A robot gait posture cooperative control method and device in a marine environment, a terminal device and a storage medium

Abstract

The application discloses a robot gait posture cooperative control method and device in a marine environment, terminal equipment and a storage medium, and belongs to the technical field of robot control. The method comprises the following steps: acquiring work platform sensing data, including robot operation data, and predicting a predicted disturbance phase sequence after extracting platform dynamic disturbance features; calculating contact reliability according to the work platform sensing data and the predicted disturbance phase sequence, combining the predicted disturbance phase sequence, current centroid state and current gait parameters in the robot operation data to calculate a stable support margin; matching a target gait from a preset gait set when the stable support margin is smaller than a preset safety threshold, obtaining an operation deviation amount of the target gait from a current gait and updating the current gait parameters, combining the predicted disturbance phase sequence to perform whole-body inertia reconstruction optimization to obtain a posture control amount, and combining the updated current gait parameters to control gait and posture switching. The application solves the problem of lacking gait and posture synchronous regulation and control.

Description

Technical Field

[0001] This invention relates to the field of robot control technology, and in particular to a method, device, terminal equipment and storage medium for coordinated control of robot gait and posture in a marine environment. Background Technology

[0002] With the increase in offshore operations, humanoid robots, with their bipedal mobility and adaptability to various work scenarios, are gradually being applied to marine operations such as offshore drilling platforms. Most existing robot gait control technologies are developed for flat, fixed land environments, and generally adopt a control framework that combines preset gait templates with real-time posture feedback, enabling basic walking and operational functions on stable ground.

[0003] However, in offshore operations, the operating platform is subjected to continuous wind, waves and ocean currents, which will generate roll, pitch and multi-degree-of-freedom complex disturbances. Existing control schemes can only switch gait or adjust posture based on a fixed platform. During the gait switching process, there is a lack of synchronous control of gait and posture, which cannot resist the balance impact caused by the dynamic disturbance of the platform. As a result, the robot's continuous operation stability in the offshore environment is insufficient and it is difficult to meet the needs of actual offshore operations. Summary of the Invention

[0004] This invention provides a method, device, terminal equipment, and storage medium for coordinated control of robot gait and posture in a marine environment. It can effectively solve the problem that existing technologies lack synchronous control of gait and posture during gait switching, cannot resist the balance impact caused by platform dynamic disturbances, and thus result in insufficient stability of robots in continuous operation in a marine environment.

[0005] One embodiment of the present invention provides a method for cooperative control of robot gait and posture in a marine environment, comprising: Acquire sensing data from the work platform; wherein, the sensing data from the work platform includes robot operation data; Based on the sensing data of the operating platform, the dynamic disturbance features of the platform are extracted, and the phase of the dynamic disturbance features of the platform is predicted to obtain the predicted disturbance phase sequence. The contact reliability is calculated based on the sensing data of the operating platform and the predicted disturbance phase sequence. Based on the contact confidence level, the predicted perturbation phase sequence, the current centroid state and current gait parameters in the robot's operating data, the stability support margin in the next gait cycle is calculated. When the stability support margin is less than the preset safety threshold, the target gait is matched from the preset gait set, the running deviation between the target gait and the current gait is obtained, and the current gait parameters are updated based on the target gait. Based on the updated current gait parameters, running deviation, and predicted perturbation phase sequence, the whole-body inertia is reconstructed and optimized to obtain the attitude control quantity. Based on the attitude control quantity and the updated current gait parameters, the robot under control is controlled to complete the smooth switching between gait and attitude.

[0006] Furthermore, the sensing data of the operating platform also includes: offshore operating area data, marine meteorological data, and foot contact status data; Based on the sensing data of the operating platform, dynamic disturbance features of the platform are extracted, and phase prediction is performed on the dynamic disturbance features of the platform to obtain a predicted disturbance phase sequence, including: Environmental features are extracted from the marine meteorological data, landing area features are extracted from the marine operation area data, attitude angle features are extracted from the robot operation data, and force contact features are extracted from the foot contact state data. Based on the environmental features, landing area features, attitude angle features, and force contact features, platform dynamic disturbance features are constructed. The platform's dynamic disturbance characteristics are subjected to sliding window time-domain filtering based on a preset filtering window, and the current disturbance phase is determined based on the filtered platform dynamic disturbance characteristics. The length of the filtering window is determined according to the next state period of the robot to be controlled. The disturbance phase includes the roll enhancement stage, roll fall stage, pitch enhancement stage, pitch fall stage, combined roll and pitch disturbance stage, and relatively stable stage. Phase prediction is performed based on the current disturbance phase and environmental characteristics to obtain the predicted disturbance phase sequence.

[0007] Furthermore, based on the sensing data of the operating platform and the predicted disturbance phase sequence, the contact confidence level is calculated, including: Based on the data of the marine operation area, the landable area of ​​the robot to be controlled is divided into grids to obtain several candidate landable areas. For each candidate footing area, the support area matching degree is calculated based on the effective contact area of ​​the foot in the foot contact state data and the grid opening area of ​​the candidate footing area. The integrity of the support edge is calculated based on the invalid side length of the candidate landing area and the outer contour side length of the candidate landing area. Based on the surface slip risk value in the data of the offshore operation area, the complementary value of surface slip risk is calculated; Based on the deck elastic deformation values ​​in the offshore operation area data, the structural elastic deformation risk complementarity value is calculated. Based on the foot landing posture parameters in the robot's operation data, the corresponding perturbation phase is matched from the predicted perturbation phase sequence, and the angle between the foot's force direction and the normal direction of the support area is determined based on the corresponding perturbation phase and the foot landing posture parameters. The stability support contribution value is determined from the preset stability support table based on the corresponding disturbance phase; The contact reliability is obtained by weighting the support area matching degree, support edge integrity, surface slip risk complementarity value, structural elastic deformation risk complementarity value, included angle, and stable support contribution value.

[0008] Furthermore, based on the contact confidence level, the predicted perturbation phase sequence, the current centroid state and current gait parameters in the robot's operational data, the stability support margin within the next gait cycle is calculated, including: Based on the current centroid state in the robot's operating data and the platform attitude change trend corresponding to the predicted perturbation phase sequence, the centroid safety margin of the robot's centroid projection relative to the support boundary is obtained; wherein, the support boundary is composed of the contours of the candidate landing area and the effective contact area of ​​the foot. Based on the reference angular momentum of the whole machine corresponding to the current gait parameters and the disturbance impact amplitude corresponding to the predicted disturbance phase sequence, the balance impact margin used to characterize the change in angular momentum is calculated. Based on the contact reliability of the candidate landing area and the preset contact reliability threshold, the support degradation margin used to characterize the change in foot contact reliability is calculated. Based on the current gait parameters and the attitude change amplitude corresponding to the predicted perturbation phase sequence, the attitude over-limit risk margin is calculated. The minimum value among the centroid safety margin, balance impact margin, support degradation margin, and attitude over-limit risk margin is taken as the stability support margin in the next state cycle.

[0009] Furthermore, matching the target gait from a preset gait set includes: Using the channel width and obstacle distribution information in the offshore operation area data as spatial constraints, each candidate gait in the preset gait set is screened, and candidate gaits whose lateral occupancy width exceeds the channel width and whose turning radius conflicts with the obstacle distribution information are eliminated, thus obtaining a feasible gait set. The perturbation phase matching degree of each feasible gait in the feasible gait set is calculated based on the predicted perturbation phase sequence, and the attitude recovery margin is calculated based on the perturbation phase matching degree and the attitude recovery capability parameters in the robot's running data. The minimum contact confidence level of the candidate landing area corresponding to each feasible gait in the feasible gait set is determined based on the contact confidence level. The feasible gait with a minimum contact confidence level greater than a preset minimum threshold and an attitude recovery margin greater than a preset recovery threshold is taken as the target gait. The target gait includes steady gait, narrowing conservative gait, widened support gait, side-stepping convergence gait, hurdle-crossing avoidance gait, and low center of gravity anti-shaking gait.

[0010] Furthermore, the operational deviations include: posture deviation, sudden change in foot contact, center of mass trajectory offset, and gait phase offset; Based on the updated current gait parameters, running deviation, and predicted perturbation phase sequence, whole-body inertia reconstruction and optimization are performed to obtain attitude control variables, including: The reference inertia distribution, reference center of mass position, and reference angular momentum are calculated based on the updated current gait parameters. Determine the direction of inertia reconstruction optimization during the current gait switching process based on the predicted perturbation phase sequence; Using attitude deviation, foot contact abrupt change, center of mass trajectory offset, and gait phase offset as real-time deviation constraints, a target optimization model is constructed based on the reference inertia distribution, reference center of mass position, and reference angular momentum. The target optimization model is solved based on the inertia reconstruction optimization direction to obtain the attitude control variables.

[0011] Furthermore, the current gait parameters are updated based on the target gait, including: Within a preset gait phase window, the current gait parameters and the parameters corresponding to the target gait are progressively updated by polynomial interpolation to obtain the updated current gait parameters. The current gait parameters include: stride length, stride width, stride frequency, support ratio, and leg swing height.

[0012] As an improvement to the above solution, another embodiment of the present invention provides a robot gait posture cooperative control device in a marine environment, comprising: The operation platform perception data acquisition module is used to acquire operation platform perception data; wherein, the operation platform perception data includes robot operation data; The disturbance phase sequence prediction module is used to extract dynamic disturbance features of the platform based on the sensing data of the operating platform, perform phase prediction on the dynamic disturbance features of the platform, and obtain a predicted disturbance phase sequence. The contact confidence calculation module is used to calculate the contact confidence based on the sensing data of the work platform and the predicted disturbance phase sequence. The stability support margin calculation module is used to calculate the stability support margin in the next gait cycle based on the contact confidence, the predicted disturbance phase sequence, the current centroid state and the current gait parameters in the robot's running data. The target gait matching module is used to match a target gait from a preset gait set when the stability support margin is less than a preset safety threshold, obtain the running deviation between the target gait and the current gait, and update the current gait parameters based on the target gait. The collaborative control module is used to reconstruct and optimize the whole-body inertia based on the updated current gait parameters, running deviation, and predicted disturbance phase sequence to obtain attitude control quantities. Based on the attitude control quantities and the updated current gait parameters, it controls the robot to be controlled to complete the smooth switching between gait and attitude.

