Self-correcting industrial robot for complex scene

Abstract

The self-correcting industrial robot for the complex scene comprises a main machine framework and a launching module, and an omnidirectional chassis transmission module, a damping module, a main body connecting frame and a launching end vertical framework ammunition supply chain are arranged between the main machine framework and the launching module. The omni-directional chassis transmission module is fixedly connected with the host framework through the damping module; