Motion control method and system for intelligent robot

Abstract

The invention provides a motion control method and system for an intelligent robot, and the method comprises the steps: collecting environment and state data through an intelligent sensor group of a humanoid robot, inputting the environment and state data into a pre-training first neural network model, and obtaining motion prediction data and an environment analysis result; constructing a motion planning model, and performing energy consumption-stability multi-objective optimization on the joint motion track by adopting a preset first algorithm; the central controller generates a joint position, speed and torque reference trajectory based on an optimization result; and the local second controller of each joint locally adjusts the reference trajectory within the prediction time domain according to the real-time feedback. According to the invention, multiple sensors are combined with the mixed attention neural network to realize environment and self state intelligent perception, and the problem of multi-sensor data fusion time sequence dependence is solved; through energy consumption-stability multi-objective optimization, the complex environment movement efficiency is remarkably improved; the central controller and the local controller work cooperatively, and in combination with an edge computing architecture, the communication delay is reduced, and the system response speed is improved.

Description

Technical Field

[0001] The present invention relates to the field of robotics technology, and in particular to a motion control method and system for an intelligent robot. Background Art

[0002] With the development of intelligent technology, intelligent robots are increasingly being used in industrial manufacturing, medical rehabilitation, and special task execution. However, because intelligent robots typically have multi-degree-of-freedom joint systems, existing robot motion control methods have shortcomings in terms of motion efficiency and response speed. Summary of the Invention

[0003] Based on the above problems, the present invention proposes a motion control method and system for intelligent robots. Through the combination of multiple sensors and hybrid attention neural networks, intelligent perception of the environment and the robot's own state is realized, and the timing dependence problem of multi-sensor data fusion is solved; energy consumption-stability multi-objective optimization is performed, which significantly improves the motion efficiency in complex environments; through the collaborative work of the central controller and the local controller, combined with the edge computing architecture, the communication delay is reduced and the system response speed is improved.

[0004] In view of this, one aspect of the present invention provides a motion control method for an intelligent robot, comprising:

[0005] Collect environmental data and robot status data through an intelligent sensor group installed on the humanoid robot;

[0006] Inputting the environmental data and the robot state data into a pre-trained first neural network model to obtain robot motion prediction data and environmental analysis results;

[0007] Constructing a motion planning model based on the motion prediction data and the environmental analysis result;

[0008] Using a preset first algorithm to perform energy consumption-stability multi-objective optimization on the joint motion trajectory in the motion planning model;

[0009] The central controller generates reference trajectories of joint positions, velocities, and torques based on the optimization results;

[0010] The second controller configured locally at each joint makes local adjustments to the reference trajectory within the prediction time domain based on real-time feedback data.

[0011] Optionally, the step of inputting the environmental data and the robot state data into a pre-trained first neural network model to obtain the robot's motion prediction data and environmental analysis results includes:

[0012] Preprocessing the environmental data and the robot status data;

[0013] Inputting the preprocessed environment data and the robot state data into the spatiotemporal attention sub-model of the first neural network model;

[0014] The spatiotemporal attention sub-model establishes an attention weight matrix for the time dimension;

[0015] The temporal attention mechanism is used to calculate the correlation between the current moment and the historical moment data, and identify the temporal dependencies of the motion patterns.

[0016] Analyze the spatial correlation between different sensor data based on the spatial attention mechanism and establish a mutual influence relationship diagram between sensors;

[0017] Fuse the temporal and spatial attention weights to generate a comprehensive spatiotemporal attention distribution matrix;

[0018] According to the spatiotemporal attention distribution matrix, the sensor fusion adaptive weight sub-model calculates the fusion weight coefficient of each sensor data;

[0019] Dynamically adjust the importance weights of different sensor data based on the current environment complexity and task requirements;

[0020] Perform weighted fusion on the weight-adjusted multimodal data to form a unified feature representation vector;

[0021] Control the information flow through the gating mechanism to filter out the sensor information that is most critical to the current movement decision;

[0022] The fused feature representation vector is input into the motion prediction sub-model to extract the robot's current motion trend and goal orientation;

[0023] The feature representation vector is simultaneously input into the environment analysis sub-model to analyze the geometric structure, obstacle distribution and dynamic changes of the surrounding environment;

[0024] Combine motion intention and environmental understanding results to generate motion prediction data and environmental analysis results.

[0025] Optionally, the step of constructing a motion planning model based on the motion prediction data and the environmental analysis result includes:

[0026] Constructing the first sub-model includes: establishing a global task decomposition tree based on the motion prediction data and the environmental analysis results, combined with preset task target parameters; generating a global path from the starting point to the target point based on the obstacle distribution information and terrain characteristics in the environmental analysis results; dividing the global path into multiple sub-path segments based on task complexity and environmental dynamics, each sub-path segment corresponding to a behavior primitive; assigning execution priority and time window to each behavior primitive, generating a time-series execution sequence including basic behaviors such as walking, turning, grasping, and obstacle avoidance; and establishing a library of alternative solutions for the global path, enabling rapid switching to alternative paths when the main path is blocked.

[0027] Constructing a second sub-model includes: receiving the behavior sequence output by the first sub-model, calling the corresponding motion pattern library for each behavior primitive; decomposing the behavior primitive into a set of motion trajectories for each joint based on the robot's kinematic and dynamic constraints; calculating the position, velocity, and acceleration target values of each joint using an inverse kinematics algorithm, taking into account the coupling relationship and coordination constraints between joints; performing feasibility testing and collision detection on the joint trajectories based on the environmental constraint information in the environmental analysis results; locally adjusting and optimizing trajectories that do not meet the constraint conditions to ensure the smoothness and continuity of the trajectories; generating a detailed joint motion trajectory table including timestamps, and establishing a checkpoint mechanism for trajectory execution;

[0028] The third sub-model is constructed, including: establishing a multi-threaded monitoring mechanism to detect environmental changes and emergencies in real time, including the appearance of dynamic obstacles, changes in ground conditions, and external interference; when an emergency is detected, immediately launching an emergency response program, pausing the current trajectory execution and entering the reaction control mode; based on the current sensor data and robot status, quickly assessing the threat level and impact range of the emergency; selecting the corresponding response strategy according to the threat level: fine-tuning the trajectory for minor interference, initiating avoidance actions for moderate threats, and executing an emergency stop for severe threats; in the reaction control mode, using a rule-based rapid decision-making algorithm to generate local avoidance trajectories or protective actions; evaluating the execution effect of the reaction action in real time, and when the threat is eliminated, replanning the connection trajectory from the current state to the original target;

[0029] Construct coordination and information transmission mechanisms among sub-models;

[0030] Planning results output and verification.

[0031] Optionally, the adopting of a preset first algorithm to perform energy consumption-stability multi-objective optimization on the joint motion trajectory in the motion planning model includes:

[0032] Construct a composite objective function including energy consumption, motion stability and task completion;

[0033] Utilize quantum annealing to rapidly search the solution space to achieve optimal motion control parameters in complex terrain and dynamic environments;

[0034] The joint stiffness and damping parameters are dynamically adjusted according to the optimization results, so that the robot can minimize energy consumption while maintaining stability.

[0035] Optionally, the step of generating reference trajectories of joint positions, velocities, and torques based on the optimization results by the central controller includes:

[0036] Perform optimization result analysis and parameter extraction;

[0037] Generate joint position reference trajectory based on the data obtained from optimization result analysis and parameter extraction;

[0038] Generate joint velocity reference trajectory based on joint position reference trajectory;

[0039] Generate joint torque reference trajectory based on robot dynamics model, joint position reference trajectory and joint velocity reference trajectory;

[0040] Perform multi-joint coordination and trajectory synchronization;

[0041] The joint position reference trajectory, joint velocity reference trajectory and joint torque reference trajectory are integrated and distributed.

[0042] Optionally, the step of locally adjusting the reference trajectory in the prediction time domain according to the real-time feedback data by the second controller configured locally at each joint includes:

[0043] The current joint state data of the joint is collected in real time through the built-in position encoder, speed sensor and torque sensor of the joint;

[0044] Align the current joint state data with the reference trajectory in time synchronization, and calculate the trajectory deviation between the current joint state and the reference trajectory;

[0045] Based on the current joint state data, establish the prediction time domain and perform state prediction;

[0046] Analyze the trajectory deviation and formulate the corresponding current adjustment strategy;

[0047] Replan and optimize the corresponding local trajectory in the parameter trajectory based on the current state data and the current adjustment strategy;

[0048] Performs adjacent joint coordination and conflict resolution;

[0049] Evaluate the effects of adjustments and provide feedback.

