Robot intelligent control method and system based on image analysis algorithm

Abstract

The invention discloses a robot intelligent control method and system based on an image analysis algorithm, and the method comprises the steps: recognizing the multi-target features of an operation environment image through an optimized Single Shot Multi-Box Detector algorithm, generating a feature vector set, inputting a robot multi-dimensional posture regulation and control model to calculate a joint angle parameter, determining a motion planning strategy through the combination of robot intelligent control parameters, and carrying out the operation of a robot. And the joint driving device is controlled to realize robot movement. During movement, joint angle and tail end position information is collected in real time and fed back to the regulation and control model, and a movement planning strategy is dynamically adjusted through comparative analysis. The system comprises six mutually connected units including an image feature recognition processing unit and a posture regulation and control calculation unit which are respectively corresponding to each step function of the method. According to the invention, the whole process intelligence of the robot from environment perception, motion planning to accurate control is realized, and the accuracy and stability of robot operation are improved.

Description

Technical Field

[0001] The present invention relates to the field of robot intelligent control, and in particular to a robot intelligent control method and system based on an image analysis algorithm. Background Art

[0002] With the rapid development of intelligent manufacturing and automation technologies, robots are increasingly being used in industrial production, logistics, warehousing, and service industries. Robotic intelligent control technology based on image analysis algorithms has become a key enabler for autonomous and precise robot operation. It empowers robots to perceive their environment and identify targets, enabling them to make autonomous decisions based on environmental changes and complete complex tasks.

[0003] However, existing intelligent robot control technology has numerous shortcomings. In terms of target recognition and processing, traditional image analysis algorithms have low efficiency and accuracy in recognizing multiple targets in complex environments. They are unable to quickly and accurately locate target objects and feature points, making it difficult for robots to accurately perceive their operating environment, which in turn affects subsequent control decisions. For example, in scenes with large lighting variations and complex backgrounds, traditional algorithms are prone to missed and false detections, preventing robots from correctly executing their tasks.

[0004] When it comes to formulating and adjusting motion control strategies, existing robot motion planning strategies often lack comprehensive consideration of the robot's structural parameters, environmental constraints, and real-time feedback. Traditional motion planning methods cannot adapt promptly to the robot's actual operating state and environmental changes, which can easily lead to collisions and unstable motion during robot movement. This not only reduces the robot's efficiency but also poses safety risks, making it difficult to meet the demands of high-precision, complex tasks. The proposed intelligent robot control method and system, based on image analysis algorithms, effectively addresses these issues. Summary of the Invention

[0005] In order to overcome the shortcomings and deficiencies of the existing technology, the present invention provides a robot intelligent control method and system based on image analysis algorithm.

[0006] The technical solution adopted by the present invention is a robot intelligent control method based on an image analysis algorithm, comprising the following steps: Step S1: Use the optimized Single Shot Multi-Box Detector algorithm to perform multi-target feature recognition on the acquired robot working environment image data, locate the target objects and feature points in the image data, and generate the corresponding feature vector set; Step S2: Input the feature vector set into a pre-built robot multi-dimensional posture control model. The model calculates the angle parameters that need to be adjusted for each joint of the robot based on the spatial position information of the target object in the feature vector set and the initial posture information of the robot itself. Step S3: Based on various parameters of the robot's intelligent control and combined with the joint angle parameters output by the robot's multi-dimensional posture control model, the robot's motion planning strategy is determined. The motion planning strategy includes the motion trajectory planning of the robot's end effector and the motion sequence planning of each joint; Step S4: According to the motion planning strategy, the driving devices of the robot joints are controlled to rotate the robot joints according to the predetermined angle parameters and motion sequence; Step S5: During the robot's motion, the actual angle information of each joint of the robot and the actual position information of the end effector are obtained in real time, and fed back to the robot's multi-dimensional posture control model; Step S6: The robot's multi-dimensional posture control model compares and analyzes the feedback information with the preset target posture information, dynamically adjusts the robot's subsequent motion planning strategy, and performs continuous and precise motion control of the robot.

[0007] Furthermore, the robot multi-dimensional posture control model is formulated as follows: in, Indicates the robot Target adjustment angle for each joint; is the weight coefficient set according to the robot structure and operation requirements; It is based on the target object feature vector and the robot's current state feature vector Nonlinear mapping function; For the robot The current angle of each joint; Represents the deviation vector between the current position of the robot end effector and the target position.

[0008] Furthermore, the various parameters of the robot intelligent control participate in constructing the following robot motion planning decision model: in, represents the finalized robot motion planning strategy; is the robot intelligent control parameter vector including robot power parameters, velocity parameters, and acceleration parameters; is the joint angle parameter vector output by the robot's multi-dimensional posture control model; Represents the robot's operating environment and its own structural constraint vector; is a complex decision function based on the above parameters.

