297 results about "Robot dynamic" patented technology

The invention discloses a dynamic path planning and self-adaptive control method and system for a coating robot, and relates to the technical field of coating automation. The method comprises the following steps: acquiring point cloud data through three-dimensional scanning equipment, constructing a dynamically updated workpiece curved surface model, and extracting curvature, edge and high-curvature mutation region features; a spraying path is generated based on a reinforcement learning algorithm, and path density, speed and coating supply are dynamically adjusted for a high-curvature area; distance, force feedback and environment parameters are fused, and mechanical arm and spray gun parameters are dynamically adjusted; dividing operation sub-areas and distributing tasks based on robot capability characteristics to realize multi-machine cooperation; and fault redundancy control and multispectral imaging are added to optimize the coating quality. The system comprises a sensing module, a decision-making module, an execution module, a redundancy control module and a communication module. According to the invention, the uniformity of the complex curved surface coating, the multi-machine cooperation efficiency and the system anti-interference capability are improved, and the method is suitable for spraying large workpieces such as aircraft fuselages and high-speed rail vehicle bodies.

The invention relates to a robot dynamic grabbing control method based on visual sense and tactile sense depth fusion. The method comprises the following steps that firstly, based on visual information, surface geometric features of a target object are extracted, grabbing adaptability scores are calculated, an optimal grabbing point is selected, and a collision-free approaching track is planned; secondly, in a contact establishment stage, detecting a contact event through a touch sensor, fusing vision and touch data to unify a coordinate system, calculating a vision-touch consistency score, and applying an initial grabbing force; and thirdly, in the stable holding stage, slip detection is conducted through wavelet packet energy entropy and pressure gradient, and the grabbing force and impedance parameters are dynamically adjusted in combination with self-adaptive impedance control. According to the method, intelligent control over the whole process from approaching to contact to stable holding is achieved, and the adaptability, stability and safety of robot grabbing in the dynamic environment are improved.

The invention relates to the technical field of service robot navigation and motion control, in particular to a service robot dynamic obstacle pre-judgment obstacle avoidance system, which comprises an environment sensing unit for outputting sensing data; the target state estimation unit outputs a target state sequence; the future evolution prediction unit is used for outputting a target future evolution result; the risk quantification unit outputs the current collision risk degree; operating the strategy management unit, and outputting strategy parameters; the evasion decision and path generation unit is used for outputting candidate evasion tracks and speed curves; the execution and control interface unit is used for outputting a chassis control instruction and implementing trajectory tracking; and the device communication and calculation platform is used for outputting a time sequence synchronization and data channel management result according to operation requirements and providing fault monitoring. According to the method, target state estimation, future evolution prediction, collision risk quantification and strategy linkage are completed within real-time constraints, consistent input and constraints are provided for planning and execution on a unified data channel, and safety and traceability are improved.

An automated system for a space includes: k navigating robots operating within the space to perform a set of orders, where k is an integer greater than or equal to two, where the space includes one or more waiting areas where any navigating robot may wait without blocking movement of another navigating robot; and a control module configured to receive the set of orders and communicate, to the k navigating robots, paths p for the k navigating robots to fulfil the set of orders, with each order in the set of orders including one or more tasks including a sequence of actions to be performed at a location in the space, having an order availability time after which any task of an order may start, and being associated with a set of precedence constraints arising from at least one of a relationship between orders and a resource of an order.

The invention belongs to the technical field of robot navigation and path planning, and particularly discloses and provides a dynamic navigation planning method and system for a service robot, which overcomes the limitation of single-dimensional analysis in a complex dynamic scene by constructing a dynamic social thermodynamic diagram and deducing group dynamic intention data. The perception precision of the environment structure and the people stream situation is improved; a linkage mechanism among a dynamic social thermodynamic diagram, individual behavior characteristics and path planning is established, and a social cost function is generated to optimize a navigation path according to the traffic efficiency and social etiquette of group future dynamic vectorization evaluation; the actual influence of the navigation behavior of the robot on the environment is monitored in real time, and a planning strategy is adjusted in combination with feedback data, so that the adaptability of the navigation decision and the continuous improvement of the environment acceptance degree are realized.