[0050] Optionally, the sensor fusion adaptive weight sub-model of the first neural network model calculates the dynamic weight of each sensor data in the following manner:

[0051] in,

[0052] W i1 (t) represents the weight coefficient of the i1th sensor at time point t; R i1 (t) represents the reliability score of the i1th sensor data, which is calculated by the following formula:

[0053]

[0054] in is the sensor data variance, D i1 (t) is the data packet loss rate, Δ i1 (t) is the consistency deviation of historical data, β1, β2 and β3 are balance coefficients and satisfy β1+β2+β3=1, λ i1 and ξ i1 is the characteristic parameter of the sensor;

[0055] C i1 (t) represents the relevance score between the i1th sensor and the current task, which is generated by the task context encoder to increase the adaptability of environmental conditions;

[0056] T i1 (t) represents the timeliness factor of the i1-th sensor, which is defined as:

[0057] T i1 (t) = exp(-γ i1 (tt last，i1 ))·(1-ν i1 exp(-κ i1 f i1 (t)));

[0058] where t last，i1 is the timestamp of the most recent valid data of the i1th sensor, γ i1 is the time decay parameter, f i1 (t) is the sensor data sampling frequency, v i1 and κ i1 is the adjustment parameter;

[0059] Q i (t) represents the newly added sensor data quality assessment factor, which is defined as:

[0060]

[0061] Where SNR i1 (t) is the sensor signal-to-noise ratio, SNRth，i1 is the threshold constant, ρ i1 and To adjust the parameters, this factor can dynamically adjust the weight according to the sensor signal-to-noise ratio;

[0062] H i1 (t) represents the newly added sensor historical performance reputation, which is updated by exponential sliding average:

[0063] H i1 (t) = α H H i1 (t-1)+(1-α H )P i1 (t);

[0064] Among them, P i1 (t) is the current performance score, α H is the smoothing coefficient;

[0065] N is the total number of multimodal sensors.

[0066] Optionally, the first algorithm includes:

[0067] S11: Construct a quantum representation of the energy-stability multi-objective optimization, evaluated by the following composite objective function:

[0068] F(θ)=α E E(θ)+α S (1-S(θ))+α T (1-T(θ))+α F ·(1-F L (θ))+α A ·A D (θ);

[0069] Where θ represents the robot joint control parameter vector;

[0070] E(θ) represents the energy consumption function, which is improved to be defined as:

[0071]

[0072] Among them, P i2 (t, θ) is the power consumption of the i2th joint at time t, a i2 (t, θ) is the joint acceleration, ζ i2 is the acceleration energy consumption coefficient, M is the total number of joints, t f is the task completion time; the additional impact of acceleration on energy consumption is taken into account;

[0073] S(θ) represents the motion stability function, which is defined as:

[0074]

[0075] Among them ZMP dist (t, θ) is the minimum distance from the zero moment point to the boundary of the supporting polygon, A CoM (t, θ) is the acceleration of the center of mass, J ang (t, θ) is the rate of change of angular momentum, and μ1, μ2, and μ3 are weight coefficients. The addition of the rate of change of angular momentum can more comprehensively evaluate dynamic stability.

[0076] T(θ) represents the task completion function, which quantifies the degree of achievement of the task goal;

[0077] F L (θ) represents the load adaptability function, which is defined as:

[0078]

[0079] Among them F ext (t, θ) is the external load force, F est (t, θ) is the estimated load force, T comp (t, θ) is the load compensation torque, ψ1 and ψ2 are weight coefficients; this function evaluates the robot's ability to adapt to changes in external load;

[0080] A D (θ) represents the action diversity index, which is defined as:

[0081]

[0082] Among them D KL is the KL divergence, p θ is the current action distribution, p ref，k is the reference action template distribution, w k is the weight coefficient, K is the number of action templates; this indicator promotes the exploration of solution space diversity;

[0083] α E , α S , α T , α F and α A is the adaptive weight coefficient, satisfying α E +α S +α T +α F +α A =1;

[0084] S12: Apply qubit probability amplitude coding to represent each candidate solution in the solution space and update the solution space by the following quantum annealing probability function:

[0085]

[0086] Among them, τ(t)=τ0·exp(-η·t / T max )·(1+δ·sin(π·t / t p )) represents the quantum annealing temperature function, with the addition of a periodic fluctuation term, τ0 being the initial temperature parameter, η being the annealing rate, and t max is the maximum number of iterations, δ is the fluctuation amplitude coefficient, t p is the fluctuation period;

[0087] is the distance adjustment factor of the solution space, h(x j2 , x k ) is the Hamming distance, χ and is the adjustment parameter;

[0088] The quantum search network is:

[0089]

[0090] in, is the unitary transformation matrix of the i3th layer, representing the evolution of the quantum system;

[0091] is the perturbation operator, parameter ε i3 Controlling quantum exploration – exploiting balance;

[0092] is the newly added quantum measurement operator, σ i3 To measure the intensity parameters, quantum collapse is implemented to prevent excessive divergence of the solution space; L is the number of quantum network layers.

[0093] Optionally, the method further comprises the steps of:

[0094] S21: The central controller generates feedforward control signals based on the robot dynamics model and optimized trajectory using the following state-space prediction equation:

[0095]

[0096] in:

[0097] represents the system state vector at time t+k1 predicted at time t; X(t) represents the actual system state vector at time t, which contains position, velocity and acceleration information; U(t+j3|t) represents the control input at time t+j3 planned at time t; Φ(t) is the time-varying state transfer matrix, which describes the dynamic characteristics of the system and is updated as follows: Φ(t) = Φ0 + ΔΦ(t); where Φ0 is the nominal model and ΔΦ(t) is the model deviation identified online; Γ(t) is the time-varying control input matrix, taking into account the change in actuator efficiency: Γ(t) = Γ0·E(t); where Γ0 is the nominal control input matrix and E(t) is the actuator efficiency diagonal matrix;

[0098] Ψ(t+k1|t) is the environmental disturbance prediction term, which is estimated by a high-order adaptive Kalman filter:

[0099]

[0100] Among them, Ψi4 is the i4th type of disturbance model, Ψi4j3 is the disturbance interaction term, ω i4 (t) is the first-order adaptive weight, ωi4j3(t) is the second-order interaction weight, N d is the size of the perturbation model library;

[0101] Ω(t+k1|t) represents the contact dynamics prediction term, which is defined as:

[0102]

[0103] where N c is the number of possible contact points, p c (t+k1|t) is the contact probability prediction, J c (t+k1|t) is the contact Jacobian matrix, F c (t+k1|t) is the contact force prediction;

[0104] Λ(t, k1) represents the model uncertainty compensation term, which is defined as:

[0105] Λ(t,k1)=B(t)·tanh(W(t)·Z(t,k1));

[0106] Where B(t) and W(t) are adaptive parameter matrices, Z(t, k) is the eigenvector, and tanh is the hyperbolic tangent function;

[0107] S22: The second controller (local controller) of each joint adopts the sliding time domain model predictive control strategy to achieve accurate tracking by solving the following constrained optimization problem:

[0108]

[0109] Satisfy the constraints:

[0110] Umin(t)≤U(t+j4|t)≤Umax(t), j4=0, 1,..., H c -1;

[0111] ΔUmin(t)≤ΔU(t+j4|t)≤ΔUmax(t), j4=0, 1,..., H c -1;

[0112]

[0113] in:

[0114] Y represents the system output at time t+j4 predicted at time t; ref (t+j4) represents the reference trajectory; ΔU(t+j4|t) represents the control increment; H p Indicates the length of the prediction time domain; H c Indicates the control time domain length, satisfying H c ≤H p ; Q(t, j4) and R(t, j4) are the time-varying output tracking error and the control increment weight matrix respectively; Y tar get is the terminal target state, S is the terminal weight matrix, and the control stability is enhanced; Γ smooth is a smooth penalty function to promote smooth transition of control signals; C safety It is a safety constraint function that ensures the robot avoids dangerous states;

[0115] S23: The edge computing architecture dynamically schedules computing loads through an adaptive resource allocation algorithm, defined as:

[0116]

[0117] Satisfy the constraints:

[0118]

[0119] R reliability (A(t))≥R min ;

[0120] in:

[0121] A(t) represents the task allocation matrix, element A i5j5 (t) represents the indicator variable that the i5th control task is assigned to the j5th edge computing node; C i5j5 (t) represents the computational cost of the i5th task at the j5th node; L i5j5 (t) represents the communication delay cost; E i5j5 (t) represents the additional energy consumption cost; Ii5j5 (t) represents the newly added information security risk indicator; D i5 (t) represents the computing resource requirement of the i5th task; K j5 (t) represents the available computing resources of the j5th node; P i5 (t) represents the power requirement of the i5th task; P max，j5 (t) represents the maximum power limit of the j5th node; R reliability (A(t)) represents the overall reliability function of the system; R min Represents the minimum reliability requirement; M1 is the total number of control tasks; N e is the number of edge computing nodes; w1(t), w2(t), w3(t) and w4(t) are time-varying weight coefficients, satisfying w1(t)+w2(t)+w3(t)+w4(t)=1.

[0122] Another aspect of the present invention provides a motion control system for an intelligent robot, which is used to execute a motion control method for an intelligent robot, comprising: a humanoid robot; a central controller; an intelligent sensor group and a second controller provided on the humanoid robot; and a server;

[0123] The intelligent sensor group is configured to: collect environmental data and robot status data;

[0124] The server is configured to:

[0125] Inputting the environmental data and the robot state data into a pre-trained first neural network model to obtain robot motion prediction data and environmental analysis results;

[0126] Constructing a motion planning model based on the motion prediction data and the environmental analysis results, the model comprising:

[0127] Using a preset first algorithm to perform energy consumption-stability multi-objective optimization on the joint motion trajectory in the motion planning model;

[0128] The central controller is configured to: generate reference trajectories of joint positions, velocities, and torques based on the optimization results;

[0129] The second controller configured locally at each joint is configured to: locally adjust the reference trajectory within the prediction time domain based on real-time feedback data.

[0130] The technical solution of the present invention is used in a motion control method for an intelligent robot, including: collecting environmental data and robot status data through an intelligent sensor group provided on a humanoid robot; inputting the environmental data and robot status data into a pre-trained first neural network model to obtain the robot's motion prediction data and environmental analysis results; constructing a motion planning model based on the motion prediction data and the environmental analysis results; using a preset first algorithm to perform energy consumption-stability multi-objective optimization on the joint motion trajectory in the motion planning model; a central controller generates a reference trajectory of joint position, velocity, and torque based on the optimization results; a second controller configured locally at each joint performs local adjustment of the reference trajectory within the prediction time domain based on real-time feedback data. By combining multiple sensors with a hybrid attention neural network, intelligent perception of the environment and the robot's own state is achieved, solving the temporal dependency problem of multi-sensor data fusion; performing energy consumption-stability multi-objective optimization, significantly improving motion efficiency in complex environments; and by the collaborative work of the central controller and local controllers, combined with an edge computing architecture, reducing communication delay and improving system response speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0131] Figure 1 is a flow chart of a motion control method for an intelligent robot provided by one embodiment of the present invention;

[0132] Figure 2 This is a schematic block diagram of a motion control system for an intelligent robot provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0133] In order to more clearly understand the above-mentioned objects, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that, in the absence of conflict, the embodiments of the present application and the features therein can be combined with each other.

[0134] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the scope of protection of the present invention is not limited to the specific embodiments disclosed below.

[0135] The terms "first," "second," and the like in the specification and claims of this application and the accompanying drawings are used to distinguish between different objects, not to describe a particular order. Furthermore, the terms "including," "having," and any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus comprising a series of steps or elements is not limited to the listed steps or elements but may optionally include steps or elements not listed, or may optionally include other steps or elements inherent to the process, method, product, or apparatus.

[0136] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0137] Refer to the following Figures 1 to 2 A motion control method and system for an intelligent robot provided according to some embodiments of the present invention will be described.

[0138] like Figure 1 As shown, one embodiment of the present invention provides a motion control method for an intelligent robot, comprising:

[0139] Collect environmental data and robot status data through an intelligent sensor group installed on the humanoid robot;

[0140] The intelligent sensor group includes a depth camera, an inertial measurement unit, a torque sensor, a tactile sensor, and a thermal imaging camera;

[0141] Inputting the environmental data and the robot state data into a pre-trained first neural network model to obtain robot motion prediction data and environmental analysis results;

[0142] The first neural network model includes a spatiotemporal attention sub-model, a sensor fusion adaptive weight sub-model, a motion prediction sub-model, and an environmental analysis sub-model. It dynamically adjusts the attention paid to different sensor data through the spatiotemporal attention mechanism, solves the temporal dependency and spatial correlation problems in multi-sensor data fusion, and outputs motion prediction data and environmental analysis results for the humanoid robot.

[0143] Based on the motion prediction data and the environmental analysis results, a motion planning model is constructed, which includes: a first sub-model for generating a global path and behavior sequence based on the task objectives and the environmental analysis results; a second sub-model for decomposing the plan of the first sub-model into a set of specific joint motion trajectories; and a third sub-model for performing real-time motion adjustments in response to unexpected situations;

[0144] Using a preset first algorithm (such as a quantum heuristic optimization algorithm) to perform energy consumption-stability multi-objective optimization on the joint motion trajectory in the motion planning model;

[0145] The central controller generates reference trajectories of joint positions, velocities, and torques based on the optimization results;

[0146] In this step, the central controller performs an energy consumption-stability multi-objective optimization on the joint motion trajectory in the motion planning model based on the first algorithm to generate a reference trajectory of the joint position, velocity and torque.

[0147] The second controller configured locally at each joint makes local adjustments to the reference trajectory within the prediction time domain based on real-time feedback data.

[0148] The technical solution of this embodiment is adopted to achieve intelligent perception of the environment and its own status through the combination of multiple sensors and hybrid attention neural networks, solving the timing dependency problem of multi-sensor data fusion; energy consumption-stability multi-objective optimization is performed, significantly improving the motion efficiency in complex environments; through the collaborative work of the central controller and the local controller, combined with the edge computing architecture, the communication delay is reduced and the system response speed is improved.

[0149] In some possible implementations of the present invention, the step of performing an adaptive learning process includes:

[0150] During the execution of the movement, the control strategy is continuously updated through reinforcement learning algorithms to adapt to different environmental conditions and task requirements; a movement experience memory library is built to record successful and failed movement cases and their environmental context; based on the data in the memory library, a meta-learning strategy is used to quickly adapt to new environments and tasks;

[0151] In some possible implementations of the present invention, the steps of achieving human-machine collaborative motion control include:

[0152] The robot perceives the motion intentions of its human collaborator through visual and force sensors; adjusts its own motion trajectory and strength according to the actions of its human collaborator; and transmits the robot's motion status information to its human collaborator through tactile feedback, achieving two-way information exchange.

[0153] In some possible implementations of the present invention, the step of inputting the environmental data and the robot state data into a pre-trained first neural network model to obtain the robot's motion prediction data and environmental analysis results includes:

[0154] Preprocessing the environmental data and the robot state data includes: performing size normalization and noise filtering on the image data from the depth camera; performing coordinate system and time synchronization calibration on the acceleration and angular velocity data of the inertial measurement unit; performing amplitude normalization and outlier detection on the data of the torque sensor, tactile sensor, and thermal imaging camera; and aligning the sensor data of different modalities according to a unified timestamp to form a time series data stream.

[0155] Inputting the preprocessed environment data and the robot state data into the spatiotemporal attention sub-model of the first neural network model;

[0156] The spatiotemporal attention sub-model establishes an attention weight matrix for the time dimension;

[0157] The temporal attention mechanism is used to calculate the correlation between the current moment and the historical moment data, and identify the temporal dependencies of the motion patterns.

[0158] Analyze the spatial correlation between different sensor data based on the spatial attention mechanism and establish a mutual influence relationship diagram between sensors;

[0159] Fuse the temporal and spatial attention weights to generate a comprehensive spatiotemporal attention distribution matrix;

[0160] According to the spatiotemporal attention distribution matrix, the sensor fusion adaptive weight sub-model calculates the fusion weight coefficient of each sensor data;

[0161] Dynamically adjust the importance weights of different sensor data based on the current environment complexity and task requirements;

[0162] Perform weighted fusion on the weight-adjusted multimodal data to form a unified feature representation vector;

[0163] Control the information flow through the gating mechanism to filter out the sensor information that is most critical to the current movement decision;

[0164] The fused feature representation vector is input into the motion prediction sub-model to extract the robot's current motion trend and goal orientation;

[0165] The feature representation vector is simultaneously input into the environment analysis sub-model to analyze the geometric structure, obstacle distribution and dynamic changes of the surrounding environment;

[0166] Combining motion intention and environmental understanding results, motion prediction data (including motion direction, speed preference, etc.) and environmental analysis results (including environmental constraint information) are generated.

[0167] In this embodiment, the dynamic weight adjustment of the spatiotemporal attention mechanism effectively addresses the fusion challenges of temporal synchronization and spatial correlation between different sensor data, improving data fusion accuracy compared to traditional fixed-weight fusion methods. The adaptive weight sub-model adjusts the fusion strategy in real time based on environmental changes and sensor status. When a sensor fails or data quality degrades, the system automatically reduces its weight and enhances the contribution of other sensors, improving the system's fault tolerance and stability. Temporal dependency modeling enables more accurate prediction of the robot's motion trends, avoiding misjudgments caused by single-moment data fluctuations and improving the accuracy of motion intention recognition. Spatial correlation analysis enables the robot to better understand the inherent connections between sensor data, enabling more accurate identification of environmental features and potential risks in complex environments, improving its environmental adaptability. Information filtering through the gating mechanism reduces the amount of irrelevant information processed, ensuring decision quality while reducing computational complexity and enabling real-time control. This hybrid attention neural network architecture provides high-quality input data for subsequent multi-level motion planning and is a key technical foundation for achieving precise, stable, and adaptive robot motion control.

[0168] In some possible implementations of the present invention, the step of constructing a motion planning model based on the motion prediction data and the environmental analysis result includes:

[0169] Constructing the first sub-model includes: establishing a global task decomposition tree based on the motion prediction data and the environmental analysis results, combined with preset task target parameters; generating a global path from the starting point to the target point based on the obstacle distribution information and terrain characteristics in the environmental analysis results; dividing the global path into multiple sub-path segments based on task complexity and environmental dynamics, each sub-path segment corresponding to a behavior primitive; assigning execution priority and time window to each behavior primitive, generating a time-series execution sequence including basic behaviors such as walking, turning, grasping, and obstacle avoidance; and establishing a library of alternative solutions for the global path, enabling rapid switching to alternative paths when the main path is blocked.

[0170] Constructing a second sub-model includes: receiving the behavior sequence output by the first sub-model, calling the corresponding motion pattern library for each behavior primitive; decomposing the behavior primitive into a set of motion trajectories for each joint based on the robot's kinematic and dynamic constraints; calculating the position, velocity, and acceleration target values of each joint using an inverse kinematics algorithm, taking into account the coupling relationship and coordination constraints between joints; performing feasibility testing and collision detection on the joint trajectories based on the environmental constraint information in the environmental analysis results; locally adjusting and optimizing trajectories that do not meet the constraint conditions to ensure the smoothness and continuity of the trajectories; generating a detailed joint motion trajectory table including timestamps, and establishing a checkpoint mechanism for trajectory execution;

[0171] The third sub-model is constructed, including: establishing a multi-threaded monitoring mechanism to detect environmental changes and emergencies in real time, including the appearance of dynamic obstacles, changes in ground conditions, and external interference; when an emergency is detected, immediately launching an emergency response program, pausing the current trajectory execution and entering the reaction control mode; based on the current sensor data and robot status, quickly assessing the threat level and impact range of the emergency; selecting the corresponding response strategy according to the threat level: fine-tuning the trajectory for minor interference, initiating avoidance actions for moderate threats, and executing an emergency stop for severe threats; in the reaction control mode, using a rule-based rapid decision-making algorithm to generate local avoidance trajectories or protective actions; evaluating the execution effect of the reaction action in real time, and when the threat is eliminated, replanning the connection trajectory from the current state to the original target;

[0172] Construct a coordination and information transmission mechanism between sub-models, including: establishing an information transmission channel between sub-models, where the first sub-model transmits behavioral instructions and constraints to the second sub-model, and the second sub-model transmits trajectory parameters and execution status to the third sub-model; the third sub-model feeds back emergency information and processing results to the second sub-model, and the second sub-model adjusts subsequent trajectory planning based on the feedback information; when the third sub-model is frequently triggered, the second sub-model reports the abnormal information to the first sub-model, and the first sub-model re-evaluates the feasibility of the global path; establish a priority arbitration mechanism between sub-models to ensure that when multiple sub-models generate control instructions simultaneously, they are selected according to the principle of safety first;

[0173] Planning result output and verification include: integrating the planning results of each sub-model to generate a complete multi-level motion control instruction set; performing consistency checks on the integrated instruction set to ensure that there are no conflicts between the instructions of each sub-model and that they comply with the physical constraints of the robot; establishing a planning quality assessment mechanism to evaluate the safety, efficiency, and executability of the generated motion plan; and outputting the verified motion planning results to the first algorithm for processing.

[0174] The solution of this embodiment realizes hierarchical control from macro-task planning to micro-action execution through the division of labor and cooperation of three motion planning sub-models, so that complex motion tasks can be decomposed in an orderly manner and executed efficiently, improving the maintainability and scalability of the system; the first sub-model ensures the optimality of the global path and the feasibility of task completion, while the local adjustment mechanism of the second sub-model and the third sub-model ensures rapid adaptation to environmental changes, realizing the organic combination of global optimization and local flexibility; the rapid response mechanism of the third sub-model can handle emergencies within milliseconds, greatly improving the safety of the robot in a dynamic environment; by assigning planning tasks of different time scales and complexities to different sub-models, the computational bottleneck of a single algorithm to handle all problems is avoided, thereby improving the overall computational efficiency; multi-model hierarchical planning ensures the continuity and smoothness of motion, reduces frequent trajectory replanning and sudden motion changes, and makes the robot motion more natural and smooth, in line with human motion habits and expectations.

[0175] In some possible implementations of the present invention, the use of a preset first algorithm to perform energy consumption-stability multi-objective optimization on the joint motion trajectory in the motion planning model includes:

[0176] Construct a composite objective function including energy consumption, motion stability and task completion;

[0177] Utilize quantum annealing to rapidly search the solution space to achieve optimal motion control parameters in complex terrain and dynamic environments;

[0178] The joint stiffness and damping parameters are dynamically adjusted according to the optimization results, so that the robot can minimize energy consumption while maintaining stability.

[0179] In some possible implementations of the present invention, the step of generating a reference trajectory of joint position, velocity, and torque by the central controller based on the optimization result includes:

[0180] Optimization results analysis and parameter extraction, including: the central controller extracts the target position sequence, expected velocity curve, and required torque distribution information of each joint from the optimization results; the optimization results include the optimal motion control parameter set and dynamically adjusted joint stiffness and damping parameters; the time allocation scheme in the optimization results is analyzed to determine the duration and transition time window of each motion phase; environmental constraint parameters and safety boundary conditions are extracted to provide a constraint basis for subsequent trajectory generation; and a mapping relationship table between the optimization parameters and the robot's physical parameters is established to ensure the accuracy of parameter conversion.

[0181] Generate joint position reference trajectories based on the data obtained from the optimization result analysis and parameter extraction, including: generating continuous position trajectories of each joint using a spline interpolation algorithm based on the extracted target position sequence and time allocation scheme; performing a smoothness test on the generated position trajectory to ensure that the trajectory is within the joint limit range and has no mutation points; performing reachability verification and singularity avoidance on the position trajectory based on the robot kinematic constraints; performing synchronization correction on the position trajectory of multi-joint collaborative motion considering the coupling relationship between joints; establishing a timestamp index for the position trajectory to ensure the precise temporal alignment of the joint trajectories; generating a position trajectory extension including the predicted time domain to provide a future state reference for predictive control;

[0182] Generate joint velocity reference trajectories based on joint position reference trajectories, including: calculating the instantaneous velocity and average velocity distribution of each joint based on the time-domain differentiation of the joint position reference trajectory; correcting and optimizing the velocity obtained by differentiation in combination with the expected velocity curve in the optimization result; applying velocity constraints to ensure that the generated velocity trajectory does not exceed the maximum velocity limit of each joint; performing continuity and differentiability tests on the velocity trajectory to eliminate velocity jumps and discontinuities; adjusting the velocity trajectory for energy efficiency based on energy consumption optimization requirements to minimize the velocity change rate while meeting task requirements; and establishing a consistency verification mechanism for the velocity trajectory and position trajectory to ensure the mathematical matching of the two.

[0183] Generate joint torque reference trajectories based on the robot dynamics model, joint position reference trajectories, and joint velocity reference trajectories, including: calculating the driving torque required for each joint based on the robot dynamics model and the generated position and velocity reference trajectories; adjusting and optimizing the calculated torque based on the torque distribution information in the optimization results; generating a complete torque trajectory including compensation terms by considering gravity compensation, friction compensation, and external disturbance compensation; applying torque constraints to ensure that the generated torque trajectory is within the output capacity of each joint; adjusting the torque trajectory to suit different control mode requirements based on the joint stiffness and damping parameters in the optimization results; establishing a safety monitoring mechanism for the torque trajectory and setting torque anomaly detection and protection thresholds;

[0184] Execute multi-joint coordination and trajectory synchronization, including: analyzing the mutual influence relationship between each joint trajectory, establishing the joint coordination matrix and coupling coefficient table; optimizing the trajectory coordination of joint groups with strong coupling relationships to ensure the stability of coordinated movement; establishing full-body dynamic balance constraints and making balance adjustments to the joint trajectories that affect the stability of the center of gravity; achieving coordinated cooperation between upper and lower limb movements to ensure the harmony and efficiency of the overall movement; establishing a timing synchronization mechanism for trajectory execution to ensure that each joint executes movement according to a unified time beat;

[0185] The joint position reference trajectory, joint velocity reference trajectory and joint torque reference trajectory are integrated and distributed, including: formatting and encapsulating the generated position, velocity and torque reference trajectories to form a standardized control instruction package; preparing a dedicated trajectory data package for the second controller (local controller) of each joint, which contains the complete reference information of the joint; establishing a real-time transmission channel for trajectory data to ensure that the reference trajectory can be transmitted to each second controller (local controller) in a timely and accurate manner; establishing a feedback receiving mechanism for trajectory execution status to provide a data basis for trajectory adjustment and update; generating a trajectory execution monitoring report to record the trajectory generation process and key parameter information; and outputting the complete reference trajectory information to the second controller (local controller) for distributed execution.

[0186] The solution of this embodiment ensures the complete mathematical consistency of the three types of reference trajectories of position, velocity, and torque through a systematic trajectory generation process and multiple verification mechanisms, eliminating the control error caused by trajectory mismatch in traditional methods and improving trajectory tracking accuracy; through the establishment of joint coordination matrix and coupling coefficient, it achieves precise synchronization and coordination of multi-joint motion, avoids mutual interference and conflict between joints, and improves the coordination of overall motion; combined with the energy consumption optimal parameters of quantum heuristic optimization results, it fully considers energy efficiency requirements in the trajectory generation process, and can reduce energy consumption and extend the robot's working life compared with traditional trajectory generation methods; through trajectory extension in the prediction time domain and standardized data encapsulation, it provides an efficient data transmission and processing basis for distributed control systems, reduces system response delay, and meets the strict requirements of real-time control. This central controller reference trajectory generation method provides a high-quality control benchmark for distributed predictive control systems and is a key technical link for achieving accurate, stable, and efficient humanoid robot motion control.

[0187] In some possible implementations of the present invention, the step of locally adjusting the reference trajectory within the prediction time domain based on the real-time feedback data by the second controller configured locally at each joint includes:

[0188] The current joint state data of the joint is collected in real time through the built-in position encoder, speed sensor and torque sensor of the joint;

[0189] The current joint state data is time-synchronized and aligned with the reference trajectory, and the trajectory deviation between the current joint state and the reference trajectory is calculated (the deviation includes position deviation, velocity deviation, and torque deviation; a deviation history database is established to track the deviation change trend and periodic characteristics; based on the deviation size and change rate, the stability and execution quality of the current control state are evaluated);

[0190] Based on the current joint state data, a prediction time domain is established and state prediction is performed, including: establishing a prediction time domain length suitable for the joint based on the current joint state data and historical motion data; using the local dynamic model of the joint to deduce the possible development trajectory of the future state within the prediction time domain; considering the coupling effect of adjacent joint motion on the current joint, and correcting the accuracy of the predicted state; analyzing possible disturbance factors in the prediction time domain, including load changes, friction changes and external interference; establishing multiple prediction scenarios corresponding to different disturbance levels and environmental changes; generating a state confidence interval within the prediction time domain and quantifying the uncertainty of the prediction results;

[0191] Analyze trajectory deviations and formulate corresponding current adjustment strategies, including: classifying and identifying detected trajectory deviations to distinguish between systematic deviations, random deviations, and periodic deviations; analyzing the root causes of the deviations to determine whether they are caused by model errors, parameter drift, or external disturbances; selecting the appropriate adjustment strategy based on the type and severity of the deviation: fine-tuning, replanning, or emergency correction; evaluating the potential impact of adjustment operations on overall motion performance to ensure that local adjustments do not disrupt global coordination; establishing a priority ranking for adjustment strategies and making reasonable choices when multiple adjustment requirements coexist; setting trigger thresholds and execution conditions for adjustment operations to avoid frequent and unnecessary trajectory modifications;

[0192] Based on the current state data and the current adjustment strategy, the corresponding local trajectory in the parameter trajectory is replanned and optimized, including: generating a new local trajectory segment within the prediction time domain based on the current state data, the current adjustment strategy and the target reference point; using a fast trajectory planning algorithm to ensure that the replanned trajectory meets the physical constraints of the joint; smoothing the replanned trajectory to ensure that there are no sudden changes at the connection with the original reference trajectory; considering the balance between energy consumption and execution time, multi-objective optimization of the local trajectory; verifying the executability and safety of the adjusted trajectory to ensure that it does not cause joint overload or abnormality; establishing a rollback mechanism for trajectory adjustment to quickly restore to the original trajectory when the adjustment effect is not good;

[0193] Perform coordination and conflict resolution of adjacent joints, including: detecting the potential impact of the current joint's trajectory adjustment on the adjacent joint's motion; communicating with the secondary controller (local controller) of the adjacent joint to exchange trajectory adjustment information and coordination requirements; establishing a negotiation mechanism between joints to perform priority arbitration when there is a conflict in trajectory adjustment; performing clustered trajectory coordination adjustment for motion patterns involving multi-joint coordination; real-time monitoring of the coupling effect between joints to promptly detect and address coordination problems; and establishing an emergency response mechanism for coordination failures to ensure the system can still operate safely when coordination is difficult.

[0194] Evaluate the adjustment effect and provide feedback, including: real-time monitoring of the execution effect after trajectory adjustment, evaluating the degree of deviation reduction and control performance improvement; establishing quantitative evaluation indicators for the adjustment effect, including tracking accuracy, energy consumption changes, and execution time; feeding back the adjustment process and effect information to the central controller to provide data support for subsequent global optimization; updating the adaptive parameters of the second controller (local controller) and improving subsequent control strategies based on the adjustment experience; establishing a learning mechanism for adjustment history to accumulate typical adjustment cases for rapid response to similar situations; generating log records of adjustment operations to provide detailed information for system maintenance and fault diagnosis;

[0195] Continuous optimization and adaptive adjustment, including: continuously optimizing the accuracy of the prediction model and the length of the prediction time domain based on continuous feedback data; adjusting the detection threshold and response sensitivity of trajectory deviation based on long-term operation data; learning and memorizing frequently occurring adjustment patterns and establishing preset adjustment plans for rapid response; regularly evaluating and updating the dynamic model parameters of the joints to adapt to equipment aging and performance changes; optimizing the overall control strategy of the local controller based on the success rate and quality evaluation of task execution; and outputting the optimization results to the edge computing architecture for distributed processing and execution.

[0196] The solution of this embodiment, through the combination of real-time feedback and predictive control, the local controller can timely detect and correct trajectory deviations, improve trajectory tracking accuracy, and significantly improve the accuracy of motion control; the local adjustment mechanism can effectively deal with uncertainties such as external disturbances, model errors and parameter changes, so that the system can still maintain stable control performance in the face of various interferences and improve anti-interference ability; through the forward-looking control and fast trajectory replanning algorithm in the predictive time domain, the response speed is improved, meeting the strict requirements of real-time control; the multi-joint coordination mechanism ensures that local adjustments will not destroy the coordination of the overall movement, and through information exchange and negotiation between joints, the smoothness and naturalness of the movement are maintained, improving the overall movement quality; through intelligent adjustment strategy selection and multi-objective optimization, while ensuring control performance, the additional energy consumption is minimized, reducing energy consumption. This distributed local controller trajectory local adjustment method realizes the organic combination of global optimization and local precise control, and is a key technical means for building a high-performance humanoid robot motion control system.

[0197] In some possible implementations of the present invention, the sensor fusion adaptive weight sub-model of the first neural network model calculates the dynamic weight of each sensor data in the following manner:

[0198] in,

[0199] w i1 (t) represents the weight coefficient of the i1th sensor at time point t;

[0200] R i1 (t) represents the reliability score of the i1th sensor data, which is calculated by the following formula:

[0201]

[0202] in is the sensor data variance, D i1 (t) is the data packet loss rate, Δ i1 (t) is the consistency deviation of historical data, β1, β2 and β3 are balance coefficients and satisfy β1+β2+β3=1, λ i1 and ξ i1 is the characteristic parameter of the sensor;

[0203] C i1 (t) represents the relevance score between the i1th sensor and the current task, which is generated by the task context encoder to increase the adaptability of environmental conditions;

[0204] T i1 (t) represents the timeliness factor of the i1-th sensor, which is defined as:

[0205] T i1 (t) = exp(-γ i1 (tt last，i1 ))·(1-v i1 exp(-κ i1 f i1 (t)));

[0206] where t last，i1 is the timestamp of the most recent valid data of the i1th sensor, γ i1 is the time decay parameter, f i1 (t) is the sensor data sampling frequency, ν i1 and κ i1 is the adjustment parameter;

[0207] Q i (t) represents the newly added sensor data quality assessment factor, which is defined as:

[0208]

[0209] Where SNR i1 (t) is the sensor signal-to-noise ratio, SNR th，i1 is the threshold constant, ρ i1 and To adjust the parameters, this factor can dynamically adjust the weight according to the sensor signal-to-noise ratio;

[0210] H i1(t) represents the newly added sensor historical performance reputation, which is updated by exponential sliding average:

[0211] H i1 (t) = α H H i1 (t-1)+(1-α H )P i1 (t);

[0212] Among them, P i1 (t) is the current performance score, α H is the smoothing coefficient;

[0213] N is the total number of multimodal sensors.

[0214] The weight calculation method in this embodiment realizes the adaptive fusion of information from various sensors under different scenarios and tasks, thereby improving the robustness and adaptability of the system in noisy environments, partial sensor failures, or data loss. The sensor data quality assessment and historical performance credibility factor realize a more comprehensive sensor reliability assessment and achieve more accurate weight allocation through multi-dimensional measurement, which can improve the accuracy of fused data in complex and changing environments, while also improving the robustness of the system in the event of partial sensor failure.

[0215] In some possible implementations of the present invention, the first algorithm includes:

[0216] S11: Construct a quantum representation of the energy-stability multi-objective optimization, evaluated by the following composite objective function:

[0217] F(θ)=α E E(θ)+α S (1-S(θ))+α T (1-T(θ))+α F ·(1-F L (θ))+α A ·A D (θ);

[0218] Where θ represents the robot joint control parameter vector;

[0219] E(θ) represents the energy consumption function, which is improved to be defined as:

[0220]

[0221] Among them, P i2 (t, θ) is the power consumption of the i2th joint at time t, a i2 (t, θ) is the joint acceleration, ζ i2 is the acceleration energy consumption coefficient, M is the total number of joints, tf is the task completion time; the additional impact of acceleration on energy consumption is taken into account;

[0222] S(θ) represents the motion stability function, which is defined as:

[0223]

[0224] Among them ZMP dist (t, θ) is the minimum distance from the zero moment point to the boundary of the supporting polygon, A CoM (t, θ) is the acceleration of the center of mass, J ang (t, θ) is the rate of change of angular momentum, and μ1, μ2, and μ3 are weight coefficients. The addition of the rate of change of angular momentum can more comprehensively evaluate dynamic stability.

[0225] T(θ) represents the task completion function, which quantifies the degree of achievement of the task goal;

[0226] F L (θ) represents the load adaptability function, which is defined as:

[0227]

[0228] Among them F ext (t, θ) is the external load force, F ext (t, θ) is the estimated load force, T comp (t,θ) is the load compensation torque, ψ1 and ψ2 are weight coefficients; this function evaluates the robot's ability to adapt to changes in external load;

[0229] A D (θ) represents the action diversity index, which is defined as:

[0230]

[0231] Among them D KL K L Divergence, p θ is the current action distribution, p ref,k is the reference action template distribution, w k is the weight coefficient, K is the number of action templates; this indicator promotes the exploration of solution space diversity;

[0232] α E , α S , α T , α F and α A is the adaptive weight coefficient, satisfying α E +α S +α T +α F +α A =1;

[0233] S12: Apply qubit probability amplitude coding to represent each candidate solution in the solution space and update the solution space by the following quantum annealing probability function:

[0234]

[0235] Among them, τ(t)=τ0·exp(-η·t / t max )·(1+δ·sin(π·t / t p )) represents the quantum annealing temperature function, with the addition of a periodic fluctuation term, τ0 being the initial temperature parameter, η being the annealing rate, and t max is the maximum number of iterations, δ is the fluctuation amplitude coefficient, t p is the fluctuation period;

[0236] is the distance adjustment factor of the solution space, h(x j2 , x k ) is the Hamming distance, χ and is the adjustment parameter;

[0237] The quantum search network is:

[0238]

[0239] in, is the unitary transformation matrix of the i3th layer, representing the evolution of the quantum system;

[0240] is the perturbation operator, parameter ε i3 Controlling quantum exploration – exploiting balance;

[0241] is the newly added quantum measurement operator, σ i3 To measure the intensity parameter, quantum collapse is implemented to prevent the solution space from diverging excessively; L is the number of quantum network layers;

[0242] In this embodiment, the quantum-inspired optimization algorithm can quickly find approximate optimal solutions in complex multi-objective optimization problems, improving computational efficiency compared to traditional optimization methods. It also adds load adaptability and action diversity evaluation, while optimizing the quantum annealing strategy and search network structure, giving the algorithm better local optimal escape capability and global search efficiency, thereby improving computational efficiency, convergence speed, and the quality of optimization results in complex multi-objective optimization problems.

[0243] In some possible implementations of the present invention, the steps are further included:

[0244] S21: The central controller generates feedforward control signals based on the robot dynamics model and optimized trajectory using the following state-space prediction equation:

[0245]

[0246] in:

[0247] represents the system state vector at time t+k1 predicted at time t; X(t) represents the actual system state vector at time t, which contains position, velocity, and acceleration information; U(t+j3|t) represents the control input at time t+j3 planned at time t; Φ(t) is the time-varying state transfer matrix, which describes the system dynamic characteristics and is updated as follows: Φ(t) = Φ0 + ΔΦ(t); where Φ0 is the nominal model and ΔΦ(t) is the model deviation identified online; Γ(t) is the time-varying control input matrix, which takes into account the changes in actuator efficiency:

[0248] Γ(t) = Γ0·E(t); where Γ0 is the nominal control input matrix and E(t) is the actuator efficiency diagonal matrix;

[0249] Ψ(t+k1|t) is the environmental disturbance prediction term, which is estimated by a high-order adaptive Kalman filter:

[0250]

[0251] Among them, Ψi4 is the i4th type of disturbance model, Ψi4j3 is the disturbance interaction term, ω i4 (t) is the first-order adaptive weight, ωi4j3(t) is the second-order interaction weight, N d is the size of the perturbation model library;

[0252] Ω(t+k1|t) represents the contact dynamics prediction term, which is defined as:

[0253]

[0254] where N c is the number of possible contact points, p c (t+k1|t) is the contact probability prediction, J c (t+k1|t) is the contact Jacobian matrix, F c (t+k1|t) is the contact force prediction;

[0255] Λ(t, k1) represents the model uncertainty compensation term, which is defined as:

[0256] Λ(t,k1)=B(t)·tanh(W(t)·Z(t,k1));

[0257] Where B(t) and W(t) are adaptive parameter matrices, Z(t, k) is the eigenvector, and tanh is the hyperbolic tangent function;

[0258] S22: The second controller (local controller) of each joint adopts the sliding time domain model predictive control strategy to achieve accurate tracking by solving the following constrained optimization problem:

[0259]

[0260] Satisfy the constraints:

[0261] Umin(t)≤U(t+j4|t)≤Umax(t), j4=0, 1,..., H c -1;

[0262] ΔUmin(t)≤ΔU(t+j4|t)≤ΔUmax(t), j4=0, 1,..., H c -1;

[0263]

[0264] in:

[0265] Y represents the system output at time t+j4 predicted at time t; ref (t+j4) represents the reference trajectory; ΔU(t+j4|t) represents the control increment; H p Indicates the length of the prediction time domain; H c Indicates the control time domain length, satisfying H c ≤H p ; Q(t, j4) and R(t, j4) are the time-varying output tracking error and the control increment weight matrix respectively; Y target is the terminal target state, S is the terminal weight matrix, and the control stability is enhanced; Γ smooth is a smooth penalty function to promote smooth transition of control signals; C safety It is a safety constraint function that ensures the robot avoids dangerous states;

[0266] S23: The edge computing architecture dynamically schedules computing loads through an adaptive resource allocation algorithm, defined as:

[0267]

[0268] Satisfy the constraints:

[0269]

[0270]

[0271] R reliability(A(t))≥R min ;

[0272] in:

[0273] A(t) represents the task allocation matrix, element A i5j5 (t) represents the indicator variable that the i5th control task is assigned to the j5th edge computing node; C i5j5 (t) represents the computational cost of the i5th task at the j5th node; L i5j5 (t) represents the communication delay cost; E i5j5 (t) represents the additional energy consumption cost;

[0274] I i5j5 (t) represents the newly added information security risk indicator; D i5 (t) represents the computing resource requirement of the i5th task; K j5 (t) represents the available computing resources of the j5th node; P i5 (t) represents the power requirement of the i5th task; P max，j5 (t) represents the maximum power limit of the j5th node; R reliability (A(t)) represents the overall reliability function of the system; R min Represents the minimum reliability requirement; M1 is the total number of control tasks; N e is the number of edge computing nodes; w1(t), w2(t), w3(t) and w4(t) are time-varying weight coefficients, satisfying w1(t)+w2(t)+w3(t)+w4(t)=1.

[0275] In this embodiment, the distributed predictive control system realizes the organic combination of global planning and local fine control through a central-local two-layer control architecture. The edge computing architecture reduces system latency and improves the response speed of the humanoid robot in a highly dynamic environment. By introducing time-varying system models, contact dynamics prediction, model uncertainty compensation, terminal constraints, and a more comprehensive edge computing resource allocation strategy, the robustness and adaptability of the control system are significantly improved, the prediction accuracy is improved in a dynamic and changeable environment, the system response latency is reduced, the control stability is improved, and the energy efficiency is improved.

[0276] See Figure 2 Another embodiment of the present invention provides a motion control system for an intelligent robot, which is used to execute a motion control method for an intelligent robot, including: a humanoid robot; a central controller; an intelligent sensor group and a second controller provided on the humanoid robot; and a server;

[0277] The intelligent sensor group is configured to: collect environmental data and robot status data;

[0278] The server is configured to:

[0279] Inputting the environmental data and the robot state data into a pre-trained first neural network model to obtain robot motion prediction data and environmental analysis results;

[0280] Constructing a motion planning model based on the motion prediction data and the environmental analysis results, the model comprising:

[0281] Using a preset first algorithm to perform energy consumption-stability multi-objective optimization on the joint motion trajectory in the motion planning model;

[0282] The central controller is configured to: generate reference trajectories of joint positions, velocities, and torques based on the optimization results;

[0283] The second controller configured locally at each joint is configured to: locally adjust the reference trajectory within the prediction time domain based on real-time feedback data.

[0284] It should be known that Figure 2 The block diagram of the motion control system for an intelligent robot is for illustration only, and the number of modules shown does not limit the scope of protection of the present invention. The motion control system for an intelligent robot provided in this embodiment can be used to implement the various embodiments of the corresponding motion control method for an intelligent robot. For specific implementation processes, please refer to the description of the various method embodiments, which will not be repeated here.

[0285] It should be noted that for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that this application is not limited by the order of the actions described, because according to this application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by this application.

[0286] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0287] In the several embodiments provided in this application, it should be understood that the disclosed devices can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the above-mentioned units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection of devices or units can be electrical or other forms.

[0288] The units described above 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 these units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0289] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0290] If the above-mentioned integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a memory and includes a number of instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the above-mentioned methods of each embodiment of the present application. The aforementioned memory includes: various media that can store program codes, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk.

[0291] Those skilled in the art will appreciate that all or part of the steps in the various methods of the above embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable memory, which may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0292] The above is a detailed introduction to the embodiments of the present application. Specific examples are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and core idea of the present application. At the same time, for those skilled in the art, according to the idea of the present application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

[0293] Although the present invention is disclosed above, it is not limited thereto. Any person skilled in the art may readily conceive of variations or substitutions, and may make various modifications and alterations without departing from the spirit and scope of the present invention. Combinations of the above-described functions and implementation steps, including software and hardware implementations, are all within the scope of protection of the present invention.

Claims

1. A motion control method for an intelligent robot, characterized in that: include: Collect environmental data and robot status data through an intelligent sensor group installed on the humanoid robot; Inputting the environmental data and the robot state data into a pre-trained first neural network model to obtain robot motion prediction data and environmental analysis results; Constructing a motion planning model based on the motion prediction data and the environmental analysis result; Using a preset first algorithm to perform energy consumption-stability multi-objective optimization on the joint motion trajectory in the motion planning model; The central controller generates reference trajectories of joint positions, velocities, and torques based on the optimization results; The second controller configured locally at each joint makes local adjustments to the reference trajectory within the prediction time domain based on real-time feedback data.

2. The motion control method for an intelligent robot according to claim 1, characterized in that: The step of inputting the environmental data and the robot state data into a pre-trained first neural network model to obtain the robot's motion prediction data and environmental analysis results includes: Preprocessing the environmental data and the robot status data; Inputting the preprocessed environment data and the robot state data into the spatiotemporal attention sub-model of the first neural network model; The spatiotemporal attention sub-model establishes an attention weight matrix for the time dimension; The temporal attention mechanism is used to calculate the correlation between the current moment and the historical moment data, and identify the temporal dependencies of the motion patterns. Analyze the spatial correlation between different sensor data based on the spatial attention mechanism and establish a mutual influence relationship diagram between sensors; Fuse the temporal and spatial attention weights to generate a comprehensive spatiotemporal attention distribution matrix; According to the spatiotemporal attention distribution matrix, the sensor fusion adaptive weight sub-model calculates the fusion weight coefficient of each sensor data; Dynamically adjust the importance weights of different sensor data based on the current environment complexity and task requirements; Perform weighted fusion on the weight-adjusted multimodal data to form a unified feature representation vector; Control the information flow through the gating mechanism to filter out the sensor information that is most critical to the current movement decision; The fused feature representation vector is input into the motion prediction sub-model to extract the robot's current motion trend and goal orientation; The feature representation vector is simultaneously input into the environment analysis sub-model to analyze the geometric structure, obstacle distribution and dynamic changes of the surrounding environment; Combine motion intention and environmental understanding results to generate motion prediction data and environmental analysis results.

3. The motion control method for an intelligent robot according to claim 2, characterized in that: The step of constructing a motion planning model based on the motion prediction data and the environmental analysis result includes: Constructing the first sub-model includes: establishing a global task decomposition tree based on the motion prediction data and the environmental analysis results, combined with preset task target parameters; generating a global path from the starting point to the target point based on the obstacle distribution information and terrain characteristics in the environmental analysis results; dividing the global path into multiple sub-path segments based on task complexity and environmental dynamics, each sub-path segment corresponding to a behavior primitive; assigning execution priority and time window to each behavior primitive, generating a time-series execution sequence including basic behaviors such as walking, turning, grasping, and obstacle avoidance; and establishing a library of alternative solutions for the global path, enabling rapid switching to alternative paths when the main path is blocked. Constructing a second sub-model includes: receiving the behavior sequence output by the first sub-model, calling the corresponding motion pattern library for each behavior primitive; decomposing the behavior primitive into a set of motion trajectories for each joint based on the robot's kinematic and dynamic constraints; calculating the position, velocity, and acceleration target values of each joint using an inverse kinematics algorithm, taking into account the coupling relationship and coordination constraints between joints; performing feasibility testing and collision detection on the joint trajectories based on the environmental constraint information in the environmental analysis results; locally adjusting and optimizing trajectories that do not meet the constraint conditions to ensure the smoothness and continuity of the trajectories; generating a detailed joint motion trajectory table including timestamps, and establishing a checkpoint mechanism for trajectory execution; The third sub-model is constructed, including: establishing a multi-threaded monitoring mechanism to detect environmental changes and emergencies in real time, including the appearance of dynamic obstacles, changes in ground conditions, and external interference; when an emergency is detected, immediately launching an emergency response program, pausing the current trajectory execution and entering the reaction control mode; based on the current sensor data and robot status, quickly assessing the threat level and impact range of the emergency; selecting the corresponding response strategy according to the threat level: fine-tuning the trajectory for minor interference, initiating avoidance actions for moderate threats, and executing an emergency stop for severe threats; in the reaction control mode, using a rule-based rapid decision-making algorithm to generate local avoidance trajectories or protective actions; evaluating the execution effect of the reaction action in real time, and when the threat is eliminated, replanning the connection trajectory from the current state to the original target; Construct coordination and information transmission mechanisms among sub-models; Planning results output and verification.

4. The motion control method for an intelligent robot according to claim 3, characterized in that: The method of using a preset first algorithm to perform energy consumption-stability multi-objective optimization on the joint motion trajectory in the motion planning model includes: Construct a composite objective function including energy consumption, motion stability and task completion; Utilize quantum annealing to rapidly search the solution space to achieve optimal motion control parameters in complex terrain and dynamic environments; The joint stiffness and damping parameters are dynamically adjusted according to the optimization results, so that the robot can minimize energy consumption while maintaining stability.

5. The motion control method for an intelligent robot according to claim 4, characterized in that: The central controller generates reference trajectories of joint positions, velocities, and torques based on the optimization results, including: Perform optimization result analysis and parameter extraction; Generate joint position reference trajectory based on the data obtained from optimization result analysis and parameter extraction; Generate joint velocity reference trajectory based on joint position reference trajectory; Generate joint torque reference trajectory based on robot dynamics model, joint position reference trajectory and joint velocity reference trajectory; Perform multi-joint coordination and trajectory synchronization; The joint position reference trajectory, joint velocity reference trajectory and joint torque reference trajectory are integrated and distributed.

6. The motion control method for an intelligent robot according to claim 5, characterized in that: The second controller locally configured at each joint locally adjusts the reference trajectory in the prediction time domain according to the real-time feedback data, including: The current joint state data of the joint is collected in real time through the built-in position encoder, speed sensor and torque sensor of the joint; Align the current joint state data with the reference trajectory in time synchronization, and calculate the trajectory deviation between the current joint state and the reference trajectory; Based on the current joint state data, establish the prediction time domain and perform state prediction; Analyze the trajectory deviation and formulate the corresponding current adjustment strategy; Replan and optimize the corresponding local trajectory in the parameter trajectory based on the current state data and the current adjustment strategy; Performs adjacent joint coordination and conflict resolution; Evaluate the effects of adjustments and provide feedback.

7. The motion control method for an intelligent robot according to claim 6, characterized in that: The sensor fusion adaptive weight sub-model of the first neural network model calculates the dynamic weight of each sensor data in the following way: in, W i1 (t) represents the weight coefficient of the i1th sensor at time point t; R i1 (t) represents the reliability score of the i1th sensor data, which is calculated by the following formula: in is the sensor data variance, D i1 (t) is the data packet loss rate, Δ i1 (t) is the consistency deviation of historical data, β1, β2 and β3 are balance coefficients and satisfy β1+β2+β3=1, λ i1 and ξ i1 is the characteristic parameter of the sensor; C i1 (t) represents the relevance score between the i1th sensor and the current task, which is generated by the task context encoder to increase the adaptability of environmental conditions; T i1 (t) represents the timeliness factor of the i1-th sensor, which is defined as: T i1 (t)=exp(-γ i1 (t-t last，i1 ))·(1-v i1 exp(-κ i1 f i1 (t))); where t last,i1 is the timestamp of the most recent valid data of the i1th sensor, γ i1 is the time decay parameter, f i1 (t) is the sensor data sampling frequency, v i1 and κ i1 To adjust the parameters; Q i (t) represents the newly added sensor data quality assessment factor, which is defined as: Where SNR i1 (t) is the sensor signal-to-noise ratio, SNR th，i1 is the threshold constant, ρ i1 and To adjust the parameters, this factor can dynamically adjust the weight according to the sensor signal-to-noise ratio; H i1 (t) represents the newly added sensor historical performance reputation, which is updated by exponential sliding average: H i1 (t)=α H H i1 (t-1)+(1-α H )P i1 (t); Among them, P i1 (t) is the current performance score, α H is the smoothing coefficient; N is the total number of multimodal sensors.

8. The motion control method for an intelligent robot according to claim 7, characterized in that: The first algorithm includes: S11: Construct a quantum representation of the energy-stability multi-objective optimization, evaluated by the following composite objective function: F(θ)=α E ·E(θ)+α S ·(1-S(θ))+α T ·(1-T(θ))+α F ·(1-F L (i))+a A ·A D (i); Where θ represents the robot joint control parameter vector; E(θ) represents the energy consumption function, which is improved to be defined as: Among them, P i2 (t, θ) is the power consumption of the i2th joint at time t, a i2 (t, θ) is the joint acceleration, ζ i2 is the acceleration energy consumption coefficient, M is the total number of joints, t f is the task completion time; the additional impact of acceleration on energy consumption is taken into account; S(θ) represents the motion stability function, which is defined as: Among them ZMP dist (t, θ) is the minimum distance from the zero moment point to the boundary of the supporting polygon, A CoM (t, θ) is the acceleration of the center of mass, J ang (t, θ) is the rate of change of angular momentum, and μ1, μ2, and μ3 are weight coefficients. The addition of the rate of change of angular momentum can more comprehensively evaluate dynamic stability. T(θ) represents the task completion function, which quantifies the degree of achievement of the task goal; F L (θ) represents the load adaptability function, which is defined as: Among them F ext (t, θ) is the external load force, F est (t, θ) is the estimated load force, T comp (t, θ) is the load compensation torque, ψ1 and ψ2 are weight coefficients; this function evaluates the robot's ability to adapt to changes in external load; A D (θ) represents the action diversity index, which is defined as: Among them D KL is the KL divergence, p θ is the current action distribution, p ref，k is the reference action template distribution, w k is the weight coefficient, K is the number of action templates; this indicator promotes the exploration of solution space diversity; α E , α S , α T , α F and α A is the adaptive weight coefficient, satisfying α E +α S +α T +α F +α A =1; S12: Apply qubit probability amplitude coding to represent each candidate solution in the solution space and update the solution space by the following quantum annealing probability function: Among them, τ(t)=τ0·exp(-η·t / t max )·(1+δ·sin(π·t / t p )) represents the quantum annealing temperature function, with the addition of a periodic fluctuation term, τ0 being the initial temperature parameter, η being the annealing rate, and t max is the maximum number of iterations, δ is the fluctuation amplitude coefficient, t p is the fluctuation period; is the distance adjustment factor of the solution space, h(x j2 , x k ) is the Hamming distance, χ and is the adjustment parameter; The quantum search network is: in, is the unitary transformation matrix of the i3th layer, representing the evolution of the quantum system; is the perturbation operator, parameter ε i3 Controlling quantum exploration – exploiting balance; is the newly added quantum measurement operator, σ i3 To measure the intensity parameters, quantum collapse is implemented to prevent excessive divergence of the solution space; L is the number of quantum network layers.

9. The motion control method for an intelligent robot according to claim 8, characterized in that: Also includes the steps: S21: The central controller generates feedforward control signals based on the robot dynamics model and optimized trajectory using the following state-space prediction equation: in: represents the system state vector at time t+k1 predicted at time t; X(t) represents the actual system state vector at time t, which contains position, velocity and acceleration information; U(t+j3|t) represents the control input at time t+j3 planned at time t; Φ(t) is the time-varying state transfer matrix, which describes the dynamic characteristics of the system and is updated as follows: Φ(t) = Φ0 + ΔΦ(t); where Φ0 is the nominal model and ΔΦ(t) is the model deviation identified online; Γ(t) is the time-varying control input matrix, taking into account the change in actuator efficiency: Γ(t) = Γ0·E(t); where Γ0 is the nominal control input matrix and E(t) is the actuator efficiency diagonal matrix; Ψ(t+k1|t) is the environmental disturbance prediction term, which is estimated by a high-order adaptive Kalman filter: Among them, Ψi4 is the i4th type of disturbance model, Ψi4j3 is the disturbance interaction term, ω i4 (t) is the first-order adaptive weight, ωi4j3(t) is the second-order interaction weight, N d is the size of the perturbation model library; Ω(t+k1|t) represents the contact dynamics prediction term, which is defined as: where N c is the number of possible contact points, p c (t+k1|t) is the contact probability prediction, J c (t+k1|t) is the contact Jacobian matrix, F c (t+k1|t) is the contact force prediction; Λ(t, k1) represents the model uncertainty compensation term, which is defined as: Λ(t,k1)=B(t)·tanh(W(t)·Z(t,k1)); Where B(t) and W(t) are adaptive parameter matrices, Z(t, k) is the eigenvector, and tanh is the hyperbolic tangent function; S22: The second controller (local controller) of each joint adopts the sliding time domain model predictive control strategy to achieve accurate tracking by solving the following constrained optimization problem: Satisfy the constraints: Umin(t)≤U(t+j4|t)≤Umax(t),j4=0,1,...,H c -1; ΔUmin(t)≤ΔU(t+j4|t)≤ΔUmax(t),j4=0,1,...,H c -1; in: Y represents the system output at time t+j4 predicted at time t; ref (t+j4) represents the reference trajectory; ΔU(t+j4|t) represents the control increment; H p Indicates the length of the prediction time domain; H c Indicates the control time domain length, satisfying H c ≤H p ; Q(t, j4) and R(t, j4) are the time-varying output tracking error and control increment weight matrix respectively; Y target is the terminal target state, S is the terminal weight matrix, and the control stability is enhanced; Γ smooth is a smooth penalty function to promote smooth transition of control signals; C safety It is a safety constraint function that ensures the robot avoids dangerous states; S23: The edge computing architecture dynamically schedules computing loads through an adaptive resource allocation algorithm, defined as: Satisfy the constraints: R reliability (A(t))≥R min ; in: A(t) represents the task allocation matrix, element A i5j5 (t) represents the indicator variable that the i5th control task is assigned to the j5th edge computing node; C i5j5 (t) represents the computational cost of the i5th task at the j5th node; L i5j5 (t) represents the communication delay cost; E i5j5 (t) represents the additional energy consumption cost; I i5j5 (t) represents the newly added information security risk indicator; D i5 (t) represents the computing resource requirement of the i5th task; K j5 (t) represents the available computing resources of the j5th node; P i5 (t) represents the power requirement of the i5th task; P max，j5 (t) represents the maximum power limit of the j5th node; R reliability (A(t)) represents the overall reliability function of the system; R min Represents the minimum reliability requirement; M1 is the total number of control tasks; N e is the number of edge computing nodes; w1(t), w2(t), w3(t) and w4(t) are time-varying weight coefficients, satisfying w1(t)+w2(t)+w3(t)+w4(t)=1.

10. A motion control system for an intelligent robot, used to execute the motion control method for an intelligent robot according to any one of claims 1 to 9, characterized in that: include: humanoid robots; A central controller; an intelligent sensor group and a second controller provided on the humanoid robot; and servers; The intelligent sensor group is configured to: collect environmental data and robot status data; The server is configured to: Inputting the environmental data and the robot state data into a pre-trained first neural network model to obtain robot motion prediction data and environmental analysis results; Constructing a motion planning model based on the motion prediction data and the environmental analysis results, the model comprising: Using a preset first algorithm to perform energy consumption-stability multi-objective optimization on the joint motion trajectory in the motion planning model; The central controller is configured to: generate reference trajectories of joint positions, velocities, and torques based on the optimization results; The second controller configured locally at each joint is configured to: locally adjust the reference trajectory within the prediction time domain based on real-time feedback data.

Images