[0013] Another embodiment of the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements a robot gait posture cooperative control method in a marine environment as described in the above embodiments.

[0014] Another embodiment of the present invention provides a computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform a robot gait posture cooperative control method in a marine environment as described in the above embodiment.

[0015] By implementing this invention, at least the following beneficial effects are achieved: This invention provides a method, device, terminal equipment, and storage medium for cooperative control of robot gait and posture in a marine environment. The method extracts dynamic disturbance characteristics of the operating platform from its sensing data and performs phase prediction to obtain a predicted disturbance phase sequence, enabling prediction of the disturbance evolution trend of the marine operating platform over a future period. After obtaining the contact confidence level based on the predicted disturbance phase sequence and the operating platform's sensing data, it calculates the stability support margin for the next gait cycle by combining the current centroid state and current gait parameters from the robot's operational data. This stability support margin can identify potential instability risks in advance and proactively trigger gait adjustments, overcoming the limitations of traditional feedback control. To address the lag issue, the system utilizes a whole-body inertia reconstruction optimization mechanism. By integrating updated gait parameters, operational deviations, and predicted disturbance phases to collaboratively solve attitude control parameters, the whole-body inertia reconstruction optimization reshapes the overall inertia distribution by adjusting the whole-body attitude. This proactively counteracts the imbalance impact caused by combined roll and pitch disturbances on the offshore platform, demonstrating stronger balance maintenance capabilities compared to independent gait and attitude control methods. Finally, based on the attitude control parameters and the updated current gait parameters, the system controls the robot to smoothly switch between gait and attitude, achieving integrated collaborative control of gait and attitude. This enhances balance under dynamic disturbances, improves the robot's continuous operational stability in the offshore environment, and meets the requirements of offshore operations. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating a robot gait and posture cooperative control method in a marine environment according to an embodiment of the present invention. Figure 2This is a schematic diagram of the structure of a robot gait posture cooperative control device in a marine environment according to an embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] See Figure 1 To address the problem in existing technologies where the lack of synchronized control over gait and posture during gait switching leads to insufficient stability of robots operating continuously in marine environments due to the inability to withstand balance shocks caused by platform dynamic disturbances, an embodiment of the present invention provides a flowchart illustrating a robot gait and posture cooperative control method in a marine environment, comprising: S1. Acquire sensing data from the work platform; wherein, the sensing data from the work platform includes robot operation data; Specifically, the perception data of the work platform represents a multi-dimensional raw dataset synchronously collected by all sensors on the offshore work platform carrying the robot to be controlled. This dataset is divided into three categories: offshore work area data, robot operation data, and foot contact status data, comprehensively representing the platform environment, robot motion, and all foot-ground interaction states. The offshore work area data represents quantitative information about the deck environment collected by LiDAR and visual depth cameras, including deck geometry, grid opening boundaries, channel width, obstacle coordinates, surface slipperiness, support edge integrity, and estimated parameters for local structural elastic deformation. This data is used to delineate the robot's foothold grid area and constrain its walking space. The robot operation data consists of real-time motion parameters collected by the robot's inertial measurement unit, whole-body joint encoders, and torque sensors. This includes body attitude angles, three-axis angular velocities or linear accelerations, joint positions, velocities, driving torques, current center of mass three-dimensional coordinates, a complete set of existing gait parameters (stride length, stride width, stride frequency, support ratio, leg swing height), and preset foot landing postures. This data is used to calculate the overall inertia, angular momentum, and center of mass safety margin. The plantar contact status data consists of foot-ground interaction parameters output by a six-dimensional force and torque sensor at the foot end, including vertical support force, tangential friction force, three-axis roll, pitch, yaw moment, plantar pressure distribution cloud map, effective contact area, pressure center coordinates, duration of single contact, foot slip velocity, and contact mutation amount, which are used to quantify the reliability of foot support and identify the risk of instantaneous slippage.

[0019] To illustrate, the body IMU, whole-body joint sensors, six-dimensional force sensor on the sole of the foot, and visual-laser fusion perception device are activated simultaneously, and a 100Hz synchronous sampling clock is set to eliminate time offset of multiple sensors. After the data acquisition is completed, all raw data are normalized to unify the dimensions, missing values ​​are filled, and abnormal mutation points are removed to form a time-aligned standardized dataset. The dataset is then split into data on the offshore operation area, robot operation data, and sole contact state data and stored in the controller cache.

[0020] In a preferred embodiment of the present invention, the work platform senses data. in, The number of state dimensions. The sampling time window length is used; the sensing data from the work platform is preprocessed using normalization, and the normalization formula is: in, This represents the i-th data point within the c-th time window. For the first The mean of the dimensional data. For the first The standard deviation of the dimensional data is calculated; the normalized sensing data of the work platform is spatiotemporally fused to provide a unified input for subsequent disturbance prediction and gait decision-making.

[0021] S2. Extract platform dynamic disturbance features based on the sensing data of the work platform, perform phase prediction on the platform dynamic disturbance features, and obtain a predicted disturbance phase sequence; Specifically, the platform dynamic disturbance characteristics represent feature vectors extracted from the fused and filtered data of the operating platform, characterizing the platform's swaying patterns. These include the platform's roll angle, pitch angle, three-axis angular velocity, vertical force fluctuation amplitude at the foot, and joint micro-vibration amplitude, distinguishing the characteristic differences between different swaying stages. The predicted disturbance phase sequence represents the chronologically ordered prediction results of the platform's disturbance state stages over a future period, with each phase corresponding to a specific motion state of the platform. The predicted disturbance phase sequence is arranged chronologically into six phase categories: increased roll, decreased roll, increased pitch, decreased pitch, combined roll and pitch disturbance, and relative stability, providing prior information on future disturbances for subsequent landing assessment, stability margin calculation, and attitude optimization. A disturbance phase represents a periodic or non-periodic stage of operational state change caused by external environmental factors affecting the offshore platform. For example, the impact of waves on the platform causes it to undulate up and down, and sway left and right. These movements have different characteristics and states at different times, and each specific state stage can be considered a disturbance phase.

[0022] Indicatively, effective feature components are extracted from the sensing data of the work platform and combined to construct the platform's dynamic disturbance features. The platform's dynamic disturbance features are then subjected to sliding window time-domain filtering, with the window length strictly matching the robot's single-step gait cycle. Based on the filtered roll and pitch angles and their angular velocity derivatives, six major disturbance phases are distinguished. Combining the disturbance amplitude and frequency variation patterns of the last three windows, the phase timing of the future gait cycle is extrapolated to generate a predicted disturbance phase sequence.

[0023] Preferably, the sensing data of the operating platform also includes: offshore operating area data, marine meteorological data, and foot contact status data; Based on the sensing data of the operating platform, dynamic disturbance features of the platform are extracted, and phase prediction is performed on the dynamic disturbance features of the platform to obtain a predicted disturbance phase sequence, including: Environmental features are extracted from the marine meteorological data, landing area features are extracted from the marine operation area data, attitude angle features are extracted from the robot operation data, and force contact features are extracted from the foot contact state data. Based on the environmental features, landing area features, attitude angle features, and force contact features, platform dynamic disturbance features are constructed. The platform's dynamic disturbance characteristics are subjected to sliding window time-domain filtering based on a preset filtering window, and the current disturbance phase is determined based on the filtered platform dynamic disturbance characteristics. The length of the filtering window is determined according to the next state period of the robot to be controlled. The disturbance phase includes the roll enhancement stage, roll fall stage, pitch enhancement stage, pitch fall stage, combined roll and pitch disturbance stage, and relatively stable stage. Phase prediction is performed based on the current disturbance phase and environmental characteristics to obtain the predicted disturbance phase sequence.

[0024] Specifically, marine meteorological data refers to the meteorological and marine environmental conditions of the offshore operation area, including wind speed, wind direction, wave height, wave frequency, and ocean current speed. These are external driving factors affecting the intensity and pattern of platform disturbances. Environmental characteristics are feature parameters directly related to platform disturbances extracted from marine meteorological data, such as wind speed amplitude, wave height amplitude, and wave frequency, which reflect the magnitude of the external driving force of the disturbance. Potential landing areas are areas pre-determined as possible landing points for the humanoid robot during its movement, based on its current position and surrounding environmental information. Landing area characteristics are features related to the platform structure's vibration characteristics extracted from the operation area data, such as deck local stiffness and grid span, reflecting the platform structure's response characteristics to disturbances. Attitude angle characteristics are platform attitude-related parameters extracted from robot operation data, including roll angle, pitch angle, yaw angle, and their angular velocity and angular acceleration, directly characterizing the platform's disturbance motion state. Force contact characteristics are the support force variation characteristics extracted from the foot contact state data, such as the fluctuation amplitude and rate of change of vertical force, which are the direct manifestation of platform disturbances transmitted to the robot's foot.

[0025] Schematic, firstly, environmental features are extracted from marine meteorological data, including parameters such as real-time wind speed, wave height, and wave dominant frequency. These are the external driving sources of platform disturbances and can reflect the overall intensity and trend of the disturbances. Secondly, landing area features are extracted from marine operation area data, including deck thickness, grid support span, and local structural stiffness, reflecting the structural vibration response characteristics of different areas. Thirdly, attitude angle features are extracted from robot operation data, including angular velocities, roll angles, and pitch angles in three axes, directly quantifying the platform's attitude motion. Fourthly, force contact features are extracted from foot contact state data, including the fluctuation amplitude and rate of change of the vertical support force on the foot, reflecting the force transmitted from the disturbance to the robot through the support surface. After aligning these four types of features synchronously over time, they are jointly constructed into a dynamic disturbance feature vector for the platform. Compared to using only the platform's attitude data, the fusion of meteorological and structural features provides a more comprehensive representation of disturbance characteristics and higher prediction accuracy.