[0009] Furthermore, in step S4, the process of controlling the driving device of each joint of the robot is as follows: in, Represents the driving force or torque vector output by the driving device; and is the driving control coefficient; is the target angle vector of each joint output by the robot's multi-dimensional posture control model; is the feedback vector of the actual angular velocity of each joint of the robot.

[0010] Furthermore, in step S5, the process of feeding back the actual angle information of each joint of the robot and the actual position information of the end effector to the multi-dimensional posture control model of the robot is achieved through the following information fusion model: in, Represents the fused feedback information vector; is the actual angle information vector of each joint of the robot; is the actual position information vector of the robot end effector; is the feedback information vector of the previous moment; is the information fusion function.

[0011] Furthermore, in step S6, the robot multi-dimensional posture control model dynamically adjusts the robot's subsequent motion planning strategy by adjusting the model as follows: in, represents the adjusted robot motion planning strategy; Plan strategies for pre-adjustment movement; is the fused feedback information vector; is the preset target posture information vector; Tuning functions for motion planning strategies.

[0012] Furthermore, the robot motion planning strategy is determined in step S3, specifically including: Step S3.1: Calculate the shortest time path from the current position to the target position of the robot end effector based on the joint angle parameters output by the robot's multi-dimensional posture control model and the speed and acceleration parameters of the robot's intelligent control; Step S3.2: Decompose the shortest time path into the motion trajectory of each joint according to the motion range and structural constraints of each joint of the robot; Step S3.3: Determine the motion sequence of each joint based on the motion trajectory of each joint and the dynamic parameters of the robot's intelligent control to avoid motion interference and ensure motion smoothness.

[0013] Furthermore, in step S4, the driving devices of the robot joints are controlled according to the motion planning strategy, specifically including: Step S4.1: activating the corresponding driving devices in sequence according to the motion sequence of each joint in the motion planning strategy; Step S4.2: According to the motion trajectory of each joint in the motion planning strategy, the output parameters of the driving device are adjusted in real time so that the joint rotates according to the predetermined angle parameters; Step S4.3: During the joint rotation process, the operating status of the drive device is monitored. If any abnormality occurs, the emergency control program is immediately activated to adjust the drive device or stop the movement.

[0014] Furthermore, in step S5, the actual angle information of each joint of the robot and the actual position information of the end effector are obtained in real time. The specific operations are as follows: Step S5.1: Use the angle sensors installed at each joint of the robot to collect the actual rotation angle of the joint in real time and convert the angle information into a digital signal; Step S5.2: Obtain the actual position coordinates of the end effector in space in real time through the position sensor installed on the end effector of the robot; Step S5.3: Encode the collected joint angle digital signals and end effector position coordinate information to form a feedback information format that can be recognized by the robot's multi-dimensional posture control model.

[0015] Robot intelligent control system based on image analysis algorithm, including: An image feature recognition processing unit, which uses the optimized Single Shot Multi-Box Detector algorithm to perform multi-target feature recognition on the acquired robot working environment image data, locate target objects and feature points, and generate corresponding feature vector sets; The posture control calculation unit is connected to the image feature recognition processing unit and is used to input the feature vector set into the robot's multi-dimensional posture control model, and calculate the angle parameters that need to be adjusted for each joint of the robot based on the spatial position information of the target object in the feature vector set and the initial posture information of the robot itself; The motion planning decision unit is connected to the posture control calculation unit. It determines the robot's motion planning strategy based on the various parameters of the robot's intelligent control and the joint angle parameters output by the robot's multi-dimensional posture control model. The joint drive control unit is connected to the motion planning and decision-making unit, and controls the drive devices of each joint of the robot according to the motion planning strategy, so that each joint of the robot rotates according to the predetermined angle parameters and motion sequence; The information acquisition and feedback unit is connected to the joint drive control unit and the posture control calculation unit. During the robot's motion, it obtains the actual angle information of each joint of the robot and the actual position information of the end effector in real time and feeds it back to the posture control calculation unit. The motion strategy adjustment unit is connected to the information acquisition and feedback unit and the motion planning and decision-making unit. The robot's multi-dimensional posture control model compares and analyzes the feedback information with the preset target posture information. Through this unit, the robot's subsequent motion planning strategy is dynamically adjusted to achieve continuous and precise motion control of the robot.

