A trajectory prediction method and medium for remote operation of unmanned systems

Abstract

This invention relates to the field of teleoperation and human-computer interaction technology for unmanned systems, and discloses a trajectory prediction method in the teleoperation of unmanned systems. The method includes: a front-end trajectory prediction and display module acquiring target parameters corresponding to user interaction input and sending them to a back-end trajectory generation module in a non-blocking manner; the back-end trajectory generation module constructing and solving a trajectory planning problem based on the most recently received target parameters, the current state of the unmanned system, and system constraints to generate a baseline trajectory that satisfies the system constraints; simultaneously generating trajectory correction model parameters to describe the relationship between changes in target parameters and changes in the baseline trajectory; and encapsulating the baseline trajectory, trajectory correction model parameters, and historical target parameters used in generating the baseline trajectory into a trajectory data frame and sending it to the front-end trajectory prediction and display module. The front-end prediction of this invention is completely independent of the delayed calculation process of the back-end, eliminating visual stuttering and operational lag caused by computational or network latency.

Description

Technical Field

[0001] This invention relates to the field of remote operation and human-computer interaction technology for unmanned systems, specifically to a trajectory prediction method and medium for remote operation of unmanned systems. Background Technology

[0002] As the complexity of unmanned systems such as drones and autonomous vehicles increases, trajectory planning based on optimal control has become a core technology. In teleoperation or human-in-the-loop mission planning scenarios, operators typically need to drag target points on a map interface and expect to observe in real time the continuous trajectory generated by the system that meets dynamic constraints and obstacle avoidance requirements.

[0003] However, generating an optimal trajectory that satisfies complex physical constraints (such as maximum overload and turning radius) and environmental constraints (such as obstacle avoidance) typically requires solving a complex nonlinear programming problem, which is computationally time-consuming, usually on the order of tens to hundreds of milliseconds. Furthermore, in cloud computing, edge computing, or remote control architectures, trajectory calculation modules are often deployed on remote servers or onboard high-performance computers, and network communication introduces additional, non-negligible transmission latency.

[0004] Current interaction solutions mainly suffer from two types of defects:

[0005] 1. Real-time feedback scheme based on geometric interpolation: To ensure the smoothness of the front-end display (such as a 60Hz refresh rate), the interface uses simple linear or spline interpolation to connect the starting point with the current mouse position. This method completely ignores the dynamic constraints of the physical model and environmental obstacle information, resulting in the generated preview trajectory often "passing through walls" or presenting a broken line that the aircraft cannot execute, forming false, unsafe, or even misleading visual feedback.

[0006] 2. Replanning scheme based on synchronous waiting: The front-end completely depends on the calculation results of the back-end. The front-end only updates and displays after the back-end solver completes the calculation and returns the data. To speed up the process, some techniques employ a warm-start strategy. However, this scheme is essentially a serial / synchronous architecture, where the interaction response frequency of the front-end is forcibly locked by the calculation frequency and network round-trip time of the back-end. Under high load or network fluctuations, the interactive interface will exhibit significant stuttering, lag, and response interruptions, severely impacting the operator's decision-making efficiency and user experience, and failing to meet the real-time requirements of highly dynamic environments.

[0007] In summary, existing technologies cannot simultaneously ensure the real-time smoothness of human-computer interaction and the physical consistency (security) of trajectory prediction under conditions of high computational / network latency. Therefore, there is an urgent need for a real-time trajectory prediction and interaction method that can fundamentally decouple the front-end and back-end and resist latency. Summary of the Invention

[0008] This invention aims to overcome the shortcomings of existing technologies and provide a trajectory prediction method and medium for remote operation of unmanned systems. Through innovation in system architecture and data processing flow, the front-end interactive interface can still provide operators with continuous, stable trajectory prediction feedback that conforms to physical and environmental constraints, even when there is a significant delay in back-end trajectory planning. This significantly improves the real-time performance, security, and availability of the system in high-latency environments.

[0009] This invention is achieved through the following technical solution:

[0010] A trajectory prediction method for remote operation of an unmanned system, the method being executed collaboratively by a back-end trajectory generation module and a front-end trajectory prediction and display module that operate asynchronously, comprising:

[0011] S1. The front-end trajectory prediction and display module obtains the target parameters corresponding to the user interaction input and sends them to the back-end trajectory generation module in a non-blocking manner.

