A robot posture calibration method in a complex environment

Abstract

The application discloses a robot posture calibration method in complex environment, which is suitable for complex environments such as slope, vibration platform, soft ground, visual degradation area and multi-contact working condition. The method comprises the following steps: system initialization, multi-source posture information and environment information acquisition, dynamic trusted reference system construction, posture distortion event monitoring and triggering, posture inversion calibration based on environmental constraints, posture uncertainty and trusted interval calculation, and posture calibration result output and operation recovery. In the normal operation stage of the robot, only posture monitoring is performed, and the calibration process is triggered when the posture distortion event is detected. Through the above technical scheme, the application can effectively avoid the instability problems caused by the failure of the fixed reference system and the posture mutation, provide a safety boundary for the robot control, and significantly improve the posture reliability and operation safety of the robot in the complex environment.

Description

Technical Field

[0001] This invention relates to the field of robot posture calibration methods, specifically a robot posture calibration method for complex environments. Background Technology

[0002] With the widespread application of robotics in industrial automation, field exploration, medical assistance, agricultural inspection, and human-robot collaboration, accurate estimation and calibration of robot pose (including position and orientation) has become one of the core technologies for ensuring the safe and reliable operation of systems. Especially in complex, unstructured, or dynamic environments (such as slopes, soft ground, vibrating platforms, visually degraded scenarios, or multi-contact conditions), robots need to perceive their own pose in real time to support path planning, obstacle avoidance, force control, and task execution. However, existing pose calibration methods still face many challenges in these environments, leading to decreased accuracy, drift accumulation, or system failure, limiting the reliable deployment of robots in real-world scenarios.

[0003] Traditional attitude calibration methods primarily rely on fixed reference frames (such as the world coordinate system, initial attitude, or a single inertial coordinate system) or continuous filtering fusion (such as IMU-based integration, visual SLAM, or EKF / UKF extended Kalman filtering). These methods have achieved good results in ideal or structured environments (such as flat indoor floors, stable lighting, and no interference), but they have the following prominent limitations: Risk of Fixed Reference Frame Failure: Existing technologies generally presuppose a world coordinate system, inertial coordinate system, or initial attitude as an absolute or relative reference. However, in complex environments, robots may encounter slope inclination, soft ground subsidence, platform vibration, or external disturbances, causing the reference frame itself to drift or become distorted. For example, IMU integration can produce cumulative drift (pose errors caused by gyro bias and accelerometer bias increase exponentially over time), while vision systems are prone to losing feature points under changes in lighting, missing textures, or occlusion, further amplifying reference frame bias. Once the fixed reference fails, the entire pose estimation chain will experience systematic errors, potentially leading to robot localization failure or safety incidents.

[0004] Resource consumption and real-time issues of continuous calibration or filtering: Most methods employ continuous attitude estimation or continuous filtering (such as EKF fusing multi-sensor data), which still perform high-frequency calculations even when the attitude is stable, resulting in unnecessary waste of computational resources and increased power consumption. Furthermore, under sudden environmental changes (such as sudden vibrations or contact abrupt changes), it is difficult to quickly distinguish between "normal disturbances" and "true attitude distortion," potentially leading to delayed response or miscalibration, affecting system real-time performance and safety.

[0005] Lack of full utilization of environmental physical constraints: Traditional methods mainly rely on sensor data to directly calculate attitude angles, with little intrinsic incorporation of environmental geometric and physical constraints (such as ground normal, gravity direction, contact feasibility, or attitude continuity). When sensor noise, degradation, or missing information occurs, physically unreasonable attitude estimates are easily produced (such as the robot "floating" or "penetrating" the ground), resulting in a lack of interpretability and robustness of the results.

[0006] Uncertainty characterization and lack of safety boundaries: Existing calibration results typically only output point-estimated attitudes, lacking a quantitative characterization of uncertainty. In complex environments, attitude errors may fluctuate significantly due to multi-source inconsistencies, constraint strength, or convergence quality, but fail to provide clear safety boundaries for the upper-level control system. This leads to high-risk actions being performed under high uncertainty, increasing the risk of collisions or loss of control.