[0016] Beneficial Effects: The present invention proposes a robot intelligent control method and system based on an image analysis algorithm. At the target recognition level, an optimized Single Shot Multi-Box Detector algorithm is used to perform multi-target feature recognition on the working environment image data, accurately locate the target object and feature points, and generate a feature vector set. Compared with traditional algorithms, it can better adapt to complex environments, greatly improve the efficiency and accuracy of target recognition, avoid missed detection and false detection problems, and lay the foundation for the robot to accurately perceive the environment. In terms of motion control, the robot's multi-dimensional posture control model calculates the joint angle parameters based on the spatial position of the target object and the robot's initial posture. At the same time, it combines various parameters of the robot's intelligent control to construct a motion planning decision model. The motion planning strategy is determined by comprehensively considering the robot's power, speed, acceleration parameters and environmental and structural constraints. Scientific planning is carried out from the end effector trajectory to the joint motion sequence. In addition, the robot's joint angle and end position information are obtained in real time through the information acquisition and feedback unit, and fed back to the control model after processing by the information fusion model. The adjustment model is used to dynamically optimize the motion planning strategy. Compared with traditional motion control methods, this closed-loop control mechanism can adjust strategies in real time according to the actual state of the robot and environmental changes, effectively avoid collisions, ensure motion smoothness, significantly improve the robot's work efficiency and operational safety, meet the requirements of high-precision and complex tasks, and achieve continuous and precise motion control of the robot. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 is a flow chart of the method steps of the present invention; Figure 2 It is a diagram of the system unit composition of the present invention. DETAILED DESCRIPTION

[0018] It should be noted that, unless there is a conflict, the embodiments and features in the embodiments of this application can be combined with each other. The present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] like Figure 1 As shown, the robot intelligent control method based on the image analysis algorithm includes the following steps: Step S1: Use the optimized Single Shot Multi-Box Detector algorithm to perform multi-target feature recognition on the acquired robot working environment image data, locate the target objects and feature points in the image data, and generate the corresponding feature vector set; Specifically, an optimized SSD (Single Shot Multi-Box Detector) algorithm is used to perform multi-target feature recognition on acquired image data of the robot's work environment, locate target objects and feature points in the image data, and generate corresponding feature vector sets. This step improves the accuracy and speed of multi-target detection in complex scenarios by using an optimized SSD algorithm, a multi-scale feature map fusion mechanism, and an improved anchor box generation strategy. The algorithm obtains multi-level semantic information from the image through a feature extraction network, and combines it with an attention mechanism to enhance the feature expression of the target area. It also uses a non-maximum suppression optimization algorithm to remove redundant detection boxes, ensuring that each real target is uniquely identified. The generated feature vector set contains multi-dimensional information such as the target object's position coordinates, geometric features, and texture features, providing accurate environmental perception data for the subsequent robot's multi-dimensional posture control model.

[0020] In terms of implementation, this step first preprocesses the input image, including normalization, rescaling, and other operations, to adapt to the optimized SSD algorithm input requirements. The algorithm fuses feature maps from different levels through a feature pyramid network to achieve efficient detection of targets of different sizes. At the same time, based on the characteristics of the robot's operating scene, the aspect ratio and scale distribution of the anchor frame are optimized to improve the detection ability of irregular targets. In the feature vector generation stage, dimensionality reduction techniques such as principal component analysis are used to compress the feature dimensions while retaining key information to ensure the efficiency and representativeness of the feature vector set, providing a reliable data foundation for subsequent posture control calculations.

[0021] Step S2: Input the feature vector set into a pre-built robot multi-dimensional posture control model. The model calculates the angle parameters that need to be adjusted for each joint of the robot based on the spatial position information of the target object in the feature vector set and the initial posture information of the robot itself. Specifically, the feature vector set is input into a pre-built multi-dimensional robot posture control model. This model calculates the angle parameters that need to be adjusted for each robot joint based on the spatial position information of the target object in the feature vector set and the robot's own initial posture information. Based on a deep learning architecture, this model integrates kinematic constraints and dynamic characteristics. Through a multi-input, multi-output neural network structure, it achieves a nonlinear mapping from the target spatial position to the joint angle parameters. During the model training phase, a reinforcement learning strategy is employed, combining forward kinematics and inverse kinematics algorithms to optimize network parameters to improve the accuracy of angle calculations. At the same time, the model introduces physical limitations of the robot joints and motion smoothness constraints to ensure that the output angle parameters conform to the robot's actual motion capabilities.

[0022] During this step, the spatial position of the target object in the feature vector set is first converted into coordinates in the robot coordinate system, and then the input tensor is constructed based on the robot's current pose information. The model processes temporal features using a multilayer perceptron and a long-short-term memory network to capture the correlation between the target's motion trajectory and the robot's posture changes. During the angle parameter calculation process, a hierarchical optimization strategy is employed, first calculating the rough angles of the main joints, then refining the parameters through local fine-tuning of the network to ensure the accuracy of the calculated results. The output angle parameters are kinematically verified to ensure that they meet the robot's kinematic constraints, providing a feasible joint configuration for subsequent motion planning.