According to the mechanical arm control method and system based on the self-adaptive force field, the motion state of a patient is monitored, the motion state is compared with a demonstration space trajectory, and training performance information of the patient is obtained. Based on the training performance information and the constructed adaptive force field model, auxiliary force applied to the patient is obtained through calculation, wherein the auxiliary force comprises normal force, tangential force and viscous force. And the applied auxiliary torque is calculated based on the tail end posture of the mechanical arm and the target posture, the joint control torque is obtained according to the auxiliary force and the auxiliary torque, and the dynamic torque and the null space control torque of the robot dynamic model are obtained. And the joint control torque, the dynamic torque and the null space control torque are combined to obtain the comprehensive control torque for controlling the mechanical arm. Therefore, the self-adaptive adjustment of the auxiliary force is realized based on the position, speed and direction of the patient, and the time degree of freedom can be obtained in the motion of the self-adaptive force field, so that the patient obtains more freedom of active motion under the assistance of the space force field.

The invention discloses a robot dynamic path planning method and system fused with lightweight LLM. The robot dynamic path planning method comprises the following steps of dynamically mapping a grid map, constructing a semantic analysis model, performing global path planning by adopting an improved A * algorithm, and performing local path optimization by adopting an improved DWA algorithm. Specifically, according to the method, the high-frequency capture and real-time updating of the dynamic obstacle are realized by introducing the fusion of the event flow of the event camera and the depth information of the laser radar; dynamic matching of semantic instructions and planning parameters is achieved by introducing lightweight LLM, meanwhile, sight distance penetration rate optimization path connectivity is fused in global planning, local obstacle avoidance adopts a semantic adaptive weight adjustment mechanism, the high-precision perception advantage of multi-sensor fusion is guaranteed, task adaptability and scene pertinence are improved, and the method is suitable for large-scale popularization and application. And the complex requirements of logistics scenes are met.

The invention provides a configuration optimization method before capturing of a space double-arm robot, and the method comprises the steps: constructing a kinematics and dynamics model of the space double-arm robot according to a three-dimensional model of the space double-arm robot; constructing a kinetic model of the to-be-captured target by using the motion parameters of the to-be-captured target in the space; according to the dynamic model of the space double-arm robot, the relative mass of the space double-arm robot under non-centering collision is obtained, and a maximum pre-contact collision force performance evaluation index is constructed; further constructing a double-arm closed chain inertia matching index performance evaluation index; constructing a configuration optimization model before the space double-arm robot captures the target; and through a multi-target optimization algorithm, the optimal configuration of the space double-arm robot before capturing the target is solved. The pre-contact collision force peak value of the double-arm robot during target capturing can be effectively reduced, and meanwhile the operability of a closed chain system formed after capturing is guaranteed.

The invention discloses a robot hierarchical path planning method and device in a complex environment, a medium and a product, and the method comprises the steps: responding to a path planning request of a robot in an airport terminal, dividing the airport terminal into a plurality of dynamic Voronoi regions according to an underlying environment grid map and people flow density information, a minimum path cost algorithm with people flow density weighting is adopted to construct a global path anchor point sequence, the global path anchor point sequence is further split into a plurality of local path planning requests, a double-heuristic cost function is constructed, an improved algorithm is adopted to generate local sub-paths, and according to surrounding environment information collected by a robot in real time, the local sub-paths are calculated. And optimizing each local sub-path by adopting a reinforcement learning algorithm, generating motion control instructions for the robot at different control time points, and realizing dynamic obstacle avoidance of the robot. The path planning efficiency and the motion safety of the mobile robot in a super-large-scale dynamic environment are remarkably improved, and the technical problems that a traditional method is high in calculation complexity and poor in real-time performance are effectively solved.

The invention discloses a spiking neural network driven robot dynamic adaptive control system, which relates to the technical field of robot control, and comprises seven modules: an environment sensing module which integrates various sensors and collects and transmits environment, attitude and interaction information; the signal preprocessing module processes data through composite filtering and feature extraction; the spiking neural network modeling module constructs a three-layer structure and performs training based on a fusion learning rule; the dynamic decision output module converts the pulse signal into a control instruction and adjusts gain; the actuating mechanism driving module drives the actuator to act; the state feedback monitoring module monitors and feeds back motion parameters and system states; and the adaptive optimization module optimizes the network and module parameters based on feedback data, and dynamically matches the environment. The control precision and the response speed of the robot in a complex environment are improved, the adaptive capacity is enhanced, the operation reliability and safety are guaranteed through multi-module cooperation, and the application scene is expanded.