[0007] Insufficient environmental adaptability and robustness: In outdoor uneven terrain, vibration interference, visual degradation (such as dust, strong / low light, vegetation obstruction), or dynamic multi-contact scenarios, single sensors (such as IMU drift, visual feature loss) or simple fusion methods struggle to maintain stable accuracy. While existing research has proposed some compensation strategies (such as vibration classification or multimodal fusion), these are mostly specific to particular scenarios and lack universal dynamic reference construction and event-driven mechanisms.

[0008] In summary, existing robot posture calibration technologies face challenges in complex environments, including fixed reference failure, resource redundancy, physical inconsistencies, lack of uncertainty, and poor adaptability. These shortcomings severely restrict the reliable application of robots in unstructured, dynamic, or high-risk scenarios. Therefore, a novel posture calibration method is urgently needed that can dynamically adapt to environmental changes, perform efficient on-demand calibration, incorporate physical constraints, and provide safety boundaries to achieve stable, interpretable, and robust posture perception. Summary of the Invention

[0009] This invention aims to address the common technical problems encountered by existing robots when performing attitude calibration in complex environments, such as reliance on a fixed reference frame, continuous attitude calibration operation, insufficient utilization of environmental physical constraints, and lack of reliability assessment of attitude results. Existing technologies are prone to attitude reference failure, amplified attitude abrupt changes, or lack of physical interpretability in calibration results under complex conditions such as slopes, vibrating platforms, soft ground, or visual degradation, thereby affecting the overall stability and safety of the robot's operation.

[0010] To address this, the present invention proposes a posture calibration method and system for robots in complex environments. By introducing a dynamic reliable reference system construction mechanism, a posture distortion event triggering mechanism, and an posture inversion calibration mechanism based on environmental constraints, and outputting posture calibration results containing a reliable interval, the method fundamentally improves the posture reliability and operational safety of robots in complex environments.

[0011] To achieve the above objectives, the present invention adopts the following technical solution.

[0012] The present invention provides a posture calibration method for robots in complex environments, including system initialization, acquisition of multi-source posture information and environmental information, construction of dynamic reliable reference system, monitoring and triggering of posture distortion events, posture inversion calibration based on environmental constraints, calculation of posture uncertainty and reliable interval, and output of posture calibration results and operation recovery.

[0013] During normal robot operation, the system continuously monitors the posture state. When a posture distortion event is detected, the posture calibration process is triggered. During the calibration process, the dynamic reliable reference system is used as a benchmark, and environmental constraints are introduced to solve the posture inversely, thereby obtaining a posture calibration result that conforms to the physical feasibility of the environment. The result is output in the form of a reliable interval, providing a safety boundary for the robot control system.

[0014] In a preferred example, the system is further configured as follows: during operation, the system performs a credibility assessment on multi-source attitude information. The credibility assessment is based on at least one or more of the following: the stability of sensor data, the consistency between multi-source attitude information, the degree of matching with the geometric features of the current environment, and the physical rationality of the contact state with the robot. Based on the credibility assessment results, the system dynamically selects or weights and fuses attitude information sources with higher credibility to construct a dynamic credibility reference system at the current moment, which serves as the reference coordinate constraint for attitude inversion calibration.

[0015] Specifically, by constructing a dynamic and reliable reference system, the overall attitude calibration collapse caused by the failure of a fixed reference system in complex environments is avoided, enabling the attitude calibration benchmark to adaptively adjust with changes in the environment.

[0016] In a preferred embodiment, the present invention can be further configured as follows: the system continuously monitors the posture state during the normal operation of the robot to determine whether a posture distortion event has occurred. The posture distortion event includes at least one or more of the following: the deviation between multi-source posture information exceeds a preset threshold, the posture change rate increases abnormally, the posture derivation result is inconsistent with the robot's contact state or support state, and the overall posture reliability drops sharply due to a sudden change in environmental perception information. When any posture distortion event is detected, the posture calibration process is triggered.