[0023] Step S3: Based on various parameters of the robot's intelligent control and combined with the joint angle parameters output by the robot's multi-dimensional posture control model, the robot's motion planning strategy is determined. The motion planning strategy includes the motion trajectory planning of the robot's end effector and the motion sequence planning of each joint; Specifically, the robot's motion planning strategy is determined based on the various parameters of the robot's intelligent control and the joint angle parameters output by the robot's multidimensional posture control model. This strategy includes planning the trajectory of the robot's end effector and the motion sequence of each joint. This step comprehensively considers factors such as the robot's dynamic parameters, kinematic constraints, and task priorities, and uses a multi-objective optimization algorithm to generate the optimal motion strategy. A cost function is constructed to evaluate different trajectory solutions for smoothness, energy consumption, execution time, and other metrics. A rapid exploration random tree algorithm is then used to search the feasible solution space, ultimately determining the optimal trajectory and joint motion sequence that meets the task requirements.

[0024] During implementation, a kinematic and dynamic model was first established based on the robot's intelligent control parameters, including constraints such as joint torque limits, speed limits, and acceleration limits. Joint angle parameters output by the robot's multidimensional posture control model, combined with the target object's motion trend prediction, were used to generate multiple candidate trajectory plans. A cost function was used to comprehensively evaluate each plan, taking into account metrics such as trajectory smoothness, obstacle avoidance, and energy consumption, to select the optimal trajectory. Furthermore, a priority scheduling algorithm was used to determine the movement sequence of each joint, ensuring stability and coordination during the robot's mission and avoiding motion interference and singularity issues.

[0025] Step S4: According to the motion planning strategy, the driving devices of the robot joints are controlled to rotate the robot joints according to the predetermined angle parameters and motion sequence; Specifically, based on the motion planning strategy, the robot's joint actuators are controlled to ensure that each joint rotates according to predetermined angle parameters and motion sequences. This step utilizes a hierarchical control architecture. The upper-level controller generates the target position sequence for each joint based on the motion planning strategy, while the lower-level controller adjusts the actuator outputs using a closed-loop feedback control algorithm to ensure that the joint's actual motion trajectory tracks the target trajectory. During the drive control process, an adaptive control algorithm is incorporated to compensate for model uncertainty and external disturbances, improving control accuracy and system robustness.

[0026] During implementation, the joint angle parameters in the motion planning strategy are first converted into control instructions for the drive device, including voltage, current, or pulse signals. A proportional-integral-differential controller is used as the basic control algorithm, and the control quantity is adjusted in real time based on the deviation between the actual joint angle and the target angle. At the same time, feedforward control is introduced to compensate for the dynamic characteristics of the system and improve the system response speed. During the movement process, sensors monitor the joint angle and velocity information in real time. A Kalman filter is used to fuse multi-sensor data, estimate the system state and predict future trends, provide more accurate feedback signals to the controller, and ensure the accuracy and stability of joint movement.

[0027] Step S5: During the robot's motion, the actual angle information of each joint of the robot and the actual position information of the end effector are obtained in real time, and fed back to the robot's multi-dimensional posture control model; Specifically, during robot motion, the actual angle information of each robot joint and the actual position of the end effector are acquired in real time and fed back into the robot's multidimensional posture control model. This step collects high-precision motion state data through a sensor network distributed across the robot's joints and end effector. Data synchronization technology is used to ensure the temporal consistency of multi-sensor data, and digital filtering algorithms are used to remove noise interference and extract authentic and valid motion information. The feedback data undergoes coordinate transformation and format conversion to generate a feedback vector that meets the input requirements of the robot's multidimensional posture control model.

[0028] During implementation, joint angle information is collected in real time via encoders or resolvers, while end-effector position information is acquired via vision sensors or inertial measurement units. A timestamp synchronization mechanism ensures temporal alignment of data from different sensors, and a Kalman filter or extended Kalman filter is used to fuse multi-source data to improve state estimation accuracy. After preprocessing, the feedback data is transmitted via a communication interface to the robot's multidimensional posture control model, providing the model with real-time motion state updates for dynamic adjustment decisions.

[0029] Step S6: The robot's multi-dimensional posture control model compares and analyzes the feedback information with the preset target posture information, dynamically adjusts the robot's subsequent motion planning strategy, and performs continuous and precise motion control of the robot.

[0030] Specifically, the robot's multi-dimensional posture control model compares and analyzes feedback information with preset target posture information, dynamically adjusting the robot's subsequent motion planning strategy to achieve continuous and precise motion control. This step uses an error analysis algorithm to calculate the deviation between the current and target postures, employs model predictive control techniques to predict future state trends, and combines this with an optimization algorithm to generate an adjustment strategy. This adjustment process comprehensively considers system constraints and task priorities, ensuring both control accuracy and the safety and stability of the robot's motion.

[0031] During implementation, feedback information and target posture information are first mapped to the same coordinate system, and multi-dimensional deviation indicators such as position error and angular error are calculated. The robot's dynamics model is used to predict state changes at several future moments, and a rolling optimization algorithm is used to solve the optimal control input sequence. When adjusting the motion planning strategy, a smooth transition algorithm is used to avoid sudden changes and ensure the continuity of the robot's motion. Simultaneously, the adjusted strategy is verified for feasibility, checking whether it meets joint limit constraints and environmental restrictions. Ultimately, an executable dynamic adjustment plan is generated, achieving continuous and precise motion control of the robot in complex environments.

