Robot advancing method and system based on inertial sensor

Abstract

The invention relates to a robot advancing method and system based on an inertial sensor, and belongs to the technical field of robot autonomous navigation and inertial guidance control. The method comprises the following steps: acquiring advancing parameter data of a robot, and analyzing a track generation mode in the advancing process of the robot to obtain autonomous advancing track data; constructing a marching analysis model, inputting the robot marching parameter data into the marching analysis model to analyze inertial navigation information, and outputting a robot marching instruction; the stability of the posture state in the advancing process is verified, a robot advancing instruction is optimized through a path optimization protocol, and an instruction optimization set of an advancing path is generated; and constructing an offset compensation mechanism, selecting an instruction optimization parameter according to the advancing path when the advancing path is detected to deviate, re-inputting the instruction optimization parameter into the advancing analysis model, and re-matching a newly selected advancing path.

Description

Technical Field

[0001] This invention belongs to the field of robot autonomous navigation and inertial guidance control technology, specifically relating to a robot movement method and system based on inertial sensors. Background Technology

[0002] In the field of autonomous movement control for mobile robots, common navigation methods mainly rely on LiDAR, visual recognition, or a single inertial measurement unit (IMU) for environmental perception and path tracking. However, visual and LiDAR are prone to recognition errors in scenarios with changing lighting, occlusion interference, and complex structures, failing to guarantee path stability during continuous movement. While single-IMU-based inertial navigation methods can continuously output attitude data without external calibration information, the cumulative drift error of inertial integrals can easily lead to heading deviation and trajectory divergence during long-distance operation. Furthermore, most existing path planning algorithms generate planned paths only based on target points, lacking the ability to adaptively identify and correct deviations from the actual trajectory during movement. They cannot compensate for instantaneous disturbances in the robot's inertial navigation parameters in real time, resulting in problems such as attitude jitter, steering lag, or failure to automatically recover from trajectory deviations during operation, affecting path execution accuracy and control stability. Therefore, there is an urgent need for a robot movement method that combines magnetometer-assisted attitude fusion and trajectory curve matching detection mechanism to achieve dynamic identification of off-track trajectories, command-level correction, and path re-matching, thereby constructing an autonomous movement control system with closed-loop compensation capability. Summary of the Invention

[0003] To address the aforementioned problems in the existing technology, this invention provides a robot movement method based on inertial sensors. The objective of this invention can be achieved through the following technical solutions: S1: Obtain robot travel parameter data, construct a path trajectory network based on the robot travel parameter data, and detect the path guidance of the path trajectory network through a trajectory curve matching algorithm. Analyze the trajectory generation mode during robot travel to obtain autonomous travel trajectory data. S2: Introduce a magnetometer-assisted attitude fusion algorithm to construct a travel analysis model, and allocate a trajectory generation mode according to the autonomous travel trajectory data. Input the robot travel parameter data into the travel analysis model to analyze the inertial navigation information and output robot travel commands. S3: Execute the robot's movement instructions, verify the stability of the posture state during movement, adjust the movement path with posture deviation based on the autonomous movement trajectory data, optimize the robot's movement instructions through the path optimization protocol, and generate an optimized set of movement path instructions. S4: Construct an offset compensation mechanism, analyze the instruction optimization set, and when an offset in the travel trajectory is detected, select instruction optimization parameters according to the travel path, and re-input the instruction optimization parameters into the travel analysis model to re-match the newly selected travel trajectory path.

[0004] Specifically, the method for obtaining the robot's travel parameter data is as follows: the robot's travel signal and odometer displacement information are timestamped, and attitude jitter and inertial drift interference are filtered out based on a low-order filtering algorithm to generate robot travel parameter data for constructing a path trajectory network.

[0005] Specifically, the path trajectory network divides the pose calculation points of the robot's continuous movement process into nodes at discrete time intervals. Each effective pose node is used as a trajectory graph node. The graph structure connection edges are constructed based on the displacement vector change characteristics between nodes, and the guiding relationship between nodes is recorded to form a path trajectory network structure with path turning angle distribution and pose gradient information.