[0026] Based on a preset filtering window, a sliding window time-domain filtering process is applied to the platform's dynamic disturbance feature vector. The filtering calculation is the arithmetic mean of the features of all sampling points within the window. The length of the filtering window is determined according to the robot's next gait cycle, typically taking the number of sampling points corresponding to 1 / 4 of a gait cycle. This value achieves an optimal balance between filtering anti-interference capability and phase recognition response speed. Based on the filtered disturbance characteristics, the disturbance phase of the current platform is determined according to the following rules: Roll enhancement phase: The absolute value of the roll angle continuously increases, and the derivative of the roll angular velocity is greater than 0, meaning the roll amplitude continuously increases and the swaying speed accelerates; Roll fall-off phase: The absolute value of the roll angle continuously decreases, and the derivative of the roll angular velocity is less than 0, meaning the roll amplitude continuously decreases and the swaying speed slows down; Pitch enhancement phase: The absolute value of the pitch angle continuously increases, and the derivative of the pitch angular velocity is greater than 0; Pitch fall-off phase: The absolute value of the pitch angle continuously decreases, and the derivative of the pitch angular velocity is less than 0; Combined roll and pitch disturbance phase: The absolute value of the roll angle is greater than the roll judgment threshold, the absolute value of the pitch angle is greater than the pitch judgment threshold, and the roll and pitch movements show synchronous changes in phase; Relatively stable phase: Both the roll and pitch angles are below their respective judgment thresholds, and the fluctuation amplitude is small.

[0027] To illustrate, the roll judgment threshold is determined based on the safe roll angle of the offshore operating platform, with a value range of 3°-8°; the pitch judgment threshold is determined based on the pitch motion characteristics of the ship, with a value range of 2°-6°. The thresholds can be adjusted according to the specific type and size of the operating platform.

[0028] Then, based on determining the current disturbance phase, and combining the disturbance amplitude, frequency, and phase change rate within the most recent continuous time window, as well as the trends in wind speed and wave height in meteorological characteristics, the platform disturbance trend within a preset time domain is extrapolated and predicted, resulting in a chronologically ordered sequence of predicted disturbance phases. The extrapolation prediction employs a combination of phase matching and trend extrapolation. Periodic, regular disturbances are extrapolated based on historical periodic patterns, while non-periodic random disturbances are extrapolated using meteorological change trends, balancing prediction accuracy and environmental adaptability.

[0029] Traditional platform disturbance identification methods mostly rely solely on data from the robot's own inertial sensors. This single data source is easily affected by the robot's own motion, resulting in low disturbance identification accuracy. Furthermore, the lack of consideration for meteorological driving factors makes it impossible to predict the overall change in disturbance intensity, leading to large short-term prediction errors. In addition, fixed filtering parameters cannot adapt to the response requirements under different time-state cycles, resulting in phase recognition lag or significant noise interference.

[0030] This embodiment utilizes a multi-source feature fusion and refined phase prediction process to construct disturbance features by integrating meteorological driving features, structural response features, body posture features, and foot force features. It comprehensively characterizes disturbance properties from three levels: disturbance source, structural response, and body perception, effectively reducing the interference of the robot's own motion on disturbance recognition and improving the accuracy of disturbance phase recognition. Simultaneously, the filter window length is dynamically adjusted with the gait cycle, ensuring filtering and noise reduction while matching the response speed requirements of the current gait, avoiding the problems of delayed response in fast gait and insufficient noise resistance in slow gait caused by a fixed window. Furthermore, the disturbance is divided into six specific phases, accurately distinguishing different states such as disturbance enhancement, decline, and compounding. Compared to the traditional method of only judging the magnitude of the disturbance, this provides a more accurate phase basis for subsequent gait adjustments, allowing for targeted adjustments before disturbance enhancement and restoration of normal gait during the decline phase. Moreover, the introduction of meteorological data as a prediction reference can predict the overall trend of disturbance intensity changes, improving prediction accuracy under irregular disturbances and reducing phase errors in the prediction time domain.

[0031] S3. Calculate the contact reliability based on the sensing data of the work platform and the predicted disturbance phase sequence; Specifically, contact reliability is a comprehensive quantitative indicator that the foot can form effective and stable support under current and future disturbances. The value range is [0,1]. The higher the value, the safer the landing, indicating that the landing area can better ensure that the robot remains stable after landing and is less likely to fall or lose balance.

[0032] Schematic, firstly, based on the geometric information of the work area, the accessible area in front of and around the robot is divided into grids to generate several candidate landing areas. For each candidate area, a quantitative evaluation is performed from six dimensions: support area matching, edge integrity, slippery risk, structural deformation risk, contact angle matching, and stability contribution under disturbance phase. Finally, the results of each dimension are weighted and fused according to preset weights to obtain the contact credibility of each candidate area. After the calculation is completed, the candidate areas are sorted from high to low contact credibility, and the next landing feasible region and the subsequent compensation landing feasible region are constructed: the former is the area that meets the safety threshold, and the latter is the adjacent reserve area for emergency landing adjustment in case of emergencies.

[0033] Preferably, the contact confidence level is calculated based on the sensing data of the operating platform and the predicted disturbance phase sequence, including: Based on the data of the marine operation area, the landable area of ​​the robot to be controlled is divided into grids to obtain several candidate landable areas. For each candidate footing area, the support area matching degree is calculated based on the effective contact area of ​​the foot in the foot contact state data and the grid opening area of ​​the candidate footing area. The integrity of the support edge is calculated based on the invalid side length of the candidate landing area and the outer contour side length of the candidate landing area. Based on the surface slip risk value in the data of the offshore operation area, the complementary value of surface slip risk is calculated; Based on the deck elastic deformation values ​​in the offshore operation area data, the structural elastic deformation risk complementarity value is calculated. Based on the foot landing posture parameters in the robot's operation data, the corresponding perturbation phase is matched from the predicted perturbation phase sequence, and the angle between the foot's force direction and the normal direction of the support area is determined based on the corresponding perturbation phase and the foot landing posture parameters. The stability support contribution value is determined from the preset stability support table based on the corresponding disturbance phase; The contact reliability is obtained by weighting the support area matching degree, support edge integrity, surface slip risk complementarity value, structural elastic deformation risk complementarity value, included angle, and stable support contribution value.

[0034] Specifically, gridding divides the continuous working plane into several square grid units of fixed size. Each grid unit is a candidate landing evaluation unit, enabling refined quantitative evaluation of the landing area. The candidate landing area is a grid unit obtained after gridding, located within the robot's reachable motion range, and is the basic unit for landing safety evaluation. Support area matching degree is the ratio of the actual contact area of ​​the foot to the effective bearing area within the candidate area, reflecting the degree of contact between the foot and the support surface after landing; a higher ratio indicates more sufficient contact. Support edge integrity is the proportion of continuous effective support edges in the outer contour of the candidate area, reflecting the condition of damage, openings, and gaps at the edges of the support area; a higher proportion indicates more reliable edge support. Surface slip risk complementarity value is the inverse quantitative indicator of surface slip risk, equal to 1 minus the slip risk value. The lower the slip risk, the higher the complementarity value, representing better friction conditions. Structural elastic deformation risk complementarity value is the inverse quantitative indicator of the risk of elastic deformation of the local deck structure under foot load, equal to 1 minus the deformation risk value. The lower the deformation risk, the higher the complementarity value, representing more rigid structural support. The stability support contribution value is the contribution coefficient of the candidate region to the overall stability support of the machine under a specific disturbance phase. Regions located in the favorable phase of the disturbance can provide better support and have a higher contribution value.

[0035] Schematic, based on the deck geometry, obstacle distribution, and robot reachable range of motion data in the offshore operation area, the reachable landing areas in front of and to the side of the robot are determined; the reachable area is uniformly divided into grids according to a preset grid size (usually 1 / 5 to 1 / 3 of the foot length), and each grid is a candidate landing area, realizing the discretization evaluation of continuous space.

[0036] For each candidate landing area, six evaluation indicators were calculated: The support area matching degree is the ratio of the effective contact area of ​​the foot to the area of ​​the continuous region that can be supported in the candidate grid, and the value is limited to between 0 and 1. If there are grid openings or gaps in the candidate area, the area of ​​the openings is deducted before calculating the effective support area; the closer the ratio is to 1, the more sufficient the contact between the foot and the support surface, and the better the support area condition.

[0037] The edge integrity score is calculated by summing the cumulative invalid edge lengths of gaps, cracks, grid opening boundaries, and damaged edges in the candidate region, then calculating the ratio of these invalid edge lengths to the total edge length of the candidate region's outer contour, and finally subtracting this ratio from 1 to obtain the edge integrity score. A value closer to 1 indicates a more continuous and complete edge, with no risk of edge misalignment.

[0038] The surface wet slip risk complementarity value is first determined by visual or laser reflection intensity to identify the proportion of water film or oil stains covering the candidate area. Combined with the ratio of historical tangential force to normal force on the sole of the foot and the micro-slip velocity of the foot, a 0-1 normalized wet slip risk value is calculated. The surface wet slip risk complementarity value is obtained by subtracting the wet slip risk value from 1. The higher the value, the better the surface friction conditions and the lower the risk of slipping.

[0039] The calculation of the structural elastic deformation risk complement value first uses the laser point cloud planar residual, the span of the grid or plate seam support identified by vision, and the historical trampling deformation, combined with the estimated vertical load on the sole of the foot, to predict the deflection deformation of the candidate area; the ratio of the predicted deflection value to the upper limit of the allowable deflection is calculated to obtain the 0-1 normalized elastic deformation risk value; the structural elastic deformation risk complement value is obtained by subtracting the deformation risk value from 1. The higher the value, the better the structural rigidity and the lower the risk of deformation and collapse.

[0040] The angle matching value is calculated based on the robot's foot landing posture parameters and the platform tilt angle corresponding to the predicted disturbance phase. The angle α between the foot's force direction and the normal direction of the support area is calculated. The cosine value of this angle, cosα, is taken as the angle matching value. The smaller the angle, the closer the cosine value is to 1, indicating that the force direction is more consistent with the normal direction of the support surface, and the better the vertical support effect.

[0041] In a preferred embodiment of the present invention, the expected foot contact time in the foot landing posture parameters is used as the matching benchmark. The time intervals of all phase units in the predicted perturbation phase sequence are traversed, and the phase unit whose expected foot contact time falls within its [start time, end time] interval is determined as the initial matching phase. If the expected foot contact time is exactly the switching time between two adjacent phase units, the phase unit with the later timing is selected as the initial matching phase. If the expected foot contact time is within a preset phase transition interval (the transition interval duration is set to 3%-5% of the single-step gait cycle), the perturbation amplitudes of two adjacent phase units are compared, and the phase unit with the larger perturbation amplitude is selected as the initial matching phase.