[0032] Preferably, the robot multi-dimensional posture control model is formulated as follows: in, Indicates the robot Target adjustment angle for each joint; is the weight coefficient set according to the robot structure and operation requirements; It is based on the target object feature vector and the robot's current state feature vector Nonlinear mapping function; For the robot The current angle of each joint; Represents the deviation vector between the current position of the robot end effector and the target position.

[0033] Specifically, the robot's multidimensional posture control model achieves accurate prediction and dynamic adjustment of the robot's joint angles through an innovative parameter fusion mechanism. The weight coefficients α, β, and γ correspond to the influence of target feature mapping, historical state continuity, and position error compensation, respectively. These factors are dynamically optimized based on the robot's structural parameters and the characteristics of the work scenario using a machine learning algorithm. The nonlinear mapping function f comprehensively considers the target object's feature vector and the robot's current state vector, employing a deep neural network to model complex relationships in high-dimensional space, ensuring the model's adaptability to environmental changes. By iteratively updating joint angle parameters, the model effectively balances target tracking accuracy and motion smoothness, providing a reliable posture control foundation for robot motion planning.

[0034] Preferably, the various parameters of the robot intelligent control participate in constructing the following robot motion planning decision model: in, represents the finalized robot motion planning strategy; is the robot intelligent control parameter vector including robot power parameters, velocity parameters, and acceleration parameters; is the joint angle parameter vector output by the robot's multi-dimensional posture control model; Represents the robot's operating environment and its own structural constraint vector; is a complex decision function based on the above parameters.

[0035] Specifically, the robot motion planning decision model systematically integrates the robot's intelligent control parameters, posture control results and environmental constraints, and uses complex decision functions to Achieve multi-objective optimization. The control parameter vector W includes key indicators such as power, speed, acceleration, etc., which directly affect the robot's motion efficiency; the joint angle vector θ provides kinematic constraints; and the environmental constraint vector G includes the working space restrictions and obstacle distribution information. Decision function A hierarchical optimization architecture is adopted, combining genetic algorithm and simulated annealing algorithm to search for the optimal motion strategy in the solution space that meets all constraints, ensuring the safety, efficiency and executability of robot motion.

[0036] Preferably, in step S4, the process of controlling the driving device of each joint of the robot is as follows: in, Represents the driving force or torque vector output by the driving device; and is the driving control coefficient; is the target angle vector of each joint output by the robot's multi-dimensional posture control model; is the feedback vector of the actual angular velocity of each joint of the robot.

[0037] Specifically, the drive control formula achieves precise control of the robot's joint drive mechanism through a dual-parameter adjustment mechanism. The drive control coefficients δ and ε adjust the influence weights of the target angle and feedback angular velocity, respectively, and are dynamically adjusted based on the robot's dynamic model and load characteristics. The target angle vector provides a position control reference, while the feedback angular velocity vector is used to compensate for system dynamic errors in real time. A Kalman filter algorithm is used to fuse multi-sensor data to improve feedback accuracy. This control strategy, through a feedforward-feedback composite control architecture, effectively suppresses system nonlinearities and external interference, ensuring that joint motion trajectory tracking errors remain within the submillimeter range.

[0038] Preferably, in step S5, the process of feeding back the actual angle information of each joint of the robot and the actual position information of the end effector to the multi-dimensional posture control model of the robot is achieved by the following information fusion model: in, Represents the fused feedback information vector; is the actual angle information vector of each joint of the robot; is the actual position information vector of the robot end effector; is the feedback information vector of the previous moment; is the information fusion function.

[0039] Specifically, the information fusion model achieves efficient integration and state estimation of multi-source sensor information through a time-series data processing mechanism. A recursive neural network architecture is used to process historical feedback information vectors, combined with current joint angle and position information, to reconstruct the state in both temporal and spatial dimensions using the information fusion function ρ. The model incorporates an attention mechanism to automatically assign weights to each sensor data point and adaptively adjust the fusion strategy for different operational scenarios. Sliding window filtering and Kalman smoothing techniques effectively reduce the impact of measurement noise, producing continuous and smooth state estimation results that provide a reliable basis for motion strategy adjustments.

[0040] Preferably, in step S6, the robot multi-dimensional posture control model dynamically adjusts the robot's subsequent motion planning strategy by adjusting the model as follows: in, represents the adjusted robot motion planning strategy; Plan strategies for pre-adjustment movement; is the fused feedback information vector; is the preset target posture information vector; Tuning functions for motion planning strategies.