[0006] Specifically, the method for acquiring the autonomous travel trajectory data is as follows: The trajectory deviation index between the autonomous path and the reference path is calculated based on the trajectory curve matching algorithm. The real-time travel trajectory of the robot is inserted into the trajectory alignment matrix, and the dynamic deviation sequence between trajectory points is calculated by dynamic time warping. Based on the dynamic deviation sequence, a trajectory deviation index is obtained. The trajectory deviation index is analyzed. When the deviation exceeds a set threshold, the interval is marked as a non-guided mode segment, and trajectory segments that meet the guided mode constraints are extracted to obtain autonomous travel trajectory data.

[0007] Specifically, the method for constructing the movement parsing model is as follows: The robot's travel parameter data is synchronized using timestamps. An initial attitude calculation dataset is constructed based on the timestamps, and geomagnetic vector information is used to analyze the initial attitude calculation dataset to obtain the attitude direction reference. An attitude calculation framework is constructed based on the attitude orientation reference. A complementary filtering algorithm is used to quickly correct short-term attitude changes, and an extended Kalman filter is introduced to constrain the attitude stability of the magnetometer data. The attitude deflection error caused by gyroscope integral drift is corrected, and an intermediate attitude calculation state parameter set is generated. The intermediate state parameter set of the attitude calculation is analyzed, an inertial navigation attitude fusion matrix is ​​established, the fusion matrix is ​​mapped to the robot's travel parameter data, and the path trajectory is matched according to the trajectory estimation and deviation identification correlation function to construct a travel analysis model.

[0008] Specifically, the matching method for the trajectory generation pattern is as follows: Based on the trajectory dataset stored in the path trajectory network, the turning angle and speed change gradient of each trajectory node are analyzed to construct a trajectory feature parameter vector. The trajectory feature parameters are then combined with trajectory recognition labels to generate a trajectory pattern matching input set. The trajectory pattern matching input set is aligned with the feature vectors of the autonomous travel trajectory data. The feature similarity score is calculated by the trajectory curve matching algorithm, and the trajectory nodes are segmented and matched by Euclidean distance clustering operation to analyze the pattern type of the autonomous travel trajectory. Based on the trajectory pattern recognition labels, a mapping relationship is established between the trajectory generation pattern and the target trajectory segment in the path trajectory network. The identified pattern labels are bound to the corresponding trajectory segment index, and the trajectory segment index is fed back to the travel parsing model to match the trajectory generation pattern.

[0009] Specifically, the method for generating the instruction optimization set of the travel path is as follows: perform local path backtracking and reprojection calculation on the trajectory curve offset segment according to the path optimization protocol, remap the trajectory nodes to the center line of the fitted curve, and generate the travel path instruction optimization set based on the real-time attitude state analysis of the center line of the fitted curve.

[0010] Specifically, the path optimization protocol is based on a multi-constraint path adjustment strategy using a cost function. It incorporates trajectory deviation, cumulative inertial drift, attitude angle offset, and travel stability index as weighting factors of the cost function into the path optimization protocol. It also defines an instruction-level penalty function to analyze the stability of the path instruction set. When the stability of the path instruction set decreases, it triggers local path replanning logic.

[0011] Specifically, the output process of the robot's movement command is as follows: Based on the inertial navigation attitude fusion matrix output by the travel analysis model, the current attitude angle, heading angle and velocity vector data are analyzed, the attitude feature parameter set for path execution is extracted, and the attitude feature parameters are correlated and compared with the autonomous travel trajectory data to generate the trajectory parameters of the current target travel segment. Based on the trajectory parameters, the attitude angle data in the attitude feature parameter set of the path execution is solved by the path guidance instruction algorithm, and a set of travel attitude analysis control instructions is generated according to the solution results. The travel control pre-instruction set is output to the travel analysis model through a bus communication protocol, and real-time closed-loop correction is performed based on attitude feedback data during execution to output robot travel instructions that optimize the accuracy of travel trajectory matching.

[0012] Specifically, the method for constructing the offset compensation mechanism is as follows: The offset detection judgment function uses the trajectory curve matching residual and abnormal angular velocity peaks as offset triggering factors to generate an offset detection threshold range. When the distance offset between the real-time trajectory node and the trajectory centerline exceeds the set tolerance, the offset compensation mechanism is triggered to enter the compensation state. Under the compensated state, the sensor disturbance sequence corresponding to the offset interval is extracted, and combined with the heading angle correction vector corrected by the magnetometer, the observation noise covariance matrix and state transition gain coefficient of the internal filter of the analytical model are adaptively adjusted to construct the offset compensation mechanism.

[0013] Specifically, the magnetometer-assisted attitude fusion algorithm performs distortion calibration on the original three-axis magnetic field data output by the magnetometer by analyzing the ellipsoid fitting, constructs a debiased magnetic field vector, and maps it to the robot body coordinate system according to the geomagnetic reference direction to establish a reference coordinate framework for calculating the heading angle, thereby obtaining the magnetic distortion dynamic suppression coefficient table of the travel analytical model.

[0014] Specifically, a robot propulsion system based on inertial sensors is characterized by comprising: The navigation data acquisition module acquires robot navigation parameter data, constructs a path trajectory network based on the robot navigation parameter data, detects the path guidance of the path trajectory network through a trajectory curve matching algorithm, analyzes the trajectory generation mode during robot navigation, and obtains autonomous navigation trajectory data. Analytical model construction module: Introduces a magnetometer-assisted attitude fusion algorithm to construct a travel analytical model, and allocates a trajectory generation mode according to the autonomous travel trajectory data. Inputs the robot travel parameter data into the travel analytical model to analyze inertial navigation information and outputs robot travel commands. Path instruction optimization module: Executes the robot's movement instructions, verifies the stability of the posture state during movement, adjusts the movement path with posture deviation based on the autonomous movement trajectory data, optimizes the robot's movement instructions through the path optimization protocol, and generates an optimized set of movement path instructions; Closed-loop offset compensation module: Constructs an offset compensation mechanism, analyzes the instruction optimization set, and when an offset in the travel trajectory is detected, selects instruction optimization parameters according to the travel path and re-inputs the instruction optimization parameters into the travel analysis model to re-match the newly selected travel trajectory path.

[0015] The beneficial effects of this invention are as follows: This invention constructs a path trajectory network based on trajectory curve matching and introduces a magnetometer-assisted attitude fusion algorithm, enabling robots to achieve highly robust autonomous path construction and attitude perception without external vision or laser environment dependence, effectively reducing drift errors caused by inertial integration. Based on this, the proposed trajectory generation pattern matching and path optimization protocol can identify and adjust attitude disturbances, steering deviations, and trajectory deviations generated during robot movement in real time, avoiding the lag and unrecoverable path offset problems inherent in traditional fixed-path execution methods. Furthermore, the offset compensation mechanism constructed in this invention has closed-loop feedback capability, dynamically generating compensation parameters and re-injecting them into the movement analysis model after detecting trajectory errors, enabling the robot to automatically correct its direction of travel during continuous operation, improving path consistency and command execution stability over long periods. In summary, this invention achieves integrated control of trajectory construction, attitude fusion, path offset detection, and closed-loop correction during autonomous robot movement, improving autonomous navigation accuracy, reliability, and environmental adaptability, and has promising engineering application prospects. Attached Figure Description

[0016] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0017] Figure 1 This is a flowchart illustrating a robot movement method and system based on inertial sensors according to the present invention.

[0018] Figure 2 This is a schematic diagram illustrating the technical flow of the analytical model for the robot movement method and system based on inertial sensors according to the present invention. Detailed Implementation

[0019] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided.

[0020] Please see Figure 1 A robot movement method based on inertial sensors includes the following steps: S1: Obtain robot travel parameter data, construct a path trajectory network based on the robot travel parameter data, and detect the path guidance of the path trajectory network through a trajectory curve matching algorithm. Analyze the trajectory generation mode during robot travel to obtain autonomous travel trajectory data. S2: Introduce a magnetometer-assisted attitude fusion algorithm to construct a travel analysis model, and allocate a trajectory generation mode according to the autonomous travel trajectory data. Input the robot travel parameter data into the travel analysis model to analyze the inertial navigation information and output robot travel commands. S3: Execute the robot's movement instructions, verify the stability of the posture state during movement, adjust the movement path with posture deviation based on the autonomous movement trajectory data, optimize the robot's movement instructions through the path optimization protocol, and generate an optimized set of movement path instructions. S4: Construct an offset compensation mechanism, analyze the instruction optimization set, and when an offset in the travel trajectory is detected, select instruction optimization parameters according to the travel path, and re-input the instruction optimization parameters into the travel analysis model to re-match the newly selected travel trajectory path.

[0021] Specifically, the method for obtaining the robot's travel parameter data is as follows: the robot's travel signal and odometer displacement information are timestamped, and attitude jitter and inertial drift interference are filtered out based on a low-order filtering algorithm to generate robot travel parameter data for constructing a path trajectory network.

[0022] Specifically, the path trajectory network divides the pose calculation points of the robot's continuous movement process into nodes at discrete time intervals. Each effective pose node is used as a trajectory graph node. The graph structure connection edges are constructed based on the displacement vector change characteristics between nodes, and the guiding relationship between nodes is recorded to form a path trajectory network structure with path turning angle distribution and pose gradient information.

[0023] Specifically, the method for acquiring the autonomous travel trajectory data is as follows: The trajectory deviation index between the autonomous path and the reference path is calculated based on the trajectory curve matching algorithm. The real-time travel trajectory of the robot is inserted into the trajectory alignment matrix, and the dynamic deviation sequence between trajectory points is calculated by dynamic time warping. Based on the dynamic deviation sequence, a trajectory deviation index is obtained. The trajectory deviation index is analyzed. When the deviation exceeds a set threshold, the interval is marked as a non-guided mode segment, and trajectory segments that meet the guided mode constraints are extracted to obtain autonomous travel trajectory data.

[0024] Specifically, the method for constructing the movement parsing model is as follows: The robot's travel parameter data is synchronized using timestamps. An initial attitude calculation dataset is constructed based on the timestamps, and geomagnetic vector information is used to analyze the initial attitude calculation dataset to obtain the attitude direction reference. An attitude calculation framework is constructed based on the attitude orientation reference. A complementary filtering algorithm is used to quickly correct short-term attitude changes, and an extended Kalman filter is introduced to constrain the attitude stability of the magnetometer data. The attitude deflection error caused by gyroscope integral drift is corrected, and an intermediate attitude calculation state parameter set is generated. The intermediate state parameter set of the attitude calculation is analyzed, an inertial navigation attitude fusion matrix is ​​established, the fusion matrix is ​​mapped to the robot's travel parameter data, and the path trajectory is matched according to the trajectory estimation and deviation identification correlation function to construct a travel analysis model.

[0025] In this embodiment, the robot movement method and system based on inertial sensors adopts a layered technical architecture during implementation. The technology stack includes a data acquisition layer, a state estimation and modeling layer, a path planning and optimization layer, and a feedback control execution layer.

[0026] The specific technology stack is as follows: Data acquisition layer: Employs an inertial measurement unit (IMU) + visual sensor + edge computing module; Data transmission channel: ROS (Robot Operating System) message queue; State estimation and modeling layer: attitude and position fusion algorithm based on Kalman filter; Trajectory planning module: Global path planning is implemented using the A* algorithm and Dijkstra's algorithm; Motion optimization and control module: Python + TensorFlow for building dynamics and motion prediction models; Optimal travel strategy solution: Motion parameter optimization is performed using the SciPy optimization library; Environment and obstacle mapping: Spatial relationship modeling using Neo4j knowledge graph database; The specific implementation technology flow includes: data acquisition and preprocessing → state estimation and sensor fusion → path planning and travel optimization → simulation verification → closed-loop feedback and environmental adaptation.