The invention discloses a reconfigurable robot control method and system for a zero-sum game optimization event triggering mechanism. The method comprises the steps that a reconfigurable robot dynamic model based on joint torque feedback is established, an error dynamic state space is established, and a preset performance function is set to restrain the transient performance of the error dynamic state space; the error dynamics state space constraining the transient performance is mapped to the unconstrained error dynamics augmented system state space, and an event triggering mechanism is introduced, so that the unconstrained error dynamics augmented system state space meets the zero-sum game of maximizing the event triggering interval and minimizing the control input; setting a performance index function based on a zero-sum game, and obtaining an optimal control law under an event triggering mechanism through the performance index function; according to the invention, the precision and stability of robot system control are improved through the predetermined performance function; the zero-sum game of an event triggering mechanism is introduced to obtain an optimal control law under a Nash equilibrium condition, so that communication resources and calculation burdens are effectively reduced; and the control torque of each subsystem is smoother under the action of the optimal control law, so that the energy consumption is reduced and the service life of the motor is prolonged.

The invention discloses a non-flat terrain autonomous navigation method based on path pruning and trajectory smoothing. The method comprises the following steps: constructing an elevation map with three-dimensional map information as a navigation map according to robot navigation space information; based on the navigation map, performing global path search by adopting a path pruning strategy to obtain an initial path from the starting point to the target point; based on the initial path, track optimization is carried out by using a B spline curve, and a smooth track meeting robot dynamics constraints is generated; in response to an obstacle which is detected on the smooth trajectory and cannot be avoided, performing three-dimensional obstacle crossing trajectory planning based on terrain elevation information of the elevation map, and generating an obstacle crossing trajectory capable of crossing the obstacle; wherein the robot performs autonomous navigation based on the smooth track or the obstacle crossing track. According to the invention, efficient global search, smooth trajectory optimization and intelligent obstacle crossing decision of the robot in unknown and non-flat terrains can be realized.

The invention discloses a robot dynamic voice interaction method based on multi-modal state management, and belongs to the technical field of intelligent robots, and the method comprises the steps: S1, collecting multi-modal instruction data inputted by a user; s2, carrying out data processing on the multi-mode instruction data, and then generating an input text packet with an environment context mark; s3, performing intention analysis on the input text packet, and binding the input text packet with a historical context to generate an intention vector; s4, updating the environment matrix in real time through the intention vector; s5, dynamically calculating the task weight W through the updated environment matrix, and dynamically adjusting the priority of each task; and S6, calling a corresponding function module according to the priority of each task through a plug-in interface, and executing the corresponding task, so as to achieve the purposes of realizing cross-scene context continuity, multi-task priority dynamic scheduling, supporting hot plug function extension and having active abnormal interaction capability.

The invention discloses an automatic calculation and compensation method and system for dynamic strike delay of a laser weeding robot, and belongs to the technical field of agricultural intelligent equipment.The method is achieved through the following steps that in the static calibration stage, a laser galvanometer, a camera and collaborative parameters of the laser galvanometer and the camera of a robot system are corrected; laser scanning nonlinear errors, camera imaging distortion errors and pose and mapping errors between a galvanometer coordinate system and a camera coordinate system are eliminated; in the dynamic testing stage, the robot advances at a constant speed, laser striking is conducted on the wood chips with preset marks, and the position deviation between an actual striking point and a theoretical target point is obtained through the image recognition technology; reversely deducing system delay time based on the deviation and the speed of the robot, and iteratively updating compensation parameters; finally, after the iterative optimization strike deviation is smaller than the set precision, the compensation parameters are solidified to a control system; the method is suitable for a laser weeding robot and other dynamic visual striking systems.

The invention relates to the technical field of intelligent robot control, in particular to a small robot dynamic balance control system based on DDPG, which comprises a sensor and a preprocessing module, provides accurate and reliable state data through denoising and standardization processing, and enhances the accuracy of subsequent processing; the ESO observer module estimates the system state and the total disturbance in real time, and the adaptive capacity and the anti-interference performance to the complex environment are improved; the information fusion and time sequence alignment module ensures information synchronization and accuracy through fusion processing and multi-dimensional state vector construction, meanwhile, the timeliness of a control decision is ensured, and the overall control effect is improved; the DDPG strategy network and the security layer module are combined with strategy optimization and multi-layer security constraints, continuous and safe control actions are output, and the stability and security of control are improved; and the actuator interface module realizes control instruction transmission, forms closed-loop feedback, adjusts a control strategy in real time, and improves the dynamic balance capability and the overall stability of the system.