[0012] S2. The backend trajectory generation module constructs and solves the trajectory planning problem based on the recently received target parameters, the current state of the unmanned system, and system constraints, generates a baseline trajectory that satisfies the system constraints, and generates trajectory correction model parameters to describe the relationship between changes in target parameters and changes in the baseline trajectory. The baseline trajectory, the trajectory correction model parameters, and the historical target parameters used when generating the baseline trajectory are encapsulated into a trajectory data frame and sent to the frontend trajectory prediction and display module.

[0013] S3. The front-end trajectory prediction and display module performs the following steps in a loop at a refresh rate higher than the trajectory planning and solving frequency of the back-end trajectory generation module:

[0014] a) Acquire current user interaction input in real time and determine real-time target parameters;

[0015] b) Read the most recently received trajectory data frame from the cache;

[0016] c) Based on the difference between the real-time target parameters and the historical target parameters, and the trajectory correction model parameters, perform algebraic operations on the baseline trajectory to generate a predicted trajectory;

[0017] In the process of generating the predicted trajectory, the front-end trajectory prediction and display module does not construct a trajectory planning problem, perform constraint solving, or perform optimization iteration. The generation of the predicted trajectory is completed only by quickly correcting the parameters of the baseline trajectory and the trajectory correction model. The predicted trajectory is continuously updated and used for interface display or control output before the back-end trajectory generation module completes the next trajectory planning calculation and returns a new trajectory data frame.

[0018] As an optimization, the trajectory correction model parameters are used to characterize the response of the baseline trajectory to changes in the target parameters, and to establish a mapping relationship between the change in the target parameters and the change in the trajectory state.

[0019] As an optimization, the trajectory correction model parameters are a local linear mapping model, including the Jacobian matrix, sensitivity matrix, or a parameterized mapping model obtained based on the solution results of the trajectory planning problem, excluding the Jacobian matrix and sensitivity matrix.

[0020] As an optimization, the front-end trajectory prediction and display module performs the fast correction calculation in the following way when generating the predicted trajectory:

[0021] Calculate the difference between real-time target parameters and historical target parameters;

[0022] Based on the differences and the trajectory correction model parameters, algebraic operations of matrices and vectors are performed on the baseline trajectory;

[0023] The calculation result is superimposed on the baseline trajectory to generate the predicted trajectory.

[0024] As an optimization, the front-end trajectory prediction and display module also includes stability control of the trajectory correction amount when generating the predicted trajectory, so as to suppress the divergence or oscillation of the predicted trajectory caused by rapid changes in the target parameters.

[0025] As an optimization, the stability control includes at least one of the following methods:

[0026] The amplitude of the trajectory correction is limited;

[0027] A damping factor is applied to the trajectory correction amount;

[0028] The predicted trajectory is then subjected to temporal smoothing filtering.

[0029] As an optimization, the triggering method for the backend trajectory generation module to perform trajectory planning calculations includes any of the following:

[0030] Triggered when the front-end trajectory prediction and display module confirms the target parameters;

[0031] The backend trajectory generation module triggers autonomously according to a preset cycle.

[0032] As an optimization, the trajectory data frames are transmitted through an asynchronous, non-blocking communication mechanism, and the front-end trajectory prediction and display module only retains the most recently received trajectory data frames for prediction calculation.

[0033] As an optimization, the system constraints include at least one of the following: dynamic constraints of the unmanned system, kinematic constraints, and obstacle collision avoidance constraints in the flight environment.

[0034] The present invention also discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement a trajectory prediction method in the teleoperation of an unmanned system as described above.

[0035] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0036] This invention employs an asynchronous and decoupled architecture between the front-end and back-end. The front-end prediction is completely independent of the back-end latency calculation process, enabling it to respond to operator input at an extremely high refresh rate (such as 60Hz), thus completely eliminating visual stuttering and operational lag caused by computation or network latency.