[0042] Extract the main disturbance direction of the initial matching phase and perform a direction consistency check in conjunction with the foot landing posture parameters. If the check passes, the phase is confirmed as the final matched disturbance phase. When the initial matching phase is the roll enhancement stage or the roll fall stage (the main disturbance direction is the roll direction), check whether the change trend of the foot roll angle is consistent with the roll disturbance direction of the phase. If the directions are consistent, the check passes. When the initial matching phase is the pitch enhancement stage or the pitch fall stage (the main disturbance direction is the pitch direction), check whether the change trend of the foot pitch angle is consistent with the pitch disturbance direction of the phase. If the directions are consistent, the check passes. When the initial matching phase is the combined roll and pitch disturbance stage (the main disturbance direction is both roll and pitch), simultaneously check the matching degree of the change trends of the foot roll angle and pitch angle with the corresponding disturbance directions. If the matching degree of both directions is not lower than the preset matching threshold (e.g., 75%), the check passes. When the initial matching phase is a relatively stable stage, no direction check is required, and the matching is directly confirmed.

[0043] The contribution value of stability support is determined by looking up the corresponding contribution coefficient in the preset stability support lookup table based on the predicted disturbance phase corresponding to the landing time. When the landing time is in a phase where the disturbance is receding and the platform is stabilizing, the contribution value is higher; when the landing time is in a phase where the disturbance is increasing and the platform is shaking violently, the contribution value is lower.

[0044] Then, according to preset weighting coefficients, the above six indicators are weighted and summed to obtain the contact credibility of the candidate areas, with the sum of all weighting coefficients being 1. The weighting coefficients are determined based on the risk priority of the marine environment: the support area matching degree has the highest weight, because insufficient area will directly lead to failure due to misalignment; the next is edge integrity, as edge damage can easily lead to sideslip and ankle sprains; wet slip risk and structural deformation risk have the next highest weights; angle matching and phase contribution have the lowest weights. After the calculation is completed, all candidate areas are sorted from high to low contact credibility. Areas with credibility higher than the preset contact credibility threshold are determined as the set of feasible landing areas for the next step; areas adjacent to feasible areas and meeting the preliminary threshold are determined as the set of subsequent compensation landing areas for emergency adjustments under sudden disturbances.

[0045] In a preferred embodiment of the present invention, a weighted summation method is used to calculate the contact reliability. The calculation formula is: Among them, the support area matching degree This indicates the effective contact area of ​​the sole of the foot. With grille opening area The degree of matching, with values ​​limited to 0 to 1; Support edge integrity , indicating invalid side length L edge The outer contour side length L of the candidate landing area total The degree of completeness, the closer C2 is to 1, the more continuous and complete the edge is.

[0046] Complementary values ​​for surface slip risk , representing the complementary value of the surface slip risk of the area; surface slip risk The system is determined by a combination of factors, including the proportion of water film or oil stains identified by visual or laser reflection intensity, the ratio of tangential to normal force on the sole of the foot, and the micro-slip velocity of the foot. When the proportion of water film or oil stains increases, the ratio of tangential to normal force approaches the friction threshold, or the micro-slip velocity of the foot exceeds a preset low threshold, the system considers the following conditions: Increase. Normalized to 0 to 1.

[0047] Complementary value of structural elastic deformation risk This represents the complementary value indicating the risk of localized elastic deformation of the deck structure; deck elastic deformation value. The deflection is determined by the ratio of the predicted deflection of the candidate landing area under the estimated foot load to the upper limit of the allowable deflection. The predicted deflection is estimated based on the planar residual of the laser point cloud, the span of the grid or plate seam support identified by visual recognition, historical trampling deformation, and the vertical load on the foot. A larger ratio indicates a higher risk of deformation.

[0048] Evaluation value of the angle between the direction of force application on the sole of the foot and the normal direction of the support area. , It is the angle between the direction of force application on the sole of the foot and the normal direction of the support area; Stable support contribution value , which represents the contribution coefficient of this region to the robot's stable support under the current perturbation phase; This represents the corresponding value in the preset stable support table; The weighting coefficients for each evaluation indicator satisfy the normalization constraint. The weighting coefficients are determined based on the priority of marine environmental risks: weighting for the matching degree of the support area. Pick Support edge integrity weights Pick Complementary values ​​for surface slipperiness risk Complementary value weights for structural elastic deformation risk Each take Angle weight Pick Stable support contribution value weight Pick .

[0049] Traditional landing area selection methods are mainly based on geometric distance and simple collision detection, with extremely limited evaluation dimensions: they only consider whether a landing is possible, without considering the reliability of the support after landing; they ignore the risks unique to offshore decks, such as slipperiness, grid openings, and structural deformation; and they do not consider the impact of platform disturbance phase on the support effect, resulting in a high failure rate of landing in complex offshore environments and risks such as slipping, stepping into empty spaces, deformation, and collapse.

[0050] This invention provides a comprehensive assessment across six dimensions, encompassing four levels: support geometry, surface characteristics, structural performance, and disturbance phase. This comprehensive approach covers all landing risks on offshore decks, reducing the risk of landing failure compared to traditional single-dimensional assessments. Weights are assigned according to the risk priority of the marine environment, prioritizing factors directly related to missteps, such as area matching and edge integrity. This aligns with the operational characteristics of offshore grid decks, resulting in assessments that better reflect actual working conditions. A dual-layer structure of a primary landing zone and a compensation reserve zone is constructed, not only planning the current landing point but also reserving emergency adjustment space. In the event of sudden disturbances, a rapid switch to a backup landing point is possible, significantly improving the robustness of landing planning. Landing safety is correlated with the disturbance phase, prioritizing landing during favorable disturbance phases to maximize the utilization of the platform's stability window and further enhance overall stability after landing.

[0051] S4. Calculate the stability support margin in the next gait cycle based on the contact confidence, predicted disturbance phase sequence, current centroid state and current gait parameters in the robot's running data. Specifically, the current center of mass state represents the three-dimensional projection coordinates of the robot's center of mass in the deck coordinate system, the center of mass movement speed, and the center of mass offset rate. These are core benchmark parameters for assessing the risk of overturning. The current gait parameters represent the complete set of control parameters for the robot's current gait, including stride length, stride width, stride frequency, support ratio, and leg swing height, determining the robot's basic walking trajectory. The stability support margin represents the remaining support capacity space that allows the humanoid robot to maintain stable movement in the current gait and posture within the next complete gait cycle. It comprehensively considers the impact of factors such as the robot's center of gravity position, foot support, and platform disturbances on the robot's stability. It takes the minimum value of four parameters: center of mass safety margin, balance impact margin, support degradation margin, and posture over-limit margin. If the value is lower than a preset safety threshold, an instability risk is identified. For example, if the robot's center of gravity is too close to the edge of the foot support and is subject to significant platform disturbances, its stability support margin will be small, making it prone to losing balance.

[0052] Indicatively, the impact of future disturbances, the reliability of support in the landing area, the motion law of the robot's center of mass, and the current gait characteristics are combined to calculate the corresponding safety margins from four dimensions: center of mass position safety, angular momentum impact, support reliability degradation, and attitude overshoot. Finally, the minimum value of the four margins is taken as the stability support margin for the future gait cycle, ensuring the conservatism and safety of the stability assessment with the most stringent dimensions.

[0053] Preferably, the stability support margin in the next gait cycle is calculated based on the contact confidence level, the predicted perturbation phase sequence, the current centroid state and current gait parameters in the robot's operating data, including: Based on the current centroid state in the robot's operating data and the platform attitude change trend corresponding to the predicted perturbation phase sequence, the centroid safety margin of the robot's centroid projection relative to the support boundary is obtained; wherein, the support boundary is composed of the contours of the candidate landing area and the effective contact area of ​​the foot. Based on the reference angular momentum of the whole machine corresponding to the current gait parameters and the disturbance impact amplitude corresponding to the predicted disturbance phase sequence, the balance impact margin used to characterize the change in angular momentum is calculated. Based on the contact reliability of the candidate landing area and the preset contact reliability threshold, the support degradation margin used to characterize the change in foot contact reliability is calculated. Based on the current gait parameters and the attitude change amplitude corresponding to the predicted perturbation phase sequence, the attitude over-limit risk margin is calculated. The minimum value among the centroid safety margin, balance impact margin, support degradation margin, and attitude over-limit risk margin is taken as the stability support margin in the next state cycle.

[0054] Specifically, the center-of-gravity safety margin is the ratio of the minimum distance from the robot's center-of-gravity projection point on the support surface to the boundary of the support polygon to a reference distance. This quantifies the safety redundancy of the center-of-gravity position; a higher ratio indicates a more inward-facing and safer center-of-gravity. The support boundary is the outer boundary of the support polygon formed by the contact contours of all current supporting feet of the robot, plus the contours of the planned candidate landing areas. It serves as the benchmark for determining whether the center-of-gravity exceeds the support range. The balance impact margin characterizes the safety redundancy of the impact of angular momentum changes on the overall balance. It is equal to 1 minus the ratio of the change in angular momentum to the upper limit of angular momentum allowance. A higher ratio indicates a smaller angular momentum impact and a more stable balance. The support degradation margin characterizes the safety redundancy of support capability degradation caused by decreased foot contact reliability. It is equal to the ratio of the lowest contact reliability of the candidate area to the contact reliability safety threshold. A higher ratio indicates more sufficient support reliability. The attitude over-limit risk margin characterizes the risk redundancy of attitude angles exceeding the safe range during gait switching. It is equal to 1 minus the ratio of attitude change to the upper limit of attitude allowance. A higher ratio indicates a safer attitude and a lower over-limit risk.