The invention discloses a dynamic obstacle avoidance and path optimization method based on a body intelligence robot, and relates to the technical field of robot path planning, and the method comprises the following steps: collecting map data and task constraint data; performing path planning analysis on the map data and the task constraint data based on a set path optimization algorithm to generate initial planning path data; according to the intelligent robot dynamic obstacle avoidance and path optimization method, a robot actual driving path prediction model is constructed, and a target function is constructed for the deviation between the robot actual driving path prediction model and the real-time path planning data of the next time period to carry out minimization analysis; and the real-time path planning data of the next time period is optimized, so that the robot runs according to the optimized real-time path planning data to obtain an actual running path which fits the original real-time path planning data, and the running error of the robot is reduced.

The invention discloses a mine robot dynamic scene modeling system based on 3D Gaussian splashing, and belongs to the technical field of mine intelligent monitoring and 3D scene modeling crossing. Comprising a multi-source data perception and accurate alignment module, a 3D Gaussian splash modeling and optimization module, a dynamic scene intelligent identification and updating module and a semantic enhancement visualization and decision support module. The method comprises the following steps of: firstly, directly generating accurate detection and semantic level classification capability on a scene change area; the system is enabled to exceed the level of purely sensing the existence of motion, and the deep understanding of the type and behavior characteristics of a dynamic object is realized. Specifically, the system can accurately distinguish different types of changes such as static obstacles, dynamic obstacles and environmental evolution, and corresponding physical attributes and semantic tags are given to the types. Therefore, the limitation of the prior art in the aspect of environment cognition dimension is effectively overcome.

The invention provides a robot dynamic obstacle avoidance method and device, control equipment and a robot, and relates to the technical field of data processing.The method comprises the steps that original environment data of a target scene where the robot is located in a preset historical time period is acquired; according to the original environment data of the preset historical time period, obtaining historical track data of the dynamic target in the target scene in the preset historical time period; performing action intention prediction by adopting a preset intention prediction network to obtain action intention prediction information of the dynamic target in preset future time; adopting a preset game model of the robot for the target scene to generate action decision information of the robot in preset future time; and according to the action decision information, generating a target control instruction of the robot in the preset future time. According to the invention, the accuracy of the control instruction and the adaptability and interaction intelligence level of the robot in a complex dynamic scene are improved.

The invention relates to the technical field of intelligent robot control and artificial intelligence, and discloses a robot dynamic trajectory control method based on reinforcement learning and multi-sensor fusion, and the method comprises the steps: firstly fusing multi-modal sensor data through extended Kalman filtering, and predicting the time sequence distribution of dynamic obstacles through a long-short-term memory network; meanwhile, the dynamic operability and singularity risk factors of the robot are calculated, on this basis, a reinforcement learning network based on an attention mechanism is constructed to generate a high-level motion strategy, the high-level motion strategy is converted into a bottom-layer execution instruction through a variable impedance controller and a convex optimization torque distribution module, and singular points are avoided by using null-space characteristics. In addition, according to the method, model parameters are finely adjusted in real time by monitoring and predicting errors and the safety intervention degree on line. According to the method, organic fusion of logic decision and physical execution is realized through a layered architecture, and the operation safety, robustness and control precision of the robot in a dynamic environment are remarkably improved.

The invention provides a robot dynamic obstacle trajectory prediction navigation method based on a visual model, and relates to the field of power equipment maintenance, the method comprises the following steps: collecting original data of dynamic scene multi-modal perception, carrying out space-time alignment, extracting thermal infrared dynamic feature vectors and visual dynamic feature vectors, and carrying out dynamic scene multi-modal perception on the thermal infrared dynamic feature vectors and the visual dynamic feature vectors; simultaneously calculating the fuzzy degree of the visual image and the dynamic complexity of the scene; calculating a fusion weight of the thermal infrared feature and the visual feature through a dynamic modal weight algorithm, and outputting a fusion feature vector; inputting the fusion feature vector into a visual model, and calculating the trajectory probability distribution of the dynamic obstacle; calculating comprehensive uncertainty and risk coefficients based on the trajectory probability distribution; and based on the fusion feature vector and the trajectory probability distribution, integrating the uncertainty and the risk coefficient, and planning an obstacle avoidance path and a decision report in combination with an A algorithm, so as to realize the dynamic navigation of the robot. According to the technical scheme, the reliability and intelligence of robot navigation can be improved, and the navigation requirement of a complex environment is met.