[0027] By deploying inertial measurement units (IMUs), gyroscopes, accelerometers, and vision sensors on the robot's chassis, such as Figure 2The system collects real-time data on the robot's acceleration, angular velocity, position, and surrounding environment, transmitting this data to edge computing nodes via a ROS message queue to ensure data integrity and real-time performance. The input data undergoes denoising, feature extraction, and normalization. Kalman filtering is used to fuse IMU and visual odometry data, achieving accurate estimation of the robot's attitude, position, and velocity. Subsequently, based on the acquired state data, a robot movement data-driven model is constructed, using state features as input variables and combining path constraints, obstacle information, and dynamic constraints to form a preliminary movement plan. A path planning algorithm (such as A* or Dijkstra) is called to generate a global movement route, and motion parameters are optimized using the SciPy optimization library to maximize path smoothness and minimize energy consumption during robot movement. The movement process is simulated and verified using a robot simulation platform (such as Gazebo), outputting the optimized robot movement strategy. The deviation between the model output and actual sensor data is monitored in real-time. When position drift or dynamic changes in obstacles are detected, a closed-loop feedback mechanism is triggered. By using knowledge graph mapping methods, environmental changes and task constraints are transformed into optimization parameters, which are then re-input into the travel data-driven model to achieve dynamic updates and adaptive optimization of the robot path, forming a closed-loop technology flow of state estimation, path optimization, and policy feedback.

[0028] Specifically, the matching method for the trajectory generation pattern is as follows: Based on the trajectory dataset stored in the path trajectory network, the turning angle and speed change gradient of each trajectory node are analyzed to construct a trajectory feature parameter vector. The trajectory feature parameters are then combined with trajectory recognition labels to generate a trajectory pattern matching input set. The trajectory pattern matching input set is aligned with the feature vectors of the autonomous travel trajectory data. The feature similarity score is calculated by the trajectory curve matching algorithm, and the trajectory nodes are segmented and matched by Euclidean distance clustering operation to analyze the pattern type of the autonomous travel trajectory. Based on the trajectory pattern recognition labels, a mapping relationship is established between the trajectory generation pattern and the target trajectory segment in the path trajectory network. The identified pattern labels are bound to the corresponding trajectory segment index, and the trajectory segment index is fed back to the travel parsing model to match the trajectory generation pattern.

[0029] Specifically, the method for generating the instruction optimization set of the travel path is as follows: perform local path backtracking and reprojection calculation on the trajectory curve offset segment according to the path optimization protocol, remap the trajectory nodes to the center line of the fitted curve, and generate the travel path instruction optimization set based on the real-time attitude state analysis of the center line of the fitted curve.

[0030] Specifically, the path optimization protocol is based on a multi-constraint path adjustment strategy using a cost function. It incorporates trajectory deviation, cumulative inertial drift, attitude angle offset, and travel stability index as weighting factors of the cost function into the path optimization protocol. It also defines an instruction-level penalty function to analyze the stability of the path instruction set. When the stability of the path instruction set decreases, it triggers local path replanning logic.

[0031] Specifically, the output process of the robot's movement command is as follows: Based on the inertial navigation attitude fusion matrix output by the travel analysis model, the current attitude angle, heading angle and velocity vector data are analyzed, the attitude feature parameter set for path execution is extracted, and the attitude feature parameters are correlated and compared with the autonomous travel trajectory data to generate the trajectory parameters of the current target travel segment. Based on the trajectory parameters, the attitude angle data in the attitude feature parameter set of the path execution is solved by the path guidance instruction algorithm, and a set of travel attitude analysis control instructions is generated according to the solution results. The travel control pre-instruction set is output to the travel analysis model through a bus communication protocol, and real-time closed-loop correction is performed based on attitude feedback data during execution to output robot travel instructions that optimize the accuracy of travel trajectory matching.

[0032] In this embodiment, the robot's motion data acquisition module collects acceleration a(t) and angular velocity w(t) through an IMU sensor during its movement, and acquires environmental feature points P through a visual sensor. i The collected data undergoes noise filtering, specifically using the following formula: , Where, n a (t), n w (t) represents Gaussian white noise.