[0055] Schematic, based on the robot's current center of mass state and the predicted platform attitude change trend corresponding to the perturbation phase sequence, the projected position of the center of mass on the support surface after landing is deduced. Simultaneously, the contour of the current effective contact area of ​​the foot and the contour of the candidate landing area together form the overall support polygon, determining the support boundary. The minimum distance from the center of mass projection point to each edge of the support polygon is calculated, and then divided by the feature reference distance of the support polygon, such as half the support width, to obtain the center of mass safety margin. The support boundary is the global support contour at the whole-machine level, distinct from the local edge of a single candidate region: the edge of a single candidate region is only used to evaluate the contact reliability of that region itself, while the support boundary is used to evaluate the overturning risk of the entire machine; the two have different dimensions.

[0056] First, calculate the robot's baseline angular momentum based on the current gait parameters. Then, estimate the change in angular momentum during gait switching based on the predicted disturbance phase sequence and the corresponding disturbance impact amplitude. Calculate the ratio of the absolute value of the change in angular momentum to the allowable upper limit of angular momentum, and subtract this ratio from 1 to obtain the equilibrium impact margin. The larger the change in angular momentum and the smaller the margin, the stronger the equilibrium impact from the disturbance and the higher the risk of instability.

[0057] The lowest contact confidence level in the candidate landing area corresponding to the target gait is selected and divided by a preset contact confidence level threshold to obtain the support degradation margin. This index reflects the redundancy of support reliability in the worst landing area relative to the safety threshold; the lower the value, the worse the support conditions and the higher the risk of support degradation.

[0058] Based on the attitude difference between the current gait parameters and the target gait, and combined with the amplitude of the additional attitude change caused by the predicted perturbation phase, the maximum attitude change during gait switching is obtained. The ratio of the absolute value of the attitude change to the upper limit of the attitude safety limit is calculated, and this ratio is subtracted from 1 to obtain the attitude over-limit risk margin. The lower the value, the greater the attitude change and the higher the risk of exceeding the safety range.

[0059] The margin values ​​of four dimensions—center of mass safety margin, balance impact margin, support degradation margin, and attitude over-limit risk margin—are compared, and the minimum value is taken as the stability support margin in the next state cycle. Using the minimum value method allows the weakest dimension to be used as the basis for judging overall stability, ensuring the conservatism and safety of the assessment results. The stability support margin threshold is determined based on the lower limit of the robot's static stability margin, with a value range of 0.3-0.5, which can be adjusted according to the safety level of the task. The contact reliability threshold is determined based on the minimum safe support condition of the grid deck, with a value range of 0.4-0.6.

[0060] In a preferred embodiment of the present invention, stability support margin The calculation formula is: Among them, the center of mass safety margin , representing the robot's center of mass projection Relative to the support boundary Safety margin; Balanced impact margin , representing the change in angular momentum For the predicted disturbance phase sequence corresponding to the disturbance impact amplitude Impact margin caused; Support degradation margin This indicates the contact credibility of the candidate landing area. Contact confidence threshold Support degradation margin; attitude over-limit risk margin , representing the change in gait parameters With attitude change amplitude The risk margin of exceeding the limit; When stable support margin When the gait falls below a preset threshold, it is determined that the gait needs to be switched in advance and the posture coordination control should be triggered synchronously.

[0061] Traditional stability assessments are mostly based on calculating real-time stability margins at the current moment, which is a post-event assessment and cannot predict future stability risks. Moreover, the assessment dimensions are singular, mostly considering only the relationship between the center of mass position and the support polygon, ignoring key risk factors in marine scenarios such as angular momentum impact, support reliability degradation, and attitude over-limit, resulting in a large deviation between the assessment results and the actual instability risk, and failing to trigger protection actions in advance.

[0062] This embodiment, based on disturbance prediction and gait planning, pre-calculates the stability margin of future gait cycles, triggering adjustments before instability risks occur, providing sufficient response time for control, and fundamentally avoiding the lag of passive remediation. It covers four core instability triggers: center of mass position, angular momentum impact, support degradation, and attitude exceeding limits, comprehensively covering various instability paths in marine environments. The assessment results are more consistent with actual working conditions, reducing the false negative rate. A multi-dimensional minimum value fusion method is adopted, using the most stringent dimension as the overall margin to ensure the conservatism of stability judgment. If any dimension has a risk, it will be identified, minimizing the possibility of instability accidents. Both the margin threshold and the confidence threshold can be adjusted according to the operational scenario. High-risk operations can increase the threshold to enhance safety redundancy, while low-risk, stable scenarios can decrease the threshold to improve motion efficiency, making it more adaptable.

[0063] S5. When the stability support margin is less than the preset safety threshold, match the target gait from the preset gait set, obtain the running deviation between the target gait and the current gait, and update the current gait parameters based on the target gait. Specifically, the preset gait set represents multiple standardized gait templates adapted to offshore deck conditions, pre-stored offline in the controller. These include six types: steady-state gait, conservative gait with shrinking steps, widened support gait, side-stepping convergence gait, cross-grid clearance gait, and low-center-of-mass anti-sway gait, each adapted to different scenarios such as stable decks, narrow passages, grid openings, and large-scale swaying. The target gait represents the gait template selected from the gait set that best adapts to the current environment and disturbances, and exhibits optimal balance performance when stability margin is insufficient. The operational deviation is the difference between the theoretical motion state of the target gait and the robot's current actual operational state, categorized into four types: posture deviation, foot contact abrupt change, center-of-mass trajectory offset, and gait phase offset, serving as a closed-loop constraint for posture optimization.

[0064] To illustrate, the gait mode is proactively switched before instability risks are anticipated, avoiding the lag of passive adjustments after disturbances occur. When the future stability support margin is detected to be lower than the safety threshold, the optimal target gait is selected from a preset gait set, taking into account factors such as the future disturbance phase, site space constraints, robot posture recovery capability, and energy consumption constraints. Then, the deviations between the current gait and the target gait in posture, contact state, center of mass position, and gait phase are calculated. Finally, the current gait parameters are gradually updated to the target gait parameters through a smooth transition, avoiding posture oscillations caused by sudden parameter changes. The gait switching is triggered at the midpoint of the support phase, when the robot's support state is most stable, minimizing the impact of the switch. Simultaneously, a dedicated transition phase window is set during the switching process to ensure continuous parameter changes.

[0065] Preferably, matching a target gait from a preset gait set includes: Using the channel width and obstacle distribution information in the offshore operation area data as spatial constraints, each candidate gait in the preset gait set is screened, and candidate gaits whose lateral occupancy width exceeds the channel width and whose turning radius conflicts with the obstacle distribution information are eliminated, thus obtaining a feasible gait set. The perturbation phase matching degree of each feasible gait in the feasible gait set is calculated based on the predicted perturbation phase sequence, and the attitude recovery margin is calculated based on the perturbation phase matching degree and the attitude recovery capability parameters in the robot's running data. The minimum contact confidence level of the candidate landing area corresponding to each feasible gait in the feasible gait set is determined based on the contact confidence level. The feasible gait with a minimum contact confidence level greater than a preset minimum threshold and an attitude recovery margin greater than a preset recovery threshold is taken as the target gait. The target gait includes steady gait, narrowing conservative gait, widened support gait, side-stepping convergence gait, hurdle-crossing avoidance gait, and low center of gravity anti-shaking gait.

[0066] Specifically, spatial constraints are gait limitations imposed by the geometry of the work area, including lateral dimension constraints determined by the width of the passageway, leg swing space constraints determined by the distribution of obstacles, and motion space constraints determined by the turning radius. Disturbance phase matching degree is an indicator of how well a candidate gait matches the phase of a future disturbance; a higher matching degree indicates better stability of the gait under future disturbances. Posture recovery margin is an indicator of the robot's ability to recover from a disturbed posture to a stable posture after adopting a corresponding gait; a higher margin indicates stronger recovery capability and more robust disturbance resistance. Steady-state gait is the conventional walking gait used by the robot in a stable environment, with stride length, stride frequency, and center of mass height all at conventional levels, balancing walking efficiency and stability. Conservative gait with reduced stride length and stride frequency reduces foot impact and instability risk by minimizing single-step displacement, suitable for moderate-intensity disturbances. Widened support gait increases stride width and widens the support polygon, improving anti-roll capability by increasing the width of the support base, suitable for scenarios with strong roll disturbances. The lateral tactile gait is a walking gait where the robot walks sideways with its limbs tactile, reducing its lateral wind-exposed area and adapting to narrow passages, making it suitable for scenarios with strong winds and narrow passages. The grid-crossing clearance gait is a gait where the stride length precisely matches the grid spacing, and the feet land precisely on the grid crossbeams, specifically adapted for grid deck scenarios to maximize the support area. The low-center-of-gravity anti-sway gait lowers the overall center of gravity height, with the knees bent and chest slightly hunched, improving anti-tipping ability and suitable for scenarios with strong disturbances.

[0067] Schematic, using the channel width and obstacle distribution information in the offshore operation area data as hard spatial constraints, all candidate gaits in the preset gait set are initially screened: if the lateral width of the gait is greater than the effective channel width, it means that normal walking in the channel is impossible, and it is directly eliminated; if the turning radius of the gait conflicts with the obstacle distribution, or the leg swing trajectory will encroach on the obstacle space, it means that there is a collision risk, and it is directly eliminated. After the initial screening, the remaining gaits constitute the feasible gait set, ensuring that all candidate gaits have the spatial conditions for execution.

[0068] For each gait in the feasible gait set, the following calculations are performed: Perturbation Phase Matching Degree: The gait's robustness is evaluated by considering the type, amplitude, and trend of the predicted perturbation phase sequence. For example, under predominantly yaw perturbations, the widened support gait has a higher matching degree; under complex strong perturbations, the low center-of-mass anti-yaw gait has a higher matching degree. Attitude Recovery Margin: The attitude recovery margin under maximum perturbation is calculated by combining the robot's own attitude recovery capability parameters. A higher margin indicates a more sufficient disturbance resistance reserve for the gait. Minimum Contact Confidence: The lowest contact confidence is extracted from the candidate landing region by matching the landing sequence corresponding to the gait, representing the worst support condition for that gait.