[0017] Specifically, by using an attitude distortion event triggering mechanism, attitude calibration can be performed on demand, avoiding redundant calculations caused by continuous calibration and improving system real-time performance and operational stability.

[0018] In a preferred embodiment, the present invention can be further configured such that when an attitude distortion event is triggered, the system enters an attitude calibration mode and locks the current dynamic reliable reference frame. During the calibration process, the reference frame is frozen to prevent reference drift, thereby ensuring the consistency and stability of the attitude inversion solution process.

[0019] Specifically, by locking the reference frame, secondary errors caused by changes in the reference datum are avoided during attitude calibration, thereby improving the convergence and reliability of the attitude calibration results.

[0020] In a preferred embodiment, the present invention can be further configured as follows: In attitude calibration mode, the system constructs a constraint optimization model for attitude inversion based on a locked dynamic reliable reference system. The introduced environmental constraints include at least the geometric normal constraint of the ground or contact surface, the constraint of the relationship between gravity direction and attitude, the physical feasibility constraint of the robot contact area, and the constraint of the continuity of attitude change. The attitude calibration result is obtained by solving the nonlinear optimization problem.

[0021] Specifically, by introducing environmental physical and geometric constraints, the attitude calibration results not only meet the consistency of multi-source information, but also comply with environmental physical feasibility, significantly reducing the risk of attitude drift and abnormal attitude output.

[0022] In a preferred embodiment, the present invention can be further configured as follows: after obtaining the attitude calibration result, the system evaluates the uncertainty of the attitude calibration result based on the number and consistency of available attitude information sources, the strength of environmental constraints, and the convergence of inversion optimization, and generates a corresponding attitude confidence interval to characterize the safety boundary of the attitude calibration result.

[0023] Specifically, by outputting attitude results in the form of a confidence interval, the robot control system can adaptively adjust motion or operation strategies according to the attitude reliability, thereby improving operational safety in complex environments.

[0024] In a preferred embodiment, the present invention can be further configured as follows: when the attitude calibration result is stable and the attitude confidence interval is less than a preset threshold, the system automatically exits the attitude calibration mode and unlocks the reference frame, restoring to the attitude monitoring state; when the attitude calibration process times out or fails to converge, an abnormal alarm is triggered or a conservative control strategy is switched.

[0025] Specifically, by using calibration exit and recovery mechanisms, the system ensures that the attitude calibration process will not cause long-term interference with the normal operation of the robot, thereby improving the overall robustness of the system.

[0026] The beneficial effects achieved by this invention are as follows: 1. This invention introduces a dynamic trusted reference system construction mechanism, avoiding the failure risks associated with relying on a fixed world coordinate system or a single sensor reference system in existing technologies. In complex environments, the system can dynamically select a trusted attitude benchmark as the calibration reference based on the stability, consistency, and matching degree of multi-source attitude information with the geometric and physical characteristics of the environment. This ensures the reliability and adaptability of the attitude calibration benchmark even under conditions such as slopes, vibrations, soft ground, or visual degradation, significantly improving the stability and robustness of the attitude calibration results.

[0027] 2. This invention employs a posture distortion event-triggered calibration mechanism. When the robot's posture is stable, it only performs continuous monitoring without executing calibration calculations. The posture calibration process is triggered only when posture distortion events such as abnormal posture deviations, sudden increases in the rate of posture change, inconsistent contact states, or sudden changes in environmental information are detected. This approach avoids the redundant calculations and real-time pressures of traditional continuous calibration or continuous filtering, making the posture calibration process more efficient. Simultaneously, it can intervene promptly at critical distortion moments, effectively reducing the impact of posture changes on system operational safety.