[0037] The front-end prediction of this invention is not a simple geometric interpolation, but rather a baseline trajectory generated at the back end that strictly satisfies the constraints, and is corrected according to a sensitivity model describing the physical characteristics of the system. Therefore, within a reasonable range of variation of the target parameters, the predicted trajectory can always maintain an approximate satisfaction of physical and environmental constraints, effectively avoiding dangerous false feedback such as "wall-penetrating" errors.

[0038] The complex trajectory optimization solution of this invention is executed only in the backend, while the frontend only performs simple algebraic operations such as matrix multiplication and vector addition, resulting in extremely low computational overhead. This allows the invention to be easily deployed on portable terminals, embedded devices, or web frontends with limited computing power, reducing system hardware costs and deployment barriers.

[0039] This invention introduces a stability control mechanism (such as amplitude limiting and damping) into the front-end prediction, which can effectively suppress the divergence or oscillation of the predicted trajectory that may occur when the target parameters change drastically or approach the constraint boundary, thus ensuring the reliable operation of the system in various interactive scenarios.

[0040] This invention eliminates display jumps that may be caused by backend data packet updates by performing time-domain smoothing on the predicted trajectory, making the trajectory feedback smooth and natural, reducing the cognitive load on the operator, and improving the efficiency of human-computer interaction. Attached Figure Description

[0041] The accompanying drawings, which are included to provide a further understanding of embodiments of the invention and form part of this application, do not constitute a limitation thereof. In the drawings:

[0042] Figure 1 A system for implementing the trajectory prediction method in the teleoperation of an unmanned system as described in this invention;

[0043] Figure 2 This is a flowchart of a trajectory prediction method for remote operation of an unmanned system as described in this invention. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0045] like Figure 1 As shown, the system architecture of this invention mainly includes three core modules that work asynchronously and collaboratively: a backend trajectory generation module, a data transmission and caching module, and a frontend trajectory prediction and display module. The backend trajectory generation module and the frontend trajectory prediction and display module run in parallel at different time scales, and non-blocking data exchange is achieved through the data transmission and caching module, forming an asynchronous and decoupled trajectory prediction interaction architecture between the frontend and backend.

[0046] like Figure 2 As shown, the trajectory prediction method for remote operation of an unmanned system according to the present invention includes the following core steps in its overall process. These steps are executed collaboratively by a back-end trajectory generation module and a front-end trajectory prediction and display module that operate asynchronously with each other:

[0047] Step S1: Front-end target parameter collection and asynchronous reporting.

[0048] The front-end trajectory prediction and display module collects real-time interactive input from the operator on the human-computer interface (e.g., dragging target points via mouse or touchscreen) and parses it to obtain the target parameters expected by the user. (For example, the three-dimensional spatial coordinates of the target point).

[0049] The front-end trajectory prediction and display module transmits the target parameters via non-blocking communication. The data is asynchronously sent to the backend trajectory generation module. The non-blocking communication method includes, but is not limited to, UDP protocol, message queue, or shared memory mechanism. This sending action does not block the subsequent display refresh process of the frontend. After sending, the frontend does not need to wait for backend confirmation or return results, and can continue to execute the interactive acquisition and prediction display task of the next frame.

[0050] Step S2: Backend trajectory planning calculation, sensitivity model generation, and data frame encapsulation.

[0051] The backend trajectory generation module operates at a relatively low frequency (e.g., 1-20Hz). When a trigger condition is met, this module generates the target parameter based on the most recently received target parameter from the front end. Given the current state of the unmanned system (including position, velocity, attitude, etc.) and preset system constraints (including dynamic constraints, kinematic constraints, environmental obstacle collision avoidance constraints, etc.), construct and solve a complete trajectory planning problem.

[0052] The trajectory planning problem typically manifests as a nonlinear optimal control problem or a nonlinear programming problem. The backend trajectory generation module uses a solver (e.g., direct collocation method, multiple target method, convex optimization method, etc.) to iteratively solve this problem, generating a trajectory from the current state to the target point. The optimal sequence of states that strictly satisfies all system constraints, i.e., the baseline trajectory Let the baseline trajectory be represented as:

[0053] ;

[0054] in, The vector represents the state vector (including position, velocity, etc.) of the trajectory at the i-th discrete moment, where N is the total number of discrete points.