[0069] The gait is selected based on safety priority. Level 1: Gaits with a minimum contact confidence level greater than a preset minimum threshold and a posture recovery margin greater than a preset recovery threshold are selected to ensure basic safety requirements are met. Level 2: Among gaits meeting the safety threshold, the gait with the highest disturbance phase matching degree and the largest future stability support margin is selected to ensure optimal disturbance rejection stability. Level 3: If multiple gaits have similar stability indicators, the gait with the lowest energy consumption per unit displacement and the lowest gait switching cost is selected as the final target gait, balancing operational efficiency and energy consumption. If all candidate gaits fail to meet the safety threshold requirements, an emergency steady-state maintenance mode is triggered: the robot stops moving forward, adjusts to a stable posture with the lowest center of mass, and replans subsequent compensation landing areas, resuming walking only after the disturbance subsides.

[0070] Traditional gait switching methods often follow fixed rules, such as slowing down when there is a large disturbance, and lack a systematic gait selection mechanism. When switching, only the intensity of the disturbance is considered, ignoring factors such as the constraints of the field space and the support conditions of the foot, which can easily lead to poor gait adaptability. Moreover, the switching process lacks a smooth transition, and sudden changes in gait parameters can easily cause posture oscillations, which can lead to instability.

[0071] This embodiment employs a hierarchical gait matching mechanism, achieving precise adaptation and smooth switching of gaits. Spatial constraints serve as the first screening threshold, eliminating unexecutable gaits to avoid subsequent invalid calculations and ensuring that gaits do not encounter execution-level issues such as collisions or excessive width. Following a priority order of safety-stability-energy consumption, the optimal gait is selected layer by layer, guaranteeing both the safety baseline for maritime operations and balancing stability and operational efficiency, thus improving adaptability. Six preset gaits cover various maritime scenarios, from smooth to highly disturbed, from wide passages to narrow spaces, and from flat ground to lattice decks, capable of handling complex and ever-changing marine environments. An emergency steady-state maintenance mode is included as a fallback; when all gaits fail to meet safety requirements, priority is given to preventing the robot from tipping over, demonstrating strong robustness and fault tolerance.

[0072] Preferably, updating the current gait parameters based on the target gait includes: Within a preset gait phase window, the current gait parameters and the parameters corresponding to the target gait are progressively updated by polynomial interpolation to obtain the updated current gait parameters. The current gait parameters include: stride length, stride width, stride frequency, support ratio, and leg swing height.

[0073] Specifically, the gait phase window is a pre-defined gait phase interval used to complete the transition of gait parameters. Within this interval, the parameter switching is completed gradually, and the length of the window determines the speed and smoothness of the switching. Gait parameters are the core parameters describing gait characteristics, including stride length (distance traveled in a single step), stride width (lateral distance between the two feet), cadence (number of steps per unit time), support phase percentage (the proportion of time spent in the support phase of a single step), and leg swing height (the height the foot lifts when stepping).

[0074] Schematic illustration: A transition phase window for gait switching is pre-set, typically with a length of half a gait cycle, corresponding to the interval from the midpoint of the support phase to the midpoint of the next support phase. This interval is chosen because the robot's bipedal support is optimal during the midpoint of the support phase, minimizing the impact on overall stability during parameter switching. The start and end phases of the window are dynamically adjusted based on the gait type to ensure the switching process remains within a phase of high support stability. Within the pre-set gait phase window, polynomial interpolation is performed on the current gait parameters and the parameters corresponding to the target gait, achieving progressive parameter updates. Cubic polynomial interpolation is typically used to ensure that parameter values ​​and rates of change are continuous at the start and end points of the window, without abrupt changes. Taking step length as an example, if the step length at the start phase of the window is the current gait step length, and the step length at the end phase of the window is the target gait step length, with the window normalized phase ranging from 0 to 1, then the step length of any phase during the transition is a cubic polynomial function, ensuring parameter continuity and a rate of change of 0 at the start and end points, achieving a smooth transition. Stride width, cadence, support ratio, leg swing height, and other gait parameters are all updated synchronously using the same interpolation method. During the interpolation process, the stability support margin is checked in real time. If the margin falls below the safety threshold during the transition, the interpolation speed will be automatically slowed down and the transition window will be extended to prioritize the stability of the switching process.

[0075] This embodiment ensures continuous gait changes by synchronously interpolating all gait parameters, such as stride length, stride width, and stride frequency, preventing inconsistencies and ensuring a smooth transition. Stability is checked in real-time during the transition, and the switching speed is automatically slowed down when the margin is insufficient, achieving adaptive switching that prioritizes stability while balancing efficiency and safety.

[0076] S6. Based on the updated current gait parameters, running deviation, and predicted disturbance phase sequence, perform whole-body inertia reconstruction optimization to obtain attitude control quantities. Based on the attitude control quantities and the updated current gait parameters, control the robot to be controlled to complete the smooth switching between gait and attitude.

[0077] Specifically, whole-body inertia reconstruction optimization refers to the coordinated adjustment of the posture of all joints in the robot's upper limbs, torso, pelvis, and lower limbs, actively changing the overall mass distribution and angular momentum transfer path to counteract the unbalanced impact caused by platform swaying. For example, by adjusting the swing amplitude and frequency of the arms and the position of the body's center of gravity, the robot's inertial parameters are optimized, improving its resistance to environmental disturbances. Attitude control quantities represent the target angles and torque commands for each joint output after inertia reconstruction optimization, used to synchronously correct the robot's torso and limb postures. Smooth gait and posture switching does not directly replace gait parameters with a step, but rather updates gait commands through progressive interpolation and synchronously outputs continuous attitude control quantities, eliminating sudden changes in the center of gravity and foot force during the switching process. In the dynamic environment at sea, the robot's gait and posture need to be adjusted collaboratively to adapt to platform disturbances and environmental changes, ensuring the stability and effectiveness of its movement. For example, when the platform tilts upwards, the robot needs to adjust its gait and posture to appropriately lower its center of gravity and move backwards to maintain balance.

[0078] Indicatively, while gait switching occurs, the robot's overall posture is simultaneously optimized, and its anti-disturbance capability is improved through inertia characteristic adjustments. Using the ideal inertia, center of mass, and angular momentum corresponding to the target gait as benchmarks, and combining this with the predicted disturbance direction, the inertia reconstruction optimization direction is determined. Simultaneously, the real-time collected operational deviation is used as a constraint to construct the inertia reconstruction optimization objective function. Solving the optimization function yields the posture adjustment quantities of the upper limbs, trunk, pelvis, and knee / ankle joints, i.e., the posture control quantities. Finally, the gait parameter commands and posture control quantities are synchronously sent to the joint controller, driving the robot to complete a smooth and coordinated gait and posture switching. The synchronous initiation and completion of inertia reconstruction and gait switching ensure that the overall anti-tipping capability remains optimal during gait changes, preventing the gait switching process from becoming a weak point in stability.

[0079] Preferably, the operational deviation includes: posture deviation, sudden change in foot contact, center of mass trajectory offset, and gait phase offset; Based on the updated current gait parameters, running deviation, and predicted perturbation phase sequence, whole-body inertia reconstruction and optimization are performed to obtain attitude control variables, including: The reference inertia distribution, reference center of mass position, and reference angular momentum are calculated based on the updated current gait parameters. Determine the direction of inertia reconstruction optimization during the current gait switching process based on the predicted perturbation phase sequence; Using attitude deviation, foot contact abrupt change, center of mass trajectory offset, and gait phase offset as real-time deviation constraints, a target optimization model is constructed based on the reference inertia distribution, reference center of mass position, and reference angular momentum. The target optimization model is solved based on the inertia reconstruction optimization direction to obtain the attitude control variables.

[0080] Specifically, the reference inertia distribution is the ideal rotational inertia distribution of the robot under the target gait, serving as the target reference for inertia reconstruction. The reference center of mass position is the ideal spatial position of the robot's center of mass under the target gait, serving as the target reference for center of mass adjustment. The reference angular momentum is the ideal angular momentum level of the robot under the target gait, serving as the reference for angular momentum suppression. The optimization direction for inertia reconstruction is determined based on the direction of the disturbance; for example, under yaw disturbances, it is necessary to increase the lateral rotational inertia and lower the center of mass, while under pitch disturbances, it is necessary to increase the longitudinal rotational inertia, etc. The target optimization model represents a mathematical optimization model constructed with the optimization objectives of inertia, angular momentum, center of mass, and support force, used to solve for the optimal attitude adjustment amount.

[0081] Schematic, based on the updated current gait parameters, the reference inertia distribution, reference center of mass position, and reference angular momentum under this gait are calculated through the robot's rigid body dynamics model, serving as the ideal target state for inertia reconstruction. The reference parameters correspond to the optimal dynamic characteristics during stable gait operation and are the ultimate goal of attitude adjustment.

[0082] Based on the predicted disturbance phase sequence, the type, direction, and intensity of the current disturbance are determined, and the optimization direction for inertia reconstruction is identified: if the disturbance is predominantly roll, the optimization direction is to increase the overall lateral rotational inertia, reduce the center of mass height, and suppress roll angular momentum; if the disturbance is predominantly pitch, the optimization direction is to increase the overall longitudinal rotational inertia, adjust the center of mass position forward and backward, and suppress pitch angular momentum; if the disturbance is a combination of lateral and longitudinal, the optimization direction is to simultaneously improve the anti-disturbance capability in both directions, prioritizing the direction with the stronger disturbance. The optimization direction determines the priority and emphasis of attitude adjustment for each component.

[0083] A multi-objective optimization model was constructed using the changes in overall inertia, angular momentum, center of mass offset, and plantar support force as optimization objectives, while real-time collected attitude deviations, plantar contact abrupt changes, center of mass trajectory offsets, and gait phase offsets as constraints. The optimization objectives were weighted according to their anti-overturning priority: center of mass offset had the highest weight because its position directly determines overturning risk; followed by angular momentum change, as angular momentum impact is a core cause of dynamic instability; then, overall inertia change, as inertia distribution affects disturbance rejection response; and finally, plantar support force change, ensuring smooth foot contact. Center of mass adjustment accounted for 0.4, angular momentum adjustment for 0.3, inertia adjustment for 0.2, and support force adjustment for 0.1, ensuring that the optimization results prioritize the most critical overturning safety.