[0028] 3. This invention, through an environmentally constrained attitude inversion calibration and reliable interval output mechanism, ensures that the attitude calibration results not only satisfy the consistency of multi-source information but also meet the requirements of environmental geometry and physical feasibility. Furthermore, it characterizes the uncertainty of the attitude results in the form of a reliable interval, providing a clear safety boundary for the robot control system. This enables the automatic limitation of high-risk actions when attitude uncertainty is high and the restoration of normal control when the attitude is reliable, significantly improving the overall reliability, safety, and interpretability of the robot in complex environments. Attached Figure Description

[0029] Figure 1 This is a schematic diagram of the overall structure of one embodiment of the present invention; Figure 2 This is a schematic diagram of the workflow structure of one embodiment of the present invention. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0031] It should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the invention.

[0032] The following is in conjunction with the appendix Figures 1-2 This invention describes a method for calibrating the attitude of a robot in complex environments, based on some embodiments of the present invention.

[0033] This invention is applicable to robot systems operating in complex environments, which include at least slopes, vibrating platforms, soft ground, narrow spaces, visually degraded areas, or multi-contact conditions. The robot may be a mobile robot, a wheeled or legged robot, an industrial robot, a mobile operation robot, or an unmanned vehicle.

[0034] In this embodiment, the robot posture calibration method in complex environments provided by the present invention adopts an event-driven hierarchical processing architecture, which includes a perception layer, an analysis layer, a calibration layer and an output layer.

[0035] The system consists of three layers: a perception layer for real-time acquisition of multi-source perception information related to robot posture and environmental perception information; an analysis layer for dynamically constructing a reliable reference system based on multi-source perception information and continuously monitoring the robot posture state to determine whether a posture distortion event has occurred; a calibration layer for inverting and calibrating the robot posture based on environmental constraints after a posture distortion event is triggered; and an output layer for generating posture output information containing posture calibration results and their reliability range.

[0036] It should be noted that the functions of each layer mentioned above can be implemented through a modular approach of hardware and software, and their functional logic does not depend on a specific hardware platform or software framework.

[0037] In this embodiment, the multi-source attitude-related information includes at least one or more of the following: The attitude or attitude change information obtained by the inertial measurement unit is used to reflect the robot's angular velocity and acceleration state in space; The pose estimation information obtained by the visual perception module is used to reflect the spatial pose relationship of the robot relative to environmental features. The pose information obtained by inverting the joint kinematics model is used to reflect the influence of the robot's structural motion on the overall pose; Attitude-related information derived from contact force or support state sensing units is used to reflect the physical contact relationship between the robot and the environment.

[0038] Meanwhile, environmental perception information may include the geometry of the environmental surface, normal direction, contact area distribution, environmental stability characteristics, etc., to provide environmental constraints for subsequent attitude inversion.

[0039] It should be noted that the present invention does not limit the specific method of obtaining the above information. Any perceptual information that can reflect the robot's posture state or the geometric and physical characteristics of the environment can be used as the input information of the present invention.

[0040] In this embodiment, the system does not preset a fixed world coordinate system or inertial coordinate system as the attitude calibration benchmark, but dynamically constructs a reliable reference system based on real-time sensing information.

[0041] Specifically, the system performs a credibility assessment on each posture information source, and the credibility assessment is based on at least one or a combination of the following factors: the stability characteristics of the posture data; the degree of consistency between the posture information sources; the degree of matching between the posture information and the geometric features of the environment; and the physical rationality between the posture information and the robot's contact state.

[0042] Based on the credibility assessment, the system selects attitude information sources with credibility higher than a preset threshold, and constructs a dynamic credible reference frame for the current moment based on the attitude information sources.

[0043] It should be noted that the dynamic reliable reference system in this invention is not the final attitude calibration result, but rather a reference benchmark in the subsequent attitude inversion calibration process. It is used to limit the reasonable solution space of attitude inversion, thereby avoiding the introduction of systematic errors by directly using the multi-source attitude fusion result as the control attitude.