[0055] The backend trajectory generation module obtains the baseline trajectory after solving. Simultaneously, trajectory correction model parameters M are generated to describe the relationship between changes in target parameters and changes in the baseline trajectory. In this embodiment, the trajectory correction model parameters M are Jacobian matrices characterizing the local linear mapping relationship of the baseline trajectory with changes in target parameters:

[0056] ;

[0057] This matrix quantifies the optimal solution. This refers to the local sensitivity information of how the optimal trajectory X should respond when the target parameter p undergoes a small perturbation. The calculation of M is a byproduct of the trajectory planning solution process and can be approximated using the second-order information of the Lagrangian function (Hessian matrix) or directly calculated using automatic differentiation techniques. Its computational cost is far less than solving the entire trajectory planning problem again.

[0058] Subsequently, the backend trajectory generation module will generate the baseline trajectory obtained in this calculation. The trajectory correction model parameters M and the historical target parameters used when generating the baseline trajectory. All three are encapsulated into a single structured trajectory data frame. This data frame is sent to the front-end trajectory prediction and display module via an asynchronous, non-blocking communication method through the data transmission and caching module. Once the back-end completes the transmission, it immediately releases resources and waits for the next trigger condition, without requiring confirmation or feedback from the front-end.

[0059] Step S3: Front-end high-frequency rapid prediction and output.

[0060] The front-end trajectory prediction and display module operates at a refresh rate (e.g., 30-120Hz) that is much higher than the solution frequency of the back-end trajectory generation module. Within each refresh cycle, this module performs the following sub-steps independently of the back-end:

[0061] a) Real-time target parameter acquisition: Real-time acquisition of current user interaction input, parsed to obtain real-time target parameters. .

[0062] b) Read the latest data frame: Read the most recently received trajectory data frame from the local cache and unpack it to obtain the reference trajectory in the frame. Trajectory correction model parameters M and corresponding historical target parameters The cache employs a "keep only the latest frame" update strategy, ensuring that the front end always makes predictions based on the latest baseline information.

[0063] c) Rapid correction calculation and prediction trajectory generation:

[0064] First, calculate the change in the target parameter:

[0065] ;

[0066] Secondly, based on the trajectory correction model parameters M, the baseline trajectory is... Perform algebraic operations to generate predicted trajectories. :

[0067] ;

[0068] The calculation process involves only one matrix-vector multiplication and one vector addition, which are pure algebraic operations, taking only microseconds. It is particularly important to emphasize that during this trajectory generation process, the front-end trajectory prediction and display module does not construct any trajectory planning problem, perform any constraint solving, or conduct any optimization iterations. This is the fundamental difference between this invention and existing technologies (such as hot-start replanning and simplified model fast solutions).

[0069] The predicted trajectory The backend trajectory generation module is based on the real-time target parameters Before completing the next trajectory planning calculation and returning a new trajectory data frame, the front-end trajectory prediction and display module continuously updates the data and uses it for real-time interface display or downlink control output.

[0070] Step S4: Stability control and smooth display of the predicted trajectory.

[0071] To further enhance the robustness and user experience of the system under extreme interaction scenarios, in a preferred embodiment of the present invention, the front-end trajectory prediction and display module 103 includes at least one of the following optimization processes after generating the predicted trajectory and before outputting and displaying it:

[0072] (1) Stability control:

[0073] To prevent the predicted trajectory from diverging or becoming physically infeasible due to linear approximation errors when the target parameters change significantly or environmental constraints exhibit strong nonlinearity, intermediate corrections are made to the algebraic operations. Implement stability control.

[0074] Amplitude limiting method: ,in This is the preset maximum allowable correction value.

[0075] Damping weighting method: ,in The damping factor, .

[0076] (2) Temporal smoothing:

[0077] To eliminate visual jumps that may be caused by backend trajectory data frame updates, the predicted trajectory... Perform time-domain filtering. For example, use a first-order exponential moving average filter:

[0078] ;

[0079] in, Smoothing coefficient , The final displayed trajectory is for the k-th frame. Display the trajectory of the previous frame.

[0080] The trajectory, after the above stability control and / or smoothing processing, is finally rendered on the human-computer interaction interface, providing the operator with continuous, smooth, stable and physically consistent visual feedback.