[0084] Finally, based on the determined inertia reconstruction optimization direction, the target optimization model is solved to obtain the attitude adjustment quantities of each joint in the body, i.e., attitude control quantities. Specific adjustment areas include: upper limb positioning adjustment, trunk pitch and yaw adjustment, pelvic translation and yaw adjustment, knee joint compliant compression and extension adjustment, and ankle joint fine-grained attitude adjustment. Through multi-part coordinated adjustment, the mass distribution and angular momentum transfer path of the entire aircraft are reshaped, forming a stable attitude adapted to the current disturbance, resistant to roll, pitch, and overturning. The solution process employs a model predictive control framework; in each control cycle, the constraints are updated and resolved based on the latest deviation, achieving dynamic closed-loop adjustment of inertia reconstruction.

[0085] In traditional gait switching, attitude control only focuses on joint trajectory tracking and lacks global control of the machine's inertia and angular momentum. Gait switching and attitude adjustment are independent of each other and are not executed synchronously, making the switching process a weak link in stability. Moreover, attitude adjustment lacks a clear direction for disturbance rejection, has poor targeting, and has limited anti-roll and anti-pitch effects, making it easy to become unstable during the switching process.

[0086] This embodiment optimizes core dynamic parameters such as inertia, angular momentum, and center of mass, rather than simply adjusting joint positions, fundamentally improving the overall machine's disturbance resistance and anti-tipping performance. The optimization direction is determined based on the disturbance direction, specifically enhancing disturbance resistance in that direction. Compared to general attitude adjustment, directional disturbance resistance efficiency is more than doubled. Optimization weights are set according to anti-tipping priority, prioritizing center of mass stabilization before suppressing angular momentum impact, meeting safety requirements in maritime scenarios and ensuring that preventing capsizing is prioritized in extreme situations. Coordinated adjustment of the upper limbs, trunk, pelvis, knees, ankles, and other body parts comprehensively reshapes mass distribution and angular momentum transfer paths, offering a wider adjustment range and better inertia reconstruction effect compared to adjusting only the lower limbs.

[0087] In a preferred embodiment of the present invention, a certain type of humanoid robot performs inspection work on the deck of a bulk carrier. The deck has a grid structure and is subject to continuous roll and pitch disturbances due to the influence of a Force 6 wind and a wave height of 2 meters.

[0088] The robot activates its multi-source sensing system, synchronously collecting data at a sampling frequency of 100Hz: LiDAR scanning reveals a deck grating opening size of 30cm x 30cm, a channel width of 1.2m, and the presence of seawater in some areas; the inertial measurement unit acquires data showing a current platform roll angle amplitude of approximately 5° and a pitch angle amplitude of approximately 3°; a six-dimensional force sensor on the foot detects a vertical force of approximately 2500N on the current supporting foot, indicating slight tangential slippage; joint sensors collect position and torque data from all 24 joints, while simultaneously retrieving historical operational data from the past three gait cycles. All data, after normalization and time alignment, is fused into a unified state vector and input into subsequent modules.

[0089] Parameters such as roll angular velocity, pitch angular velocity, foot vertical force fluctuation, and joint micro-vibration amplitude are extracted from the state vector to construct a dynamic disturbance feature vector of the platform. A sliding window filter with a length of 1 / 4 gait cycle is used to obtain a smooth disturbance trend. It is determined that the current platform is in the roll enhancement stage. Combined with historical disturbance frequency (about 0.8 Hz) extrapolation prediction, the disturbance phase sequence in the next 1.5 gait cycles is obtained, and it is predicted that the platform will enter the roll peak stage after 0.6 s.

[0090] The reachable area within 1.5m in front of the robot was divided into 10cm×10cm grids, resulting in 120 candidate landing areas. Six indicators, including support area matching degree, edge integrity, and wet slip risk complementarity value, were calculated for each area and weighted according to preset weights to obtain the contact confidence. 22 areas with a contact confidence higher than 0.5 were selected as the next landing feasible domain, and 18 adjacent areas were marked as compensation reserve domains.

[0091] Combining the predicted perturbation phase sequence, candidate region contact confidence, current centroid position (0.12m from the support boundary), and normal walking gait parameters, the following margins were calculated: centroid safety margin 0.42, angular momentum impact margin 0.51, support degradation margin 0.48, and attitude overshoot margin 0.55. The minimum value of 0.42 was taken as the future stability support margin. The preset safety threshold is 0.45. If this margin is lower than the threshold, an instability risk is determined, triggering a gait switch.

[0092] Using a channel width of 1.2m and grid openings as spatial constraints, gaits that occupy excessive lateral width are eliminated; the disturbance matching degree and attitude recovery margin of the remaining gaits are calculated, and the low center of gravity anti-sway gait is selected as the target gait; the stride length deviation of 0.2m and the center of gravity height deviation of 0.15m between the current normal gait and the target gait are calculated; the gait parameters are updated step by step within a window of 1 / 2 gait cycle.

[0093] Using the baseline inertia, center of mass, and angular momentum of the low-center-of-mass anti-sway gait as targets, and combining the direction of the yaw-enhancing disturbance, the optimization direction is determined, and the optimization model is constructed with real-time deviation as a constraint. The attitude control quantities such as upper limb swing, trunk flexion, and knee joint compression are obtained by solving. The gait parameters and attitude control quantities are synchronously sent to the joint controller, and the robot smoothly completes the gait switching, remains stable during the peak yaw phase, and does not exhibit instability.

[0094] Most existing gait control methods for humanoid robots are designed for stable land surfaces, and have significant shortcomings when applied to offshore platforms: First, they are slow to perceive platform disturbances, and mostly rely on passive compensation after disturbances occur, failing to anticipate phase changes; second, their landing assessment dimensions are singular, considering only geometric distance and collision detection, ignoring unique risks at sea such as slipperiness, structural deformation, and disturbance phase, resulting in insufficient landing safety; third, gait switching is disconnected from posture control, lacking coordinated regulation of whole-body inertia and angular momentum during switching, resulting in weak anti-roll and anti-pitch capabilities and easy instability during switching; fourth, they lack an advanced prediction mechanism, often triggering adjustments only after stability has decreased, leading to delayed responses and a high risk of falls.

[0095] This invention integrates environmental, body, and foot data to comprehensively cover the disturbance characteristics and site risks of offshore platforms. It solves the problem that a single data source cannot accurately characterize complex offshore environments, providing a reliable data foundation for subsequent control and improving data accuracy. Through phase recognition and short-term prediction, disturbance response is upgraded from passive compensation to proactive prediction, allowing for awareness of disturbance trends 1-2 gait cycles in advance. This provides sufficient response time for gait adjustments and significantly reduces the impact of platform disturbances on robot stability. Landing safety is comprehensively evaluated from six dimensions: geometry, surface, structure, and phase, replacing the traditional single-geometric assessment method. This effectively reduces the probability of landing failure caused by unique offshore risks such as slipperiness, grid openings, and structural deformation. Gait switching and whole-body inertia reconstruction are performed simultaneously. By reshaping the mass distribution and angular momentum transfer path, anti-overturning capability is improved, solving the problem of gait and attitude control being disconnected in traditional methods and enhancing anti-roll and anti-pitch capabilities during the switching process. By calculating the future stability margin in advance based on the prediction results, gait switching is triggered before the risk of instability occurs, thus preventing problems before they occur and avoiding the lag of passive adjustment, ensuring the continuity and stability of robot movement.

[0096] By implementing this embodiment, dynamic disturbance characteristics of the platform are extracted from the platform's sensing data, and phase prediction is performed to obtain a predicted disturbance phase sequence, enabling the prediction of the disturbance evolution trend of the offshore operating platform over a future period. After obtaining the contact confidence level based on the predicted disturbance phase sequence and the platform's sensing data, the stability support margin in the next gait cycle is calculated by combining the current centroid state and current gait parameters from the robot's operating data. The stability support margin can identify potential instability risks in advance, proactively trigger gait adjustments, and overcome the lag problem of traditional feedback control. Based on the whole-body inertia reconstruction optimization mechanism... By integrating updated gait parameters, operational deviations, and predicted disturbance phases to collaboratively solve attitude control parameters, the whole-body inertia reconstruction optimization reshapes the overall inertia distribution by adjusting the whole-body attitude, actively counteracting the unbalanced impact caused by combined roll and pitch disturbances on the offshore platform. Compared with independent gait and attitude control methods, this approach has stronger balance maintenance capabilities. Finally, based on the attitude control parameters and the updated current gait parameters, the robot under control is controlled to smoothly switch between gait and attitude, achieving integrated collaborative control of gait and attitude. This improves balance under dynamic disturbances, enhances the robot's continuous operational stability in the offshore environment, and meets the needs of offshore operations.

[0097] See Figure 2 This is a schematic diagram of the structure of a robot gait posture cooperative control device in a marine environment according to an embodiment of the present invention, comprising: The operation platform sensing data acquisition module is used for the operation platform sensing data, and also includes: offshore operation area data, offshore meteorological data, and foot contact status data; The disturbance phase sequence prediction module is used to extract dynamic disturbance features of the platform based on the sensing data of the operating platform, perform phase prediction on the dynamic disturbance features of the platform, and obtain a predicted disturbance phase sequence. The contact confidence calculation module is used to calculate the contact confidence based on the sensing data of the work platform and the predicted disturbance phase sequence. The stability support margin calculation module is used to calculate the stability support margin in the next gait cycle based on the contact confidence, the predicted disturbance phase sequence, the current centroid state and the current gait parameters in the robot's running data. The target gait matching module is used to match a target gait from a preset gait set when the stability support margin is less than a preset safety threshold, obtain the running deviation between the target gait and the current gait, and update the current gait parameters based on the target gait. The collaborative control module is used to reconstruct and optimize the whole-body inertia based on the updated current gait parameters, running deviation, and predicted disturbance phase sequence to obtain attitude control quantities. Based on the attitude control quantities and the updated current gait parameters, it controls the robot to be controlled to complete the smooth switching between gait and attitude.