[0044] Example 4: Monitoring and Triggering of Attitude Distortion Events In this embodiment, the system continuously monitors the robot's posture during normal operation, but does not perform posture calibration calculations.

[0045] The system determines whether a posture distortion event has occurred through at least one of the following methods: the posture deviation between different posture information sources exceeds a preset threshold; the posture change rate increases abnormally; there is an inconsistency between the posture result and the contact force or support state; or the environmental perception information undergoes a sudden change, resulting in a significant decrease in posture reliability.

[0046] When an attitude distortion event is detected, the system triggers the attitude calibration process and locks the current dynamic trusted reference frame as the calibration benchmark. After the attitude state stabilizes, the system automatically exits the attitude calibration process and returns to the attitude monitoring state.

[0047] Through the above methods, the present invention achieves discontinuous, on-demand execution of attitude calibration, effectively avoiding unnecessary continuous calibration calculations.

[0048] In this embodiment, attitude calibration adopts an inversion solution method based on environmental constraints.

[0049] During the calibration phase, the system constructs an attitude inversion model to ensure that the attitude solution simultaneously satisfies one or a combination of the following constraints: consistency constraints with the dynamic reliable reference system; spatial constraints with the geometric features of the environment surface; physical constraints with the direction of gravity or external force; feasibility constraints with the robot's contact state or support state; and attitude change continuity constraints.

[0050] By inverting the attitude under the above constraints, attitude calibration results that satisfy environmental physical feasibility and system consistency are obtained.

[0051] In other embodiments, the environmental constraints are not limited to ground constraints, but may also include geometric or physical constraints such as walls, slopes, clamping structures, step edges, or other accessible environmental structures.

[0052] In this embodiment, after completing the attitude inversion calibration, the system not only outputs the attitude calibration value, but also simultaneously calculates the uncertainty of the attitude result and generates the corresponding attitude confidence interval.

[0053] The attitude confidence interval is used to characterize the reliability of the current attitude calibration result and can be determined comprehensively based on the number of attitude information sources, consistency level, environmental constraint strength, and inversion convergence. During subsequent control, the robot can adaptively adjust its motion or operation strategy according to the size of the attitude confidence interval. When the attitude confidence interval is greater than a preset threshold, the execution of highly dynamic or high-risk actions is restricted, thereby improving the robot's operational safety in complex environments.

[0054] In summary, the robot attitude calibration method in complex environments provided by this invention includes at least the following steps: Collect multi-source attitude-related information and environmental perception information; dynamically evaluate the reliability of attitude information and construct a reliable reference system; continuously monitor the attitude state and determine whether an attitude distortion event has occurred; when an attitude distortion event is triggered, perform attitude inversion calibration based on environmental constraints; output attitude information including attitude calibration results and their reliability range.

[0055] Through the above technical solution, the present invention achieves stable, reliable and physically interpretable attitude calibration in complex environments.

[0056] The hardware focuses on multi-source sensing and environmental monitoring to ensure data reliability. The system can be integrated into the robot itself, eliminating the need for additional large equipment.

[0057] 1. Multi-source attitude information acquisition module: Inertial Measurement Unit (IMU): Employs a Bosch BMI088 or similar 6-axis IMU with a sampling rate >1kHz, providing angular velocity and acceleration data, and supporting attitude information output.

[0058] Visual perception module: RGB-D camera (such as Intel RealSense D435), resolution 640x480, depth accuracy <1cm, used to calculate pose (e.g., through SLAM algorithm).

[0059] Joint kinematics module: Robot joint encoder (such as absolute encoder, accuracy 0.01°), inverting attitude information.

[0060] Contact force / support status module: Force / torque sensor (such as ATI Nano17), mounted on the foot or end effector, measures contact force (range 0-50N).

[0061] Interface: Unified I2C / SPI connection to the main control board (such as NVIDIA Jetson Nano), 5V / 12V power supply, shielded cable to reduce noise.