[0033] The analytical model construction module assigns a trajectory generation mode based on the autonomous travel trajectory data, inputs the robot's travel parameter data into the travel analytical model to analyze inertial navigation information, calculates a preliminary path based on the known obstacle map M and target point, and optimizes the generated speed v during travel. i The optimal combination of the steering angle and the turning angle minimizes the cost function: , Among them, E energy (v) i ( ) represents the energy consumption index, a and b are weighting coefficients, and E deviation (x) i ,p) represents the deviation from the reference path error.

[0034] The robot monitors the deviation between its current position and the reference path in real time. When the deviation exceeds a threshold, feedback control is triggered, and SciPy is used to solve for the optimal speed and turning angle. , , Meanwhile, by using knowledge graph mapping, the dynamic changes of environmental obstacles and task constraints are transformed into updated parameters, and the optimized path is recalculated to achieve dynamic adjustment.

[0035] The specific implementation steps are as follows: Figure 1 As shown, an IMU, gyroscope, accelerometer, and vision sensor are deployed on the robot, and the ROS message queue is initialized. Real-time robot state data is collected, and the collected data is filtered and standardized. Extended Kalman filtering is used to fuse sensor data to estimate the robot's pose. Based on the obstacle map and target position, the A* algorithm is executed to generate a global path. An optimization algorithm is used to solve for the optimal speed and steering angle at each path node. The robot moves according to the optimization strategy while monitoring the actual pose. When the deviation does not meet the settings, closed-loop feedback control is triggered to correct the motion parameters. Environmental changes and task constraints are mapped to a knowledge graph to dynamically update optimization parameters, achieving adaptive movement. After completing the task, the motion trajectory, energy consumption, and execution efficiency are recorded for subsequent model optimization.

[0036] Specifically, the method for constructing the offset compensation mechanism is as follows: The offset detection judgment function uses the trajectory curve matching residual and abnormal angular velocity peaks as offset triggering factors to generate an offset detection threshold range. When the distance offset between the real-time trajectory node and the trajectory centerline exceeds the set tolerance, the offset compensation mechanism is triggered to enter the compensation state. Under the compensated state, the sensor disturbance sequence corresponding to the offset interval is extracted, and combined with the heading angle correction vector corrected by the magnetometer, the observation noise covariance matrix and state transition gain coefficient of the internal filter of the analytical model are adaptively adjusted to construct the offset compensation mechanism.

[0037] Specifically, the magnetometer-assisted attitude fusion algorithm performs distortion calibration on the original three-axis magnetic field data output by the magnetometer by analyzing the ellipsoid fitting, constructs a debiased magnetic field vector, and maps it to the robot body coordinate system according to the geomagnetic reference direction to establish a reference coordinate framework for calculating the heading angle, thereby obtaining the magnetic distortion dynamic suppression coefficient table of the travel analytical model.

[0038] Specifically, a robot propulsion system based on inertial sensors is characterized by comprising: The navigation data acquisition module acquires robot navigation parameter data, constructs a path trajectory network based on the robot navigation parameter data, detects the path guidance of the path trajectory network through a trajectory curve matching algorithm, analyzes the trajectory generation mode during robot navigation, and obtains autonomous navigation trajectory data. Analytical model construction module: Introduces a magnetometer-assisted attitude fusion algorithm to construct a travel analytical model, and allocates a trajectory generation mode according to the autonomous travel trajectory data. Inputs the robot travel parameter data into the travel analytical model to analyze inertial navigation information and outputs robot travel commands. Path instruction optimization module: Executes the robot's movement instructions, verifies the stability of the posture state during movement, adjusts the movement path with posture deviation based on the autonomous movement trajectory data, optimizes the robot's movement instructions through the path optimization protocol, and generates an optimized set of movement path instructions; Closed-loop offset compensation module: Constructs an offset compensation mechanism, analyzes the instruction optimization set, and when an offset in the travel trajectory is detected, selects instruction optimization parameters according to the travel path and re-inputs the instruction optimization parameters into the travel analysis model to re-match the newly selected travel trajectory path.