[0041] Specifically, the motion planning adjustment model, based on the principles of model predictive control, dynamically optimizes the robot's motion strategy. This model processes the current motion strategy through a rolling optimization window, integrates feedback information with the target pose, and employs a quadratic programming algorithm to solve the optimal control sequence. The adjustment function σ includes submodules such as kinematic feasibility verification, dynamic constraint satisfaction assessment, and task prioritization to ensure that the adjusted strategy meets the robot's physical limitations and operational requirements. Through multi-step optimization within the predictive time domain, potential collision risks are avoided in advance, achieving smooth transitions in motion trajectories and continuous task execution.

[0042] Preferably, determining the robot motion planning strategy in step S3 specifically includes: Step S3.1: Calculate the shortest time path from the current position to the target position of the robot end effector based on the joint angle parameters output by the robot's multi-dimensional posture control model and the speed and acceleration parameters of the robot's intelligent control; Step S3.2: Decompose the shortest time path into the motion trajectory of each joint according to the motion range and structural constraints of each joint of the robot; Step S3.3: Determine the motion sequence of each joint based on the motion trajectory of each joint and the dynamic parameters of the robot's intelligent control to avoid motion interference and ensure motion smoothness.

[0043] Specifically, based on time-optimal control theory and combined with the robot's dynamic constraints, a rapid marching method is used to calculate the shortest path for the end effector. Secondly, an inverse kinematics solution is used to map the Cartesian space trajectory to the joint space, generating reference trajectories for each joint. Finally, a task decomposition and priority sorting algorithm is used to determine the joint motion sequence, employing a heuristic search algorithm to avoid singular configurations and motion interference. This solution reduces energy consumption and extends the robot's service life by optimizing joint motion trajectories while ensuring motion efficiency.

[0044] Preferably, the step S4 controls the driving devices of the joints of the robot according to the motion planning strategy, specifically including: Step S4.1: activating the corresponding driving devices in sequence according to the motion sequence of each joint in the motion planning strategy; Step S4.2: According to the motion trajectory of each joint in the motion planning strategy, the output parameters of the driving device are adjusted in real time so that the joint rotates according to the predetermined angle parameters; Step S4.3: During the joint rotation process, the operating status of the drive device is monitored. If any abnormality occurs, the emergency control program is immediately activated to adjust the drive device or stop the movement.

[0045] Specifically, precise control of the drive unit is achieved through a three-tiered cascade control architecture. The first tier activates the drive system for the corresponding joint based on a task scheduling algorithm, employing a state machine model to manage transitions between different motion phases. The second tier uses an adaptive PID controller to track the reference trajectory in real time, dynamically adjusting control parameters based on load changes. The third tier implements fault diagnosis and fault-tolerant control, monitoring state parameters such as drive current and temperature, and employing threshold detection and pattern recognition algorithms to promptly detect anomalies and trigger emergency protection mechanisms. This architecture ensures the robot's operational reliability and safety under complex working conditions.

[0046] Preferably, in step S5, the actual angle information of each joint of the robot and the actual position information of the end effector are obtained in real time, and the specific operations are as follows: Step S5.1: Use the angle sensors installed at each joint of the robot to collect the actual rotation angle of the joint in real time and convert the angle information into a digital signal; Step S5.2: Obtain the actual position coordinates of the end effector in space in real time through the position sensor installed on the end effector of the robot; Step S5.3: Encode the collected joint angle digital signals and end effector position coordinate information to form a feedback information format that can be recognized by the robot's multi-dimensional posture control model.

[0047] Specifically, the data acquisition process utilizes a standardized sensor network to achieve synchronous data collection and preprocessing from multiple sources. Angle sensors utilize absolute encoders to provide high-precision position feedback. Position sensors integrate a laser rangefinder with a visual positioning system to measure the end effector's six-degree-of-freedom pose. Data acquisition utilizes a distributed architecture, with time synchronization of sensor nodes achieved via a fieldbus. The preprocessing stage implements digital filtering, coordinate transformation, and data compression, using a lossless coding algorithm to convert raw data into an efficient feature vector representation, ensuring the real-time and validity of feedback information.

[0048] like Figure 2 As shown, the robot intelligent control system based on image analysis algorithm includes: An image feature recognition processing unit, which uses the optimized Single Shot Multi-Box Detector algorithm to perform multi-target feature recognition on the acquired robot working environment image data, locate target objects and feature points, and generate corresponding feature vector sets; The posture control calculation unit is connected to the image feature recognition processing unit and is used to input the feature vector set into the robot's multi-dimensional posture control model, and calculate the angle parameters that need to be adjusted for each joint of the robot based on the spatial position information of the target object in the feature vector set and the initial posture information of the robot itself; The motion planning decision unit is connected to the posture control calculation unit. It determines the robot's motion planning strategy based on the various parameters of the robot's intelligent control and the joint angle parameters output by the robot's multi-dimensional posture control model. The joint drive control unit is connected to the motion planning and decision-making unit, and controls the drive devices of each joint of the robot according to the motion planning strategy, so that each joint of the robot rotates according to the predetermined angle parameters and motion sequence; The information acquisition and feedback unit is connected to the joint drive control unit and the posture control calculation unit. During the robot's motion, it obtains the actual angle information of each joint of the robot and the actual position information of the end effector in real time and feeds it back to the posture control calculation unit. The motion strategy adjustment unit is connected to the information acquisition and feedback unit and the motion planning and decision-making unit. The robot's multi-dimensional posture control model compares and analyzes the feedback information with the preset target posture information. Through this unit, the robot's subsequent motion planning strategy is dynamically adjusted to achieve continuous and precise motion control of the robot.