[0081] System operating logic: As the operator continuously drags the target point, the front end continuously updates the data based on "old" data packets (based on historical target points). (Calculation) and "new" mouse position Perform rapid predictions. Simultaneously, the backend asynchronously receives data reported from the frontend. (or a snapshot thereof), and based on this, initiate new optimization calculations to generate new data packets (based on the new target point). (Calculation) Update the front end. In this way, the high-frequency prediction of the front end and the low-frequency optimization of the back end can be carried out in parallel, and the back end latency is perfectly masked.

[0082] Compared with existing technologies, this invention, by constructing an asynchronous and decoupled trajectory prediction interaction mechanism between the front-end and back-end, has at least the following beneficial effects:

[0083] (1) Improve the real-time performance of human-computer interaction in high-latency environments

[0084] This invention decouples the trajectory planning calculation process from the front-end interactive response process, so that the front-end trajectory prediction and display process no longer depends on the real-time return of the back-end trajectory planning results. By performing rapid correction calculations on the baseline trajectory based on the trajectory correction model parameters at the front end, the front end can maintain continuous and stable trajectory prediction and display even when there are delays in back-end calculations or communication, thereby significantly improving the real-time interactive performance of the system in high-latency computing environments.

[0085] (2) Ensure the physical consistency and security of the predicted trajectory

[0086] The front-end predicted trajectory of this invention is not based on simple geometric interpolation, but is corrected based on the baseline trajectory generated in the back-end and its corresponding trajectory correction model parameters.

[0087] Therefore, the predicted trajectory can inherit the physical and environmental constraints introduced in the back-end trajectory planning process, maintain the physical feasibility of the trajectory as the target parameters change, and effectively avoid trajectory distortion or potential collision risks caused by interpolation prediction in existing technologies.

[0088] (3) Reduce the front-end computing burden and improve the system deployment flexibility.

[0089] In this invention, the complex trajectory planning or optimization process is completed by the backend computing module, while the frontend trajectory prediction process only involves algebraic operations on predefined model parameters. This design significantly reduces the computational complexity of the frontend, enabling the invention to be deployed on terminal devices or embedded platforms with limited computing power, thereby improving system deployment flexibility and reducing hardware costs.

[0090] (4) Enhance the system's operational stability in complex interactive scenarios

[0091] This invention introduces a stability control mechanism during the front-end trajectory correction process to limit the amount of trajectory correction. This ensures that the predicted trajectory can maintain numerical stability and display continuity even when the target parameters change significantly or the environmental constraints exhibit strong nonlinear characteristics, thereby improving the overall robustness of the system in complex interactive scenarios.

[0092] (5) Improve the operator's interactive experience

[0093] By smoothing the predicted trajectory, this invention effectively reduces display jumps caused by backend trajectory updates or changes in environmental conditions, making the frontend trajectory feedback more continuous and natural, thereby reducing the cognitive load on operators and improving the overall interactive experience.

[0094] It should be understood that the core of this invention does not lie in the specific calculation method of trajectory correction model parameters, but in the trajectory prediction interactive engineering mechanism that generates a baseline trajectory and its corresponding trajectory correction model parameters through the backend, and performs rapid correction based on the model parameters by the frontend without re-executing trajectory planning calculation.

[0095] Without altering the aforementioned engineering mechanism, those skilled in the art can make equivalent substitutions or modifications to some implementation methods. The following alternative embodiments are only used to illustrate the scope of implementation of the present invention and do not constitute a limitation on the scope of protection of the present invention.

[0096] (1) Regarding the generation method of trajectory correction model parameters

[0097] In one embodiment of the present invention, the trajectory correction model parameters can be obtained based on the analysis of the local characteristics of the trajectory planning problem. Without changing the overall mechanism of the present invention—where the backend generates the trajectory correction model and the frontend performs rapid correction based on the model—the trajectory correction model parameters can also be generated in other ways, such as parameter mapping relationships obtained based on different numerical analysis methods or modeling techniques.

[0098] The above alternatives only change the method of generating trajectory correction model parameters and do not affect the engineering process of supporting rapid front-end prediction and execution through trajectory correction model.

[0099] (2) Implementation of the front-end trajectory prediction model

[0100] In one embodiment of the present invention, the front-end trajectory prediction performs a fast correction calculation on the baseline trajectory based on the trajectory correction model parameters. Without changing the technical concept that the front-end does not re-execute trajectory planning calculations but only performs predictions based on model parameters, the specific implementation of the front-end prediction model can be adjusted according to system requirements.

[0101] The above alternative implementation methods only affect the implementation form of the prediction model and do not change the core engineering mechanism of front-end and back-end decoupling and front-end fast execution of the present invention.

[0102] (3) Modeling methods for environmental constraints

[0103] In one embodiment of the present invention, environmental constraints are incorporated into the backend trajectory planning calculation process to generate baseline trajectory data that meets the environmental constraints. Without altering the characteristic of the present invention that the baseline trajectory inherits environmental constraints, the specific modeling form of the environmental constraints can be adjusted according to the application scenario.

[0104] (4) Regarding the system deployment architecture

[0105] In one embodiment of the present invention, the backend trajectory calculation module and the frontend trajectory prediction module are deployed on different computing nodes. Without changing the asynchronous collaboration mechanism between the frontend and backend, the above modules can also be deployed on different processing units of the same physical device or in different computing environments.

[0106] Meanwhile, the key technical point of this invention does not lie in the specific trajectory optimization algorithm or mathematical solution method itself, but in a trajectory prediction interactive engineering mechanism for high-latency computing environments. This mechanism, through system architecture and data flow design, effectively decouples the trajectory planning calculation process from the front-end interactive response process. Its core technical points include:

[0107] (1) Asynchronous decoupling of front-end and back-end trajectory prediction architecture

[0108] This invention divides the trajectory planning process into two independent processing stages: a back-end low-frequency high-precision calculation and a front-end high-frequency fast prediction execution. This eliminates the dependence of the front-end trajectory prediction and display process on the real-time return of the back-end trajectory planning results, thereby eliminating the direct impact of computational and communication delays on interactive response from the system architecture level.

[0109] (2) Trajectory correction model generation mechanism based on reference trajectory

[0110] While generating baseline trajectory data that satisfies physical and environmental constraints, the backend calculation module simultaneously generates trajectory correction model parameters to describe the relationship between changes in target parameters and trajectory changes. These trajectory correction model parameters, as part of the backend calculation results, form a one-to-one correspondence with the baseline trajectory, thus providing the frontend with a directly applicable basis for trajectory correction without re-executing trajectory planning calculations.

[0111] (3) Unified trajectory data frame and non-blocking transmission mechanism

[0112] This invention defines a unified trajectory data frame structure that includes baseline trajectory data, trajectory correction model parameters, and corresponding target parameter information, and transmits it asynchronously and non-blockingly. The front-end interaction module performs prediction calculations only based on the latest received trajectory data frame, thereby avoiding blocking coupling between the front-end and back-end and improving the system's robustness in high-latency or network fluctuation environments.

[0113] (4) Front-end fast trajectory correction execution mechanism

[0114] The front-end interaction module does not perform trajectory planning or optimization. Instead, based on the received trajectory correction model parameters, it performs rapid correction calculations on the baseline trajectory data to generate a predicted trajectory. This prediction process only involves algebraic operations on predefined model parameters, enabling the front-end trajectory prediction and display to run at a refresh rate far exceeding the back-end computing frequency.

[0115] (5) Stability control mechanism of predicted trajectory

[0116] To prevent the predicted trajectory from diverging or becoming numerically unstable in areas with large changes in target parameters or strong nonlinear environmental constraints, this invention introduces an amplitude constraint or damping control mechanism during the front-end trajectory correction process to limit the amount of trajectory correction, thereby ensuring the physical feasibility and display continuity of the predicted trajectory.

[0117] Through the synergistic effect of the above-mentioned engineering mechanisms, this invention achieves continuous output of trajectory prediction results in a high-latency computing environment without increasing the front-end computing burden, while taking into account interactive real-time performance, trajectory physical consistency and system operation stability.

[0118] Example 2 also discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements a trajectory prediction method in the teleoperation of an unmanned system as described above.