[0062] 2. Environmental perception auxiliary hardware: Geometric feature sensor: LiDAR (such as Velodyne Puck Lite), with a scanning range of 10m, used to extract ground normals and constraints.

[0063] Interference monitoring sensors: humidity / temperature sensor (DHT22, accuracy ±2%RH), vibration sensor (IMU built-in accelerometer) and electromagnetic shielding layer (aluminum foil cage).

[0064] Installation considerations: Sensors are distributed and embedded into the robot joints / chassis to ensure redundancy (e.g., dual IMU backup).

[0065] Optimized for complex environments: The hardware features an IP67 waterproof / dustproof design and supports operating temperatures from -20°C to 60°C.

[0066] Working principle and usage process of this invention: In practical applications of this invention, the attitude calibration method is executed in the following order to form a closed-loop process of continuous monitoring and on-demand calibration: S1: System Initialization and Establishment of Initial Trusted Reference Frame After the robot system is powered on, the attitude calibration module first performs a self-test, initializing the multi-source attitude information acquisition units (including the inertial measurement unit, visual perception module, joint kinematics module, and contact force / support state module) and the environmental perception unit to ensure that all sensors are functioning properly. In the initial, relatively stable environmental phase, the system dynamically constructs an initial reliable reference frame based on the currently acquired multi-source attitude information and environmental geometric / physical feature information (such as ground normal and gravity direction) through a reliability assessment mechanism. This initial reference frame serves as the baseline coordinate constraint for subsequent attitude monitoring and calibration.

[0067] S2: Real-time acquisition of multi-source attitude and environmental information Once the robot enters normal operation, the system continuously and synchronously collects multi-source attitude-related information, including but not limited to: attitude information output by the inertial measurement unit; attitude information calculated by the visual perception module; attitude information obtained from joint kinematics inversion; attitude information derived from contact forces or support states; and environmental perception information (such as ground geometric normal, gravity direction, and physical characteristics of the contact area). The acquisition frequency is set according to the sensor characteristics (typically 1kHz for IMU, and 10-30Hz for vision / laser).

[0068] S3: Construction of Dynamic Trusted Reference Systems The system performs real-time reliability assessment on the collected multi-source attitude information. The assessment indicators include at least: stability of sensor data (variance or noise level); consistency between multi-source information; matching degree with the current environmental geometric features; and physical rationality of the contact state with the robot.

[0069] Based on the evaluation results, the system dynamically selects or weights and fuses attitude sources with high reliability to construct a dynamic reliable reference frame for the current moment. This reference frame is not directly used as the final attitude output, but rather as a stable reference coordinate constraint for subsequent attitude inversion calibration.

[0070] S4: Continuous monitoring of attitude distortion events During normal robot operation, the system monitors posture changes in real time to determine whether posture distortion events have occurred. Posture distortion events include, but are not limited to, one or more of the following: deviations in multi-source posture information exceeding a preset threshold (e.g., angle deviation > 5°); abnormally sudden increases in the rate of posture change (e.g., angular velocity > 10° / s); inconsistencies between posture derivation results and contact / support states; and sudden changes in environmental perception information leading to a sharp drop in overall posture reliability.

[0071] If no distortion event is detected, the system continues to maintain the monitoring state and does not trigger calibration to save computing resources.

[0072] S5: Attitude Distortion Event Triggering and Reference Frame Locking When any attitude distortion event is detected, the system immediately enters attitude calibration mode and locks the current dynamic reliable reference frame (i.e., freezes the reference frame constructed by S3) to prevent the reference benchmark from drifting during the calibration process, thereby ensuring the stability and consistency of the inversion process.