[0049] The robot intelligent control method and system based on image analysis algorithms are difficult to handle in complex environments due to the difficulty of traditional image analysis algorithms in target recognition and environmental perception. They suffer from low target recognition efficiency and false detection and missed detection. This method and system utilizes an optimized Single Shot Multi-Box Detector algorithm to quickly and accurately identify multiple targets in image data of the robot's operating environment. This not only locates the target object, but also captures its feature points and generates a set of feature vectors. This enables the robot to accurately perceive its surroundings even in scenes with changing lighting and complex backgrounds, providing a reliable basis for subsequent decision-making and fundamentally addressing the shortcomings of traditional algorithms in environmental adaptability.

[0050] In the motion planning and control link, the existing technology lacks comprehensive consideration of robot parameters, environmental constraints and real-time feedback, resulting in unreasonable robot motion planning and unstable execution. This system uses the robot's multi-dimensional posture control model to combine the spatial position of the target object and the robot's initial posture to calculate the joint angle parameters, and then uses the power, speed and other parameters of the robot's intelligent control to build a motion planning decision model to scientifically plan the end effector motion trajectory and joint motion sequence. At the same time, during the movement process, the information acquisition and feedback unit obtains joint angle and end position information in real time, and after information fusion, it feeds back to the control model, and realizes dynamic optimization of motion planning through the motion strategy adjustment unit. This multi-model collaboration and real-time feedback adjustment mechanism allows the robot to avoid collisions while ensuring motion stability during movement, greatly improving work efficiency and safety, and significantly outperforming traditional motion control technology.

[0051] At the system architecture level, the six modules of this intelligent control system have clear divisions of labor and closely collaborate. The image feature recognition and processing unit, the posture control calculation unit, and the motion planning and decision-making unit work together in sequence, from environmental information acquisition to motion strategy generation, and then to drive control and feedback adjustment, forming an efficient and orderly control process. Compared to traditional systems with loose architectures and poor coordination between components, this system, through modular design and tight connectivity, achieves precise and efficient intelligent control of robots, providing solid technical support for their application in complex mission scenarios.

[0052] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "disposed," "installed," "connected," "connected," and "fixed" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; they may refer to mechanical connections or electrical connections; they may refer to direct connections or indirect connections through an intermediate medium; and they may refer to internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0053] Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various equivalent changes, modifications, substitutions and variations may be made to these embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A robot intelligent control method based on image analysis algorithm, characterized in that: The steps include: Step S1: Use the optimized Single Shot Multi-Box Detector algorithm to perform multi-target feature recognition on the acquired robot working environment image data, locate the target objects and feature points in the image data, and generate the corresponding feature vector set; Step S2: Input the feature vector set into a pre-built robot multi-dimensional posture control model. The model calculates the angle parameters that need to be adjusted for each joint of the robot based on the spatial position information of the target object in the feature vector set and the initial posture information of the robot itself. Step S3: Based on various parameters of the robot's intelligent control and combined with the joint angle parameters output by the robot's multi-dimensional posture control model, the robot's motion planning strategy is determined. The motion planning strategy includes the motion trajectory planning of the robot's end effector and the motion sequence planning of each joint; Step S4: According to the motion planning strategy, the driving devices of the joints of the robot are controlled to rotate the joints of the robot according to the predetermined angle parameters and motion sequence; Step S5: During the robot's motion, the actual angle information of each joint of the robot and the actual position information of the end effector are obtained in real time, and fed back to the robot's multi-dimensional posture control model; Step S6: The robot's multi-dimensional posture control model compares and analyzes the feedback information with the preset target posture information, dynamically adjusts the robot's subsequent motion planning strategy, and performs continuous motion control of the robot.

2. The robot intelligent control method based on image analysis algorithm according to claim 1 is characterized in that: The robot multi-dimensional posture control model is formulated as follows: in, Indicates the robot Target adjustment angle for each joint; is the weight coefficient set according to the robot structure and operation requirements; It is based on the target object feature vector and the robot's current state feature vector Nonlinear mapping function; For the robot The current angle of each joint; Represents the deviation vector between the current position of the robot end effector and the target position.

3. The robot intelligent control method based on image analysis algorithm according to claim 1, characterized in that: The various parameters of the robot intelligent control are used to construct the following robot motion planning decision model: in, represents the finalized robot motion planning strategy; is the robot intelligent control parameter vector including robot power parameters, velocity parameters, and acceleration parameters; is the joint angle parameter vector output by the robot's multi-dimensional posture control model; Represents the robot's operating environment and its own structural constraint vector; is a complex decision function based on the above parameters.

