at the flight control rate, fully utilizing the controller's performance. The TDESAPO structure incorporates real-time estimation and compensation capabilities for total disturbances (including target platform maneuvers, wind disturbances, and model uncertainties), transforming the complex tracking problem into a simple integrator cascade system control problem. This framework does not rely on platform motion information; it effectively suppresses internal and external disturbances using only onboard visual information, ensuring that the UAV maintains smooth, stable, and high-precision tracking and landing performance even under complex conditions such as severe target platform maneuvers and external wind disturbances.

[0141] This embodiment eliminates the reliance on cooperative markers, platform communication, and high-precision IMU / LiDAR. It extracts target platform features (such as natural features or general markers) solely through airborne visual sensors and processes them using the algorithm proposed in this invention, achieving full autonomy from identification and tracking to landing. This method significantly improves the system's applicability and deployability, enabling its application in critical scenarios such as emergency rescue and military reconnaissance where cooperation with the platform is unpredictable or impossible, while reducing system cost and complexity.

[0142] Example 2

[0143] Based on the same inventive concept as Embodiment 1, this embodiment also provides a non-cooperative robot servo anti-interference control system to cope with visual sampling delay. Since the principle of this system in solving the problem is similar to the aforementioned non-cooperative robot servo anti-interference control method to cope with visual sampling delay, the implementation of this system can refer to the implementation of the non-cooperative robot servo anti-interference control method to cope with visual sampling delay.

[0144] like Figure 6 As shown, a non-cooperative robot servo anti-interference control system for dealing with visual sampling delay includes:

[0145] The first determining module 10 is used to determine the relative position measurement values ​​between the UAV and the target mobile platform.

[0146] The judgment module 20 is used to determine whether the current moment is the sampling moment of the UAV's airborne camera in each control cycle of the airborne observer.

[0147] The second determining module 30 is used to determine, when the determining module determines that the current time is the sampling time, to input the relative position measurement value of the target moving platform to the observer.

[0148] The third determining module 40 is used to determine, when the judgment module determines that the current time is not the sampling time, to generate a predicted value of the relative position between the UAV and the target mobile platform based on the UAV dynamics model and the state estimate value of the previous control cycle, and input it to the observer.

[0149] The full state vector generation module 50 is used to perform state estimation based on the values ​​input to the observer and compensate for time delay. At the same time, it expands the total disturbance of the UAV into a state and estimates it, and outputs a full state vector including the UAV state estimate and the disturbance estimate after time delay compensation.

[0150] The fourth determining module 60 is used to determine the control quantity of the current control cycle based on the full state vector and the preset control law, so as to control the UAV to land on the target mobile platform; wherein, the control law includes feedforward compensation for the disturbance estimate.

[0151] For example, the first determining module includes:

[0152] The first calculation unit is used to calculate the translation matrix of the spatial relationship between the landmark coordinate system M and the camera coordinate system C according to the following formula. :

[0153] ;

[0154] Among them, P C The 3D coordinates of the AprilTag origin in the camera coordinate system C; P is the rotation matrix representing the spatial relationship between the landmark coordinate system M and the camera coordinate system C; m The three-dimensional coordinates of the AprilTag origin in the landmark coordinate system M;

[0155] The second calculation unit is used to calculate the translation matrix based on the spatial relationship between the landmark coordinate system M and the camera coordinate system C. Calculate the relative position measurement r between the UAV and the target mobile platform. pt :

[0156] ;

[0157] Where R is the rotation matrix, used to describe the coordinate transformation from the UAV coordinate system B to the inertial coordinate system W; This is the transformation matrix between the camera coordinate system and the UAV coordinate system.

[0158] Optionally, the third determining module includes:

[0159] Building blocks are used to construct expressions for the dynamic model of the UAV:

[0160] .