[0073] S6: Attitude Inversion Calibration Based on Environmental Constraints In calibration mode, the system constructs a constrained optimization problem based on a locked dynamic reliable reference frame, and solves it by inversion of the robot's current posture. The environmental constraints introduced in the inversion process include at least: geometric normal constraints of the ground or contact surface, constraints on the relationship between gravity direction and posture, constraints on the physical feasibility of the robot's contact area, and constraints on the continuity of posture changes. Simultaneously, the solved posture result must also satisfy the robot's own structural kinematic constraints to avoid conflicts between the posture solution and the robot's joint range of motion, mechanism assembly relationships, or structural kinematic model. A nonlinear optimization solver iteratively calculates the above constraints to obtain posture calibration results that simultaneously satisfy dynamic reference consistency, environmental physical feasibility, and robot structural kinematic constraints.

[0074] By iteratively calculating using a nonlinear optimization solver (such as Ceres Solver), attitude calibration results that simultaneously satisfy dynamic reference consistency, environmental physical feasibility, and robot structural kinematic constraints are obtained.

[0075] S7: Calculation of Attitude Uncertainty and Confidence Interval After obtaining the attitude calibration results, the system further calculates the uncertainty of the results. This uncertainty is determined based on a combination of factors such as the number and consistency of available attitude sources, the strength of environmental constraints, and the convergence of inversion optimization, forming a corresponding attitude confidence interval (e.g., expressed as ±2σ or 95% confidence level).

[0076] S8: Output of attitude calibration results The system outputs the final attitude calibration value and its confidence interval together for use by the robot's upper-level control system (such as motion planning, trajectory tracking, or obstacle avoidance modules). The output format can be ROS messages or custom structures, including attitude (Euler angles / quaternions), position (if applicable), and confidence interval boundaries.

[0077] S9: Calibration Mode Exit and Operation Resumption When the calibration results are stable (e.g., the confidence interval is less than a preset threshold, such as <0.5°) and no new distortion events occur, the system automatically exits the calibration mode, unlocks the reference frame, and returns to the S4 attitude monitoring state to continue evaluating the robot's attitude in real time. If the calibration times out or convergence fails, an abnormal alarm can be triggered or a conservative control strategy can be switched (e.g., reducing the movement speed).

[0078] Through the above working principle and usage process, this invention realizes an attitude calibration mechanism of "on-demand triggering, physical constraint inversion, and safety boundary output" under complex environmental conditions, which effectively avoids system instability caused by fixed reference system failure, sudden attitude change or sensor degradation, and significantly improves the overall reliability and safety of robot operation.

[0079] In the description of this specification, the terms "one embodiment," "some embodiments," "specific embodiment," etc., refer to a specific feature, structure, material, or characteristic described in connection with that embodiment or example, which is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0080] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A method for calibrating the attitude of a robot in complex environments, characterized in that, Includes the following steps: S1. System initialization and establishment of initial credible reference frame: After the robot system is powered on and started, the multi-source posture information acquisition unit and the environmental perception unit are initialized and tested. In the initial stage when the environment is relatively stable, an initial credible reference frame is constructed based on the acquired multi-source posture information and environmental geometric or physical feature information through credibility assessment. S2. Real-time acquisition of multi-source attitude information and environmental information: During the normal operation of the robot, multi-source attitude-related information and environmental perception information are continuously acquired; S3. Construction of dynamic trusted reference system: Real-time trustedness assessment of the collected multi-source attitude information, and dynamic selection or weighted fusion of attitude information sources with higher trustedness based on the assessment results to construct the dynamic trusted reference system at the current moment. S4. Continuous monitoring of posture distortion events: During the normal operation of the robot, the posture state is continuously monitored to determine whether a posture distortion event has occurred. S5, Attitude Distortion Event Trigger and Reference Frame Locking: When an attitude distortion event is detected, enter attitude calibration mode and lock the current dynamic trusted reference frame; S6. Attitude Inversion Calibration Based on Environmental Constraints: Using a locked dynamic reliable reference system as a benchmark, environmental constraints are introduced to invert and solve the robot's attitude to obtain attitude calibration results. S7. Attitude Uncertainty and Confidence Interval Calculation: Evaluate the uncertainty of the attitude calibration results and generate the corresponding attitude confidence interval; S8. Attitude calibration result output: Outputs attitude results including attitude calibration values ​​and their confidence intervals; S9. Calibration Mode Exit and Operation Resumption: When the attitude calibration result meets the stability requirements, exit the attitude calibration mode and resume the attitude monitoring state. The method only monitors the posture during the normal operation of the robot. When a posture distortion event is detected, the posture calibration process is triggered. During the calibration process, the posture inversion solution is performed based on a dynamic reliable reference system and environmental constraints are introduced. The posture calibration result containing the reliable interval is output.