[0039] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A robot movement method based on inertial sensors, characterized in that, include: S1: Obtain robot travel parameter data, construct a path trajectory network based on the robot travel parameter data, and detect the path guidance of the path trajectory network through a trajectory curve matching algorithm. Analyze the trajectory generation mode during robot travel to obtain autonomous travel trajectory data. S2: Introduce a magnetometer-assisted attitude fusion algorithm to construct a travel analysis model, and allocate a trajectory generation mode according to the autonomous travel trajectory data. Input the robot travel parameter data into the travel analysis model to analyze the inertial navigation information and output robot travel commands. S3: Execute the robot's movement instructions, verify the stability of the posture state during movement, adjust the movement path with posture deviation based on the autonomous movement trajectory data, optimize the robot's movement instructions through the path optimization protocol, and generate an optimized set of movement path instructions. S4: Construct an offset compensation mechanism, analyze the instruction optimization set, and when an offset in the travel trajectory is detected, select instruction optimization parameters according to the travel path, and re-input the instruction optimization parameters into the travel analysis model to re-match the newly selected travel trajectory path.

2. The method according to claim 1, characterized in that, The method for obtaining the robot's travel parameter data is as follows: the robot's travel signal and odometer displacement information are timestamped, and attitude jitter and inertial drift interference are filtered out based on a low-order filtering algorithm to generate robot travel parameter data for constructing a path trajectory network.

3. The method according to claim 1, characterized in that, The path trajectory network divides the pose calculation points of the robot's continuous movement into nodes at discrete time intervals. Each effective pose node is used as a trajectory graph node. Graph structure connection edges are constructed based on the displacement vector change characteristics between nodes, and the guiding relationship between nodes is recorded to form a path trajectory network structure with path turning angle distribution and pose gradient information.

4. The method according to claim 1, characterized in that, The method for obtaining the autonomous travel trajectory data is as follows: The trajectory deviation index between the autonomous path and the reference path is calculated based on the trajectory curve matching algorithm. The real-time travel trajectory of the robot is inserted into the trajectory alignment matrix, and the dynamic deviation sequence between trajectory points is calculated by dynamic time warping. Based on the dynamic deviation sequence, a trajectory deviation index is obtained. The trajectory deviation index is analyzed. When the deviation exceeds a set threshold, the interval is marked as a non-guided mode segment, and trajectory segments that meet the guided mode constraints are extracted to obtain autonomous travel trajectory data.

5. The method according to claim 1, characterized in that, The method for constructing the aforementioned moving parsing model is as follows: The robot's travel parameter data is synchronized using timestamps. An initial attitude calculation dataset is constructed based on the timestamps, and geomagnetic vector information is used to analyze the initial attitude calculation dataset to obtain the attitude direction reference. An attitude calculation framework is constructed based on the attitude orientation reference. A complementary filtering algorithm is used to quickly correct short-term attitude changes, and an extended Kalman filter is introduced to constrain the attitude stability of the magnetometer data. The attitude deflection error caused by gyroscope integral drift is corrected, and an intermediate attitude calculation state parameter set is generated. The intermediate state parameter set of the attitude calculation is analyzed, an inertial navigation attitude fusion matrix is ​​established, the fusion matrix is ​​mapped to the robot's travel parameter data, and the path trajectory is matched according to the trajectory estimation and deviation identification correlation function to construct a travel analysis model.

6. The method according to claim 1, characterized in that, The matching method for the trajectory generation pattern is as follows: Based on the trajectory dataset stored in the path trajectory network, the turning angle and speed change gradient of each trajectory node are analyzed to construct a trajectory feature parameter vector. The trajectory feature parameters are then combined with trajectory recognition labels to generate a trajectory pattern matching input set. The trajectory pattern matching input set is aligned with the feature vectors of the autonomous travel trajectory data. The feature similarity score is calculated by the trajectory curve matching algorithm, and the trajectory nodes are segmented and matched by Euclidean distance clustering operation to analyze the pattern type of the autonomous travel trajectory. Based on the trajectory pattern recognition labels, a mapping relationship is established between the trajectory generation pattern and the target trajectory segment in the path trajectory network. The identified pattern labels are bound to the corresponding trajectory segment index, and the trajectory segment index is fed back to the travel parsing model to match the trajectory generation pattern.