[0098] This invention provides a robot gait and posture collaborative control device for a marine environment. It extracts dynamic disturbance characteristics of the platform based on the platform's perception data and performs phase prediction to obtain a predicted disturbance phase sequence, enabling the prediction of the disturbance evolution trend of the marine platform over a future period. After obtaining the contact confidence level based on the predicted disturbance phase sequence and the platform's perception data, it calculates the stability support margin for the next gait cycle by combining the current centroid state and current gait parameters from the robot's operational data. This stability support margin can identify potential instability risks in advance and proactively trigger gait adjustments, overcoming the lag problem of traditional feedback control. Based on the full... The whole-body inertia reconstruction optimization mechanism integrates updated gait parameters, operational deviations, and predicted disturbance phases to collaboratively solve attitude control quantities. By adjusting the whole-body attitude, the whole-body inertia reconstruction optimization reshapes the overall inertia distribution, actively counteracting the unbalanced impact caused by the combined yaw and pitch disturbances of the offshore platform. Compared with independent gait and attitude control methods, it has stronger balance maintenance capabilities. Finally, based on the attitude control quantities and the updated current gait parameters, the robot under control is controlled to smoothly switch between gait and attitude, realizing integrated collaborative control of gait and attitude, improving balance capabilities under dynamic disturbances, enhancing the robot's continuous operational stability in the offshore environment, and meeting the needs of offshore operations.

[0099] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0100] Those skilled in the art will understand that, for convenience and brevity, the specific working process of the device described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0101] Another embodiment of the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements a robot gait and posture cooperative control method in a marine environment as described in the above embodiments. The terminal device may be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0102] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the terminal device, connecting all parts of the terminal device via various interfaces and lines.

[0103] The memory can be used to store the computer program. The processor implements various functions of the terminal device by running or executing the computer program stored in the memory and calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function, etc.; the data storage area may store data created based on the use of the mobile phone, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, RAM, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device or other volatile solid-state storage device.

[0104] Another embodiment of the present invention provides a computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform a robot gait posture cooperative control method in a marine environment as described in the above embodiment.

[0105] The storage medium is a computer-readable storage medium, and the computer program is stored in the computer-readable storage medium. When the computer program is executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0106] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for cooperative control of robot gait and posture in a marine environment, characterized in that, include: Acquire sensing data from the work platform; wherein, the sensing data from the work platform includes robot operation data; Based on the sensing data of the operating platform, the dynamic disturbance features of the platform are extracted, and the phase of the dynamic disturbance features of the platform is predicted to obtain the predicted disturbance phase sequence. The contact reliability is calculated based on the sensing data of the operating platform and the predicted disturbance phase sequence. Based on the contact confidence level, the predicted perturbation phase sequence, the current centroid state and current gait parameters in the robot's operating data, the stability support margin in the next gait cycle is calculated. When the stability support margin is less than the preset safety threshold, the target gait is matched from the preset gait set, the running deviation between the target gait and the current gait is obtained, and the current gait parameters are updated based on the target gait. Based on the updated current gait parameters, running deviation, and predicted perturbation phase sequence, the whole-body inertia is reconstructed and optimized to obtain the attitude control quantity. Based on the attitude control quantity and the updated current gait parameters, the robot under control is controlled to complete the smooth switching between gait and attitude.

2. The method for cooperative control of robot gait and posture in a marine environment as described in claim 1, characterized in that, The sensing data of the operation platform also includes: offshore operation area data, marine meteorological data, and foot contact status data; Based on the sensing data of the operating platform, dynamic disturbance features of the platform are extracted, and phase prediction is performed on the dynamic disturbance features of the platform to obtain a predicted disturbance phase sequence, including: Environmental features are extracted from the marine meteorological data, landing area features are extracted from the marine operation area data, attitude angle features are extracted from the robot operation data, and force contact features are extracted from the foot contact state data. Based on the environmental features, landing area features, attitude angle features, and force contact features, platform dynamic disturbance features are constructed. The platform's dynamic disturbance characteristics are subjected to sliding window time-domain filtering based on a preset filtering window, and the current disturbance phase is determined based on the filtered platform dynamic disturbance characteristics. The length of the filtering window is determined according to the next state period of the robot to be controlled. The disturbance phase includes the roll enhancement stage, roll fall stage, pitch enhancement stage, pitch fall stage, combined roll and pitch disturbance stage, and relatively stable stage. Phase prediction is performed based on the current disturbance phase and environmental characteristics to obtain the predicted disturbance phase sequence.

3. The method for cooperative control of robot gait and posture in a marine environment as described in claim 2, characterized in that, The contact reliability is calculated based on the sensing data and predicted disturbance phase sequence of the operating platform, including: Based on the data of the marine operation area, the landable area of ​​the robot to be controlled is divided into grids to obtain several candidate landable areas. For each candidate footing area, the support area matching degree is calculated based on the effective contact area of ​​the foot in the foot contact state data and the grid opening area of ​​the candidate footing area. The integrity of the support edge is calculated based on the invalid side length of the candidate landing area and the outer contour side length of the candidate landing area. Based on the surface slip risk value in the data of the offshore operation area, the complementary value of surface slip risk is calculated; Based on the deck elastic deformation values ​​in the offshore operation area data, the structural elastic deformation risk complementarity value is calculated. Based on the foot landing posture parameters in the robot's operation data, the corresponding perturbation phase is matched from the predicted perturbation phase sequence, and the angle between the foot's force direction and the normal direction of the support area is determined based on the corresponding perturbation phase and the foot landing posture parameters. The stability support contribution value is determined from the preset stability support table based on the corresponding disturbance phase; The contact reliability is obtained by weighting the support area matching degree, support edge integrity, surface slip risk complementarity value, structural elastic deformation risk complementarity value, included angle, and stable support contribution value.

4. The method for cooperative control of robot gait and posture in a marine environment as described in claim 3, characterized in that, Based on the contact confidence level, the predicted perturbation phase sequence, the current centroid state and current gait parameters in the robot's operational data, the stability support margin within the next gait cycle is calculated, including: Based on the current centroid state in the robot's operating data and the platform attitude change trend corresponding to the predicted perturbation phase sequence, the centroid safety margin of the robot's centroid projection relative to the support boundary is obtained; wherein, the support boundary is composed of the contours of the candidate landing area and the effective contact area of ​​the foot. Based on the reference angular momentum of the whole machine corresponding to the current gait parameters and the disturbance impact amplitude corresponding to the predicted disturbance phase sequence, the balance impact margin used to characterize the change in angular momentum is calculated. Based on the contact reliability of the candidate landing area and the preset contact reliability threshold, the support degradation margin used to characterize the change in foot contact reliability is calculated. Based on the current gait parameters and the attitude change amplitude corresponding to the predicted perturbation phase sequence, the attitude over-limit risk margin is calculated. The minimum value among the centroid safety margin, balance impact margin, support degradation margin, and attitude over-limit risk margin is taken as the stability support margin in the next state cycle.

5. The method for cooperative control of robot gait and posture in a marine environment as described in claim 3, characterized in that, Match the target gait from a preset set of gaits, including: Using the channel width and obstacle distribution information in the offshore operation area data as spatial constraints, each candidate gait in the preset gait set is screened, and candidate gaits whose lateral occupancy width exceeds the channel width and whose turning radius conflicts with the obstacle distribution information are eliminated, thus obtaining a feasible gait set. The perturbation phase matching degree of each feasible gait in the feasible gait set is calculated based on the predicted perturbation phase sequence, and the attitude recovery margin is calculated based on the perturbation phase matching degree and the attitude recovery capability parameters in the robot's running data. The minimum contact confidence level of the candidate landing area corresponding to each feasible gait in the feasible gait set is determined based on the contact confidence level. The feasible gait with a minimum contact confidence level greater than a preset minimum threshold and an attitude recovery margin greater than a preset recovery threshold is taken as the target gait. The target gait includes steady gait, narrowing conservative gait, widened support gait, side-stepping convergence gait, hurdle-crossing avoidance gait, and low center of gravity anti-shaking gait.

6. The method for cooperative control of robot gait and posture in a marine environment as described in claim 1, characterized in that, The operational deviations include: posture deviation, sudden change in foot contact, center of mass trajectory offset, and gait phase offset; Based on the updated current gait parameters, running deviation, and predicted perturbation phase sequence, whole-body inertia reconstruction and optimization are performed to obtain attitude control variables, including: The reference inertia distribution, reference center of mass position, and reference angular momentum are calculated based on the updated current gait parameters. Determine the direction of inertia reconstruction optimization during the current gait switching process based on the predicted perturbation phase sequence; Using attitude deviation, foot contact abrupt change, center of mass trajectory offset, and gait phase offset as real-time deviation constraints, a target optimization model is constructed based on the reference inertia distribution, reference center of mass position, and reference angular momentum. The target optimization model is solved based on the inertia reconstruction optimization direction to obtain the attitude control variables.

7. The method for cooperative control of robot gait and posture in a marine environment as described in claim 1, characterized in that, Update the current gait parameters based on the target gait, including: Within a preset gait phase window, the current gait parameters and the parameters corresponding to the target gait are progressively updated by polynomial interpolation to obtain the updated current gait parameters. The current gait parameters include: stride length, stride width, stride frequency, support ratio, and leg swing height.

8. A robot gait and posture cooperative control device for use in a marine environment, characterized in that, include: The operation platform perception data acquisition module is used to acquire operation platform perception data; wherein, the operation platform perception data includes robot operation data; The disturbance phase sequence prediction module is used to extract dynamic disturbance features of the platform based on the sensing data of the operating platform, perform phase prediction on the dynamic disturbance features of the platform, and obtain a predicted disturbance phase sequence. The contact confidence calculation module is used to calculate the contact confidence based on the sensing data of the work platform and the predicted disturbance phase sequence. The stability support margin calculation module is used to calculate the stability support margin in the next gait cycle based on the contact confidence, the predicted disturbance phase sequence, the current centroid state and the current gait parameters in the robot's running data. The target gait matching module is used to match a target gait from a preset gait set when the stability support margin is less than a preset safety threshold, obtain the running deviation between the target gait and the current gait, and update the current gait parameters based on the target gait. The collaborative control module is used to reconstruct and optimize the whole-body inertia based on the updated current gait parameters, running deviation, and predicted disturbance phase sequence to obtain attitude control quantities. Based on the attitude control quantities and the updated current gait parameters, it controls the robot to be controlled to complete the smooth switching between gait and attitude.

9. A terminal device, characterized in that, The system includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements a robot gait posture cooperative control method in a marine environment as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform a robot gait posture cooperative control method in a marine environment as described in any one of claims 1 to 7.

Images