The invention provides a split type robot dynamic task flow engine system. Comprising a basic actuator used for generating a task instance according to an externally input task request; the operator set comprises at least one operator, and each operator is used for calling hardware resources of the corresponding function cabin to execute actions corresponding to the operator types when receiving an execution instruction of the basic actuator; the task model is used for maintaining identification, types, priorities and state information of task instances between the basic actuator and the operator set and supporting persistent storage and recovery of the task instances; and the communication cooperation module is used for providing a unified message transmission interface among the chassis, the functional cabin and the edge computing node of the split type robot so as to realize the transmission of instruction and state data between the basic actuator and the operator set. According to the invention, plug-and-play and dynamic adaptation of the cabin body, task scheduling based on capability, unified communication and state persistence can be realized, and task continuity is guaranteed.

The invention belongs to the technical field of intelligent robot control, particularly relates to a multi-target balance reinforcement learning robot flexible control strategy optimization method, and aims to solve the problem that operation efficiency, energy consumption, safety boundary and man-machine interaction flexibility are difficult to synchronously consider in complex tasks in traditional reinforcement learning. And a fixed weight multi-target reward mechanism lacks a state sensing capability, so that the strategy adaptability and robustness are poor. The method comprises the following steps: constructing a robot dynamic environment model; designing a state-aware nonlinear multi-target reward function; introducing a security constraint guided exploration mechanism; constructing a dual-path policy network architecture; implementing an online self-adaptive tradeoff adjustment mechanism; and end-to-end strategy training and deployment are completed. According to the technical scheme, task execution efficiency is remarkably improved, energy consumption is reduced, safety and environmental adaptability are enhanced, and the method is suitable for high-flexibility control scenes such as industrial cooperation, medical assistance and service robots.

A robot sensorless external force estimation method in combination with a physical embedded kinetic model comprises the following steps: generating a robot real motion data set by collecting the angle, angular velocity and current data of each joint of a robot, and realizing robot dynamics and friction modeling based on PINN; robot inertia matrix derivative and momentum derivative online reasoning is carried out based on the trained dynamics and friction model, a joint torque error is obtained through calculation, and a joint gap error data set is obtained through separation. The joint gap model based on the Gaussian basis function is used for predicting and compensating the model uncertainty caused by the joint gap on line so as to correct the AKF external force estimation algorithm, an auxiliary observation variable is constructed, self-adaptive online filtering is implemented to observe and reduce the noise influence, and finally the robot external force estimation based on the AKF is achieved. According to the method, the problem that the system inertia matrix derivative solution is complex and the model uncertainty influences the observation precision can be effectively solved, and better robot sensorless external force estimation is realized.

The invention belongs to the technical field of robot control, and discloses a fuzzy adaptive control method and system for uncertain robot system constraints. A dead zone model is integrated into an uncertain robot kinetic equation, a robot system module is established, feedback joint displacement state information is output in real time through the robot system module, and unmeasurable speed and acceleration state information is processed through a filter module, a virtual controller module and a self-adaptive law module. And a virtual control signal and a fuzzy adaptive law parameter are obtained, and dynamic change information is obtained through a fuzzy logic system in a fuzzy adaptive controller module. According to the method, parameter uncertainty caused by unknown robot system parameters or uncertainty and dead zone characteristics caused by non-controlled object factors such as external disturbance are solved, and accurate tracking control of the system state under boundary constraint is realized.

The invention provides a robot dynamic matching and cooperation method and system based on capability discovery. The method comprises the following steps: constructing an ecological integrated collaborative interface set, acquiring physical capability parameter data and real-time operation state parameter data, generating a dynamic capability evaluation map, analyzing a task demand instruction, generating a task demand vector, performing intelligent matching, and performing analysis in combination with a preset robot digital twinborn model. Obtaining a dynamic resource scheduling strategy, controlling the robot to execute actions, obtaining execution state detection data, further processing to obtain execution state abnormity evaluation parameters, and finally determining the dynamic matching and cooperation state of the robot through threshold comparison; according to the method, closed-loop processing is formed by constructing a multi-layer interface abstract architecture, relying on a dynamic capability discovery mechanism, constructing a dynamic capability graph and a task demand vector, intelligently matching a dynamic resource scheduling strategy and evaluating execution feedback, so that robot dynamic matching and cooperation based on capability discovery are realized.