7. The method according to claim 1, characterized in that, The method for generating the instruction optimization set of the travel path is as follows: perform local path backtracking and reprojection calculation on the trajectory curve offset segment according to the path optimization protocol, remap the trajectory nodes to the center line of the fitted curve, and generate the travel path instruction optimization set based on the real-time attitude state analysis of the center line of the fitted curve.

8. The method according to claim 3, characterized in that, The path optimization protocol is based on a multi-constraint path adjustment strategy using a cost function. It incorporates trajectory deviation, cumulative inertial drift, attitude angle offset, and travel stability index as weighting factors of the cost function into the path optimization protocol. It also defines an instruction-level penalty function to analyze the stability of the path instruction set. When the stability of the path instruction set decreases, it triggers local path replanning logic.

9. The method according to claim 4, characterized in that, The process of outputting the robot's movement command is as follows: Based on the inertial navigation attitude fusion matrix output by the travel analysis model, the current attitude angle, heading angle and velocity vector data are analyzed, the attitude feature parameter set for path execution is extracted, and the attitude feature parameters are correlated and compared with the autonomous travel trajectory data to generate the trajectory parameters of the current target travel segment. Based on the trajectory parameters, the attitude angle data in the attitude feature parameter set of the path execution is solved by the path guidance instruction algorithm, and a set of travel attitude analysis control instructions is generated according to the solution results. The travel control pre-instruction set is output to the travel analysis model through a bus communication protocol, and real-time closed-loop correction is performed based on attitude feedback data during execution to output robot travel instructions that optimize the accuracy of travel trajectory matching.

10. The method according to claim 5, characterized in that, The method for constructing the offset compensation mechanism is as follows: The offset detection judgment function uses the trajectory curve matching residual and abnormal angular velocity peaks as offset triggering factors to generate an offset detection threshold range. When the distance offset between the real-time trajectory node and the trajectory centerline exceeds the set tolerance, the offset compensation mechanism is triggered to enter the compensation state. Under the compensated state, the sensor disturbance sequence corresponding to the offset interval is extracted, and combined with the heading angle correction vector corrected by the magnetometer, the observation noise covariance matrix and state transition gain coefficient of the internal filter of the analytical model are adaptively adjusted to construct the offset compensation mechanism.

11. The method according to claim 8, characterized in that, The magnetometer-assisted attitude fusion algorithm performs distortion calibration on the original three-axis magnetic field data output by the magnetometer by analyzing the ellipsoid fitting, constructs a debiased magnetic field vector, and maps it to the robot body coordinate system according to the geomagnetic reference direction to establish a reference coordinate framework for calculating the heading angle, thereby obtaining the magnetic distortion dynamic suppression coefficient table of the travel analytical model.

12. A robot propulsion system based on inertial sensors, characterized in that, include: The navigation data acquisition module acquires robot navigation parameter data, constructs a path trajectory network based on the robot navigation parameter data, detects the path guidance of the path trajectory network through a trajectory curve matching algorithm, analyzes the trajectory generation mode during robot navigation, and obtains autonomous navigation trajectory data. Analytical model construction module: Introduces a magnetometer-assisted attitude fusion algorithm to construct a travel analytical model, and allocates a trajectory generation mode according to the autonomous travel trajectory data. Inputs the robot travel parameter data into the travel analytical model to analyze inertial navigation information and outputs robot travel commands. Path instruction optimization module: Executes the robot's movement instructions, verifies the stability of the posture state during movement, adjusts the movement path with posture deviation based on the autonomous movement trajectory data, optimizes the robot's movement instructions through the path optimization protocol, and generates an optimized set of movement path instructions; Closed-loop offset compensation module: Constructs an offset compensation mechanism, analyzes the instruction optimization set, and when an offset in the travel trajectory is detected, selects instruction optimization parameters according to the travel path and re-inputs the instruction optimization parameters into the travel analysis model to re-match the newly selected travel trajectory path.

Images