[0161] in, The current control time t k The predicted state; e is the natural constant; T is the sampling time; For the previous time t k-1 The estimated state of the unmanned aerial vehicle; The new state matrix is ​​formed by discretizing the original state matrix B using Euler; u(t) k-1 ) represents the previous time t k-1 The control quantity.

[0162] The third calculation unit is used to calculate the predicted relative position y between the UAV and the target mobile platform according to the following formula. p (t k ):

[0163] .

[0164] Where C is the output state matrix.

[0165] Optionally, the full-state vector generation module includes:

[0166] The fourth calculation unit is used to calculate the current control time t according to the following formula. k Full-state estimation vector :

[0167] .

[0168] in, This is an estimate of the error between the relative position and the expected relative position; This is an estimate of the derivative of the error between the relative position and the expected relative position; This is the estimated value of the disturbance; The data consists of a mixture of predicted and estimated observations; Q1 is the first state matrix. This represents the new state after expansion; M is a multiple of the sampling time; T is the sampling time. Q1 represents the control input before time 2MT; Q2 represents the second state matrix. Q1 represents the control quantity before time MT; Q2 represents the third state matrix.

[0169] Optionally, the fourth determining module includes:

[0170] The fifth calculation unit is used to calculate the control quantity u(t) for the current control cycle according to the following formula. k ):

[0171] .

[0172] Where K is the overall controller gain; k1 is the first controller gain; and k2 is the second controller gain. The current control time t k The full-state estimation vector.

[0173] For more detailed information on the working process of each of the above modules, please refer to the relevant content disclosed in Example 1, which will not be repeated here.

[0174] Example 3

[0175] This embodiment provides a computer device, including a processor and a memory; wherein, when the processor executes a computer program stored in the memory, it implements the steps of the non-cooperative robot servo anti-interference control method for dealing with visual sampling delay described in Embodiment 1.

[0176] For a more detailed explanation of the above method, please refer to the relevant content disclosed in Example 1, which will not be repeated here.

[0177] Example 4

[0178] This embodiment provides a computer-readable storage medium for storing a computer program; when the computer program is executed by a processor, it implements the steps of the non-cooperative robot servo anti-interference control method for dealing with visual sampling delay described in Embodiment 1.

[0179] For a more detailed explanation of the above method, please refer to the relevant content disclosed in Example 1, which will not be repeated here.

[0180] Example 5

[0181] This embodiment provides a computer program product, including computer-executable instructions or a computer program. When the computer-executable instructions or the computer program are executed by a processor, they implement the steps of the non-cooperative robot servo anti-interference control method for dealing with visual sampling delay described in Embodiment 1.

[0182] For a more detailed explanation of the above method, please refer to the relevant content disclosed in Example 1, which will not be repeated here.

[0183] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems, devices, storage media, and computer program products disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple, and relevant parts can be referred to the method section.

[0184] Those skilled in the art will clearly understand that the techniques in the embodiments of the present invention can be implemented using software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solutions in the embodiments of the present invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or certain parts of the embodiments of the present invention.

[0185] In some embodiments, computer-executable instructions may take the form of programs, software, software modules, scripts, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as stand-alone programs or as modules, components, subroutines, or other units suitable for use in a computing environment.

[0186] As an example, computer-executable instructions may, but do not necessarily, correspond to files in a file system. They may be stored as part of a file that holds other programs or data, for example, in one or more scripts in a Hyper Text Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple co-located files (e.g., files that store one or more modules, subroutines, or code sections).

[0187] As an example, computer-executable instructions can be deployed to execute on a single electronic device, or on multiple electronic devices located in one location, or on multiple electronic devices distributed across multiple locations and interconnected via a communication network.

[0188] The present invention has been described in detail above with reference to specific embodiments and exemplary examples; however, these descriptions should not be construed as limiting the present invention. Those skilled in the art will understand that various equivalent substitutions, modifications, or improvements can be made to the technical solutions and embodiments of the present invention without departing from the spirit and scope of the invention, and all such modifications and improvements fall within the scope of the present invention. The scope of protection of the present invention is defined by the appended claims.