4. The robot intelligent control method based on image analysis algorithm according to claim 1, characterized in that: The step S4 is a process of controlling the driving devices of the robot joints, and the formula is: in, Represents the driving force or torque vector output by the driving device; and is the driving control coefficient; is the target angle vector of each joint output by the robot's multi-dimensional posture control model; is the feedback vector of the actual angular velocity of each joint of the robot.

5. The robot intelligent control method based on image analysis algorithm according to claim 1, characterized in that: The step S5, in which the actual angle information of each joint of the robot and the actual position information of the end effector are fed back to the multi-dimensional posture control model of the robot, is implemented by the following information fusion model: in, Represents the fused feedback information vector; is the actual angle information vector of each joint of the robot; is the actual position information vector of the robot end effector; is the feedback information vector of the previous moment; is the information fusion function.

6. The robot intelligent control method based on image analysis algorithm according to claim 1, characterized in that: Step S6, in which the robot's multi-dimensional posture control model dynamically adjusts the robot's subsequent motion planning strategy, is performed by adjusting the model as follows: in, represents the adjusted robot motion planning strategy; Plan strategies for pre-adjustment movement; is the fused feedback information vector; is the preset target posture information vector; Tuning functions for motion planning strategies.

7. The robot intelligent control method based on image analysis algorithm according to claim 1 is characterized in that: The step S3, determining the robot motion planning strategy, specifically includes: Step S3.1: Calculate the shortest time path from the current position to the target position of the robot end effector based on the joint angle parameters output by the robot's multi-dimensional posture control model and the speed and acceleration parameters of the robot's intelligent control; Step S3.2: Decompose the shortest time path into the motion trajectory of each joint according to the motion range and structural constraints of each joint of the robot; Step S3.3: Determine the motion sequence of each joint based on the motion trajectory of each joint and the dynamic parameters of the robot's intelligent control to avoid motion interference and ensure motion smoothness.

8. The robot intelligent control method based on image analysis algorithm according to claim 1 is characterized in that: The step S4, controlling the driving devices of the robot joints according to the motion planning strategy, specifically includes: Step S4.1: activating the corresponding driving devices in sequence according to the motion sequence of each joint in the motion planning strategy; Step S4.2: According to the motion trajectory of each joint in the motion planning strategy, the output parameters of the driving device are adjusted in real time so that the joint rotates according to the predetermined angle parameters; Step S4.3: During the joint rotation process, the operating status of the drive device is monitored. If any abnormality occurs, the emergency control program is immediately activated to adjust the drive device or stop the movement.

9. The robot intelligent control method based on image analysis algorithm according to claim 1, characterized in that: The step S5, obtaining the actual angle information of each joint of the robot and the actual position information of the end effector in real time, specifically includes: Step S5.1: Use the angle sensors installed at each joint of the robot to collect the actual rotation angle of the joint in real time and convert the angle information into a digital signal; Step S5.2: Obtain the actual position coordinates of the end effector in space in real time through the position sensor installed on the end effector of the robot; Step S5.3: Encode the collected joint angle digital signals and end effector position coordinate information to form a feedback information format for recognition by the robot's multi-dimensional posture control model.

10. A robot intelligent control system based on image analysis algorithm, characterized in that: include: An image feature recognition processing unit, which uses the optimized Single Shot Multi-Box Detector algorithm to perform multi-target feature recognition on the acquired robot working environment image data, locate target objects and feature points, and generate corresponding feature vector sets; The posture control calculation unit is connected to the image feature recognition processing unit and is used to input the feature vector set into the robot's multi-dimensional posture control model, and calculate the angle parameters that need to be adjusted for each joint of the robot based on the spatial position information of the target object in the feature vector set and the initial posture information of the robot itself; The motion planning decision unit is connected to the posture control calculation unit. It determines the robot's motion planning strategy based on the various parameters of the robot's intelligent control and the joint angle parameters output by the robot's multi-dimensional posture control model. The joint drive control unit is connected to the motion planning and decision-making unit, and controls the drive devices of each joint of the robot according to the motion planning strategy, so that each joint of the robot rotates according to the predetermined angle parameters and motion sequence; The information acquisition and feedback unit is connected to the joint drive control unit and the posture control calculation unit. During the robot's motion, it obtains the actual angle information of each joint of the robot and the actual position information of the end effector in real time and feeds it back to the posture control calculation unit. The motion strategy adjustment unit is connected to the information acquisition and feedback unit and the motion planning and decision-making unit. The robot's multi-dimensional posture control model compares and analyzes the feedback information with the preset target posture information. Through this unit, the robot's subsequent motion planning strategy is dynamically adjusted to achieve continuous and precise motion control of the robot.

Images