[0119] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A trajectory prediction method for remote operation of an unmanned system, characterized in that, The method is executed collaboratively by a backend trajectory generation module and a frontend trajectory prediction and display module that operate asynchronously, including: S1. The front-end trajectory prediction and display module obtains the target parameters corresponding to the user interaction input and sends them to the back-end trajectory generation module in a non-blocking manner. S2. The backend trajectory generation module constructs and solves the trajectory planning problem based on the recently received target parameters, the current state of the unmanned system, and system constraints, generates a baseline trajectory that satisfies the system constraints, and generates trajectory correction model parameters to describe the relationship between changes in target parameters and changes in the baseline trajectory. The baseline trajectory, the trajectory correction model parameters, and the historical target parameters used when generating the baseline trajectory are encapsulated into a trajectory data frame and sent to the frontend trajectory prediction and display module. S3. The front-end trajectory prediction and display module performs the following steps in a loop at a refresh rate higher than the trajectory planning and solving frequency of the back-end trajectory generation module: a) Acquire current user interaction input in real time and determine real-time target parameters; b) Read the most recently received trajectory data frame from the cache; c) Based on the difference between the real-time target parameters and the historical target parameters, and the trajectory correction model parameters, perform algebraic operations on the baseline trajectory to generate a predicted trajectory; In the process of generating the predicted trajectory, the front-end trajectory prediction and display module does not construct a trajectory planning problem, perform constraint solving, or perform optimization iteration. The generation of the predicted trajectory is completed only by quickly correcting the parameters of the baseline trajectory and the trajectory correction model. The predicted trajectory is continuously updated and used for interface display or control output before the back-end trajectory generation module completes the next trajectory planning calculation and returns a new trajectory data frame.

2. The trajectory prediction method for remote operation of an unmanned system according to claim 1, characterized in that, The trajectory correction model parameters are used to characterize the response of the baseline trajectory to changes in the target parameters, and to establish a mapping relationship between the changes in the target parameters and the changes in the trajectory state.

3. The trajectory prediction method for remote operation of an unmanned system according to claim 2, characterized in that, The trajectory correction model parameters are a local linear mapping model, including the Jacobian matrix, sensitivity matrix, or a parameterized mapping model obtained based on the solution results of the trajectory planning problem, excluding the Jacobian matrix and sensitivity matrix.

4. The trajectory prediction method for remote operation of an unmanned system according to claim 1, characterized in that, When generating the predicted trajectory, the front-end trajectory prediction and display module performs the rapid correction calculation in the following manner: Calculate the difference between real-time target parameters and historical target parameters; Based on the differences and the trajectory correction model parameters, algebraic operations of matrices and vectors are performed on the baseline trajectory; The calculation result is superimposed on the baseline trajectory to generate the predicted trajectory.

5. The trajectory prediction method for remote operation of an unmanned system according to claim 1, characterized in that, When generating the predicted trajectory, the front-end trajectory prediction and display module also includes stability control of the trajectory correction amount to suppress the divergence or oscillation of the predicted trajectory caused by rapid changes in the target parameters.

6. The trajectory prediction method for remote operation of an unmanned system according to claim 5, characterized in that, The stability control includes at least one of the following methods: The amplitude of the trajectory correction is limited; A damping factor is applied to the trajectory correction amount; The predicted trajectory is then subjected to temporal smoothing filtering.

7. The trajectory prediction method for remote operation of an unmanned system according to claim 1, characterized in that, The triggering methods for the backend trajectory generation module to perform trajectory planning calculations include any of the following: Triggered when the front-end trajectory prediction and display module confirms the target parameters; The backend trajectory generation module triggers autonomously according to a preset cycle.

8. The trajectory prediction method for remote operation of an unmanned system according to claim 1, characterized in that, The trajectory data frames are transmitted through an asynchronous, non-blocking communication mechanism, and the front-end trajectory prediction and display module only retains the most recently received trajectory data frames for prediction calculation.

9. The trajectory prediction method for remote operation of an unmanned system according to claim 1, characterized in that, The system constraints include at least one of the following: dynamic constraints, kinematic constraints, and obstacle collision avoidance constraints in the flight environment of the unmanned system.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements a trajectory prediction method for remote operation of an unmanned system as described in any one of claims 1 to 9.

Images