2. The robot attitude calibration method in complex environments according to claim 1, characterized in that, The construction of the S3 dynamic trusted reference system includes: The credibility of multi-source pose information is evaluated, and the credibility evaluation is based on at least one or more of the following indicators: The stability of sensor data, the degree of consistency between multi-source attitude information, the degree of matching with the current environmental geometry, and the physical rationality of the robot's contact state; Based on the credibility assessment results, attitude information sources with higher credibility are dynamically selected or weighted and fused to construct a dynamic credible reference system for the current moment, which serves as the reference coordinate constraint for subsequent attitude inversion calibration.

3. The robot attitude calibration method in complex environments according to claim 1, characterized in that, In the continuous monitoring of S4 attitude distortion events, the attitude distortion event includes at least one of the following: the deviation between multi-source attitude information exceeds a preset threshold; An abnormally rapid increase in the rate of attitude change; inconsistency between the attitude derivation results and the robot's contact or support states; and a sudden change in environmental perception information leading to a sharp drop in the overall attitude reliability.

4. The robot attitude calibration method in complex environments according to claim 1, characterized in that, The S5 attitude distortion event triggering and reference frame locking includes: when any attitude distortion event is detected, immediately enter the attitude calibration mode and lock the current dynamic reliable reference frame to prevent the reference reference from drifting during the attitude calibration process.

5. The robot attitude calibration method in complex environments according to claim 1, characterized in that, The S6 attitude inversion calibration based on environmental constraints includes: constructing a constraint optimization model for attitude inversion based on a locked dynamic reliable reference system; the environmental constraints include at least: geometric normal constraints of the ground or contact surface, constraints on the relationship between gravity direction and attitude, constraints on the physical feasibility of the robot contact area, and constraints on the continuity of attitude changes; and obtaining attitude calibration results that simultaneously satisfy dynamic reference consistency, environmental physical feasibility, and robot structural kinematic constraints through nonlinear optimization solution.

6. The robot attitude calibration method in complex environments according to claim 1, characterized in that, The S7 attitude uncertainty and confidence interval calculation includes: calculating the uncertainty of the attitude calibration result based on the number and consistency of available attitude information sources, the strength of environmental constraints, and the convergence of the inversion optimization; and generating an attitude confidence interval based on the uncertainty to characterize the safety boundary of the attitude calibration result.

7. The robot attitude calibration method in complex environments according to claim 1, characterized in that, The S9 calibration mode exit and operation recovery include: when the attitude calibration result is stable and the attitude confidence interval is less than the preset threshold, automatically exiting the attitude calibration mode and unlocking the reference frame; when the attitude calibration process times out or fails to converge, triggering an abnormal alarm or switching to a conservative control strategy.

8. The robot attitude calibration method in complex environments according to any one of claims 1 to 7, characterized in that, The multi-source attitude information includes at least one or more of the following: attitude information output by the inertial measurement unit; attitude information calculated by the visual perception module; attitude information obtained by joint kinematics inversion; and attitude information derived from contact force or support state.

9. The robot attitude calibration method in complex environments according to any one of claims 1 to 7, characterized in that, The environmental perception information includes at least one or more of the following: ground geometric normal information, gravity direction information, and physical characteristic information of the contact area.

Images