Claims

1. A non-cooperative robot servo anti-interference control method to address visual sampling delay, characterized in that, include: Determine the relative position measurements between the UAV and the target mobile platform; Within each control cycle of the airborne observer, determine whether the current moment is the sampling moment of the UAV's airborne camera; If the current time is the sampling time, the relative position measurement value of the target moving platform is input to the observer; If the current time is not the sampling time, then based on the UAV dynamics model and the state estimate of the previous control cycle, the predicted relative position of the UAV and the target mobile platform is generated and input to the observer. Based on the values ​​input to the observer, state estimation is performed and time delay is compensated. At the same time, the total disturbance of the UAV is extended into a state and estimated. The output is a full state vector including the UAV state estimate and the disturbance estimate after time delay compensation. The control quantity for the current control cycle is determined based on the full state vector and the preset control law to control the UAV to land on the target mobile platform; wherein, the control law includes feedforward compensation for the disturbance estimate.

2. The non-cooperative robot servo anti-interference control method for dealing with visual sampling delay according to claim 1, characterized in that, The determination of the relative position measurement values ​​between the UAV and the target mobile platform includes: The translation matrix of the spatial relationship between the landmark coordinate system M and the camera coordinate system C is calculated using the following formula. : ； Among them, P C The 3D coordinates of the AprilTag origin in the camera coordinate system C; P is the rotation matrix representing the spatial relationship between the landmark coordinate system M and the camera coordinate system C; m The three-dimensional coordinates of the AprilTag origin in the landmark coordinate system M; The relative position measurement value r between the UAV and the target mobile platform is calculated based on the translation matrix of the spatial relationship between the landmark coordinate system M and the camera coordinate system C. pt : ； Where R is the rotation matrix, used to describe the coordinate transformation from the UAV coordinate system B to the inertial coordinate system W; This is the transformation matrix between the camera coordinate system and the UAV coordinate system.

3. The non-cooperative robot servo anti-interference control method for dealing with visual sampling delay according to claim 1, characterized in that, The step of generating a predicted relative position between the UAV and the target mobile platform based on the UAV dynamics model and the state estimate from the previous control cycle, and inputting this predicted value to the observer, includes: Constructing the dynamic model expression for the UAV: ； in, The current control time t k The predicted state; e is the natural constant; T is the sampling time; For the previous time t k-1 The estimated state of the unmanned aerial vehicle; The new state matrix is ​​formed by discretizing the original state matrix B using Euler; u(t) k-1 ) represents the previous time t k-1 The control quantity; The predicted relative position y between the UAV and the target mobile platform is calculated using the following formula. p (t k ): ； Where C is the output state matrix.

4. The non-cooperative robot servo anti-interference control method for dealing with visual sampling delay according to claim 1, characterized in that, The process involves performing state estimation based on the values ​​input to the observer, compensating for time delays, and simultaneously expanding the total UAV disturbance into states and estimating them. The output is a full state vector comprising the delayed UAV state estimate and the disturbance estimate, including: Calculate the current control time t using the following formula. k Full-state estimation vector : ； in, This is an estimate of the error between the relative position and the expected relative position; This is an estimate of the derivative of the error between the relative position and the expected relative position; This is the estimated value of the disturbance; The data consists of a mixture of predicted and estimated observations; Q1 is the first state matrix. This represents the new state after expansion; M is a multiple of the sampling time; T is the sampling time. Q1 represents the control input before time 2MT; Q2 represents the second state matrix. Q1 represents the control quantity before time MT; Q2 represents the third state matrix.

5. The non-cooperative robot servo anti-interference control method for dealing with visual sampling delay according to claim 4, characterized in that, The step of determining the control quantity for the current control cycle based on the full state vector and a preset control law to control the UAV to land on the target mobile platform includes: Calculate the control quantity u(t) for the current control cycle using the following formula. k ): ； Where K is the overall controller gain; k1 is the first controller gain; and k2 is the second controller gain. The current control time t k The full-state estimation vector.

6. A non-cooperative robot servo anti-interference control system for coping with visual sampling delay, characterized in that, include: The first determining module is used to determine the relative position measurement values ​​between the UAV and the target mobile platform; The judgment module is used to determine whether the current moment is the sampling moment of the UAV's airborne camera in each control cycle of the airborne observer; The second determining module is used to determine, when the judging module determines that the current time is the sampling time, to input the relative position measurement value of the target moving platform to the observer; The third determination module is used to determine, when the judgment module determines that the current time is not the sampling time, to generate a predicted value of the relative position between the UAV and the target mobile platform based on the UAV dynamics model and the state estimate value of the previous control cycle, and input it to the observer. The full state vector generation module is used to perform state estimation based on the values ​​input to the observer and compensate for time delay. At the same time, it expands the total disturbance of the UAV into a state and estimates it, and outputs a full state vector including the UAV state estimate and the disturbance estimate after time delay compensation. The fourth determination module is used to determine the control quantity of the current control cycle based on the full state vector and the preset control law, so as to control the UAV to land on the target mobile platform; wherein, the control law includes feedforward compensation for the disturbance estimate.

7. The non-cooperative robot servo anti-interference control system for coping with visual sampling delay according to claim 6, characterized in that, The first determining module includes: The first calculation unit is used to calculate the translation matrix of the spatial relationship between the landmark coordinate system M and the camera coordinate system C according to the following formula. : ； Among them, P C The 3D coordinates of the AprilTag origin in the camera coordinate system C; P is the rotation matrix representing the spatial relationship between the landmark coordinate system M and the camera coordinate system C; m The three-dimensional coordinates of the AprilTag origin in the landmark coordinate system M; The second calculation unit is used to calculate the relative position measurement value r between the UAV and the target mobile platform based on the translation matrix of the spatial relationship between the landmark coordinate system M and the camera coordinate system C. pt : ； Where R is the rotation matrix, used to describe the coordinate transformation from the UAV coordinate system B to the inertial coordinate system W; This is the transformation matrix between the camera coordinate system and the UAV coordinate system.

8. The non-cooperative robot servo anti-interference control system for coping with visual sampling delay according to claim 6, characterized in that, The third determining module includes: Building blocks are used to construct expressions for the dynamic model of the UAV: ； in, The current control time t k The predicted state; e is the natural constant; T is the sampling time; For the previous time t k-1 The estimated state of the unmanned aerial vehicle; The new state matrix is ​​formed by discretizing the original state matrix B using Euler; u(t) k-1 ) represents the previous time t k-1 The control quantity; The third calculation unit is used to calculate the predicted relative position y between the UAV and the target mobile platform according to the following formula. p (t k ): ； Where C is the output state matrix.

9. The non-cooperative robot servo anti-interference control system for coping with visual sampling delay according to claim 6, characterized in that, The full-state vector generation module includes: The fourth calculation unit is used to calculate the current control time t according to the following formula. k Full-state estimation vector : ； in, This is an estimate of the error between the relative position and the expected relative position; This is an estimate of the derivative of the error between the relative position and the expected relative position; This is the estimated value of the disturbance; The data consists of a mixture of predicted and estimated observations; Q1 is the first state matrix. This represents the new state after expansion; M is a multiple of the sampling time; T is the sampling time. Q1 represents the control input before time 2MT; Q2 represents the second state matrix. Q1 represents the control quantity before time MT; Q2 represents the third state matrix.

10. The non-cooperative robot servo anti-interference control system for coping with visual sampling delay according to claim 9, characterized in that, The fourth determining module includes: The fifth calculation unit is used to calculate the control quantity u(t) for the current control cycle according to the following formula. k ): ； Where K is the overall controller gain; k1 is the first controller gain; and k2 is the second controller gain. The current control time t k The full-state estimation vector.

Images