The invention relates to the field of robot path planning, in particular to a multi-robot path planning method based on group control, which comprises the following steps: on the basis of a visibility graph, generating a communication undirected weighted graph, introducing a Laplacian matrix, and reflecting the connectivity of the graph according to a second small feature value of the Laplacian matrix; the arrival time of the robots is controlled by adjusting the side weights, space-time conflicts are avoided, the robot path planning sequence is determined according to the contribution value of the second small feature value, and the overall performance of multi-robot cooperation is optimized; according to the method, data in robot path planning are integrated through the Transform model, the comprehensiveness of environmental understanding is improved, spatial constraints of the path are optimized through a message passing mechanism of the graph neural network, the robot dynamically adapts to environmental changes, the path is generated through the generative adversarial network, diversity and high efficiency are achieved, local optimum is avoided, and the method is suitable for being popularized and applied. And dynamic obstacle avoidance and real-time path optimization are realized through cooperation of multiple models.

The present application discloses a positioning method based on a dynamic region of a robot, comprising: step 1, upon verifying that a repositioning result fails, determining whether a matching score in the repositioning result is greater than a first preset score threshold, and if yes, proceeding to step 2; step 2, performing weighted calculation on the basis of the matching scores of point cloud points falling on a corresponding region of a target map to obtain a target matching score, and then proceeding to step 3; step 3, determining whether the target matching score is greater than a second preset score threshold, if yes, indicating that the positioning is successful, or otherwise, proceeding to step 4; step 4, setting a search distance on the basis of the relationship between a mask extracted from a reference map and a mask extracted from a depth dynamic region of the target map, and selecting a navigation target point within a preset positioning distance range on the basis of the search distance, and then executing step 5; and step 5, when a robot moves to the navigation target point, obtaining a repositioning result by means of a repositioning matching algorithm, and verifying the repositioning result to determine whether the positioning is successful.

The invention provides an intelligent early warning monitoring method for oil and gas pipe network optical fiber sensing and robot collaborative inspection, and relates to the field of oil and gas pipe network monitoring, which comprises the steps of acquiring scattered signals of an optical fiber sensing system to identify abnormal events, setting a robot dynamic inspection strategy, controlling a robot to move to collect data and determining an abnormal source position. Dense detection is carried out to construct a spatial diffusion matrix and a time evolution matrix, the abnormal expansion direction and speed are determined through tensor operation, and finally a control instruction is generated. According to the invention, collaborative operation of optical fiber sensing and the robot is realized, the early warning accuracy of abnormal events is improved, the emergency response speed is accelerated, and the safety risk is reduced.

The invention relates to the technical field of orthopedic medical treatment, in particular to a dynamic path planning method and system for an orthopedic surgical robot, and the method comprises the steps: fusing multi-modal medical diagnosis information to construct a three-dimensional model, and formulating an initial scheme containing an obstacle avoidance boundary; in the operation, multi-source data are collected through multi-device cooperation and time sequence alignment is carried out; performing two-dimensional dynamic calibration to generate a compensation value; updating paths and parameters according to a hierarchical strategy, and verifying a security boundary; real-time feedback is realized; and convenient manual intervention is provided. The system comprises a multi-modal data processing module, an intra-operative collaborative acquisition module, a dynamic calibration module, a path adjustment module, a man-machine interaction module and a main control module, and full-process automation is achieved. The operation accuracy and safety can be improved, manual operation of doctors is reduced, clinical requirements are met, and the method has important clinical application value.

The invention discloses an industrial robot dynamic feedback normalization reinforcement learning optimization method and system. The method comprises the following steps: S1, constructing an experience playback buffer area D, Q network parameter theta and a target network parameter theta-in an industrial robot memory; s2, the industrial robot executes an action based on a control instruction output by the current Q network, and an original feedback signal rraw and a next state st + 1 are observed; s3, performing nonlinear normalization processing on the original feedback signal rraw; s4, if the industrial robot triggers a system failure protection mechanism, determining a negative system failure penalty term p; s5, the complete state transition tuple of the final feedback signal rfinal is stored in an experience playback buffer area D; s6, sampling training data from the experience playback buffer area D, calculating a target Q value, and updating a Q network parameter theta; and S7, when a preset network updating condition is satisfied, synchronizing the Q network parameter theta to the target network theta-, and updating the parameters alpha t and beta t in the nonlinear normalization based on the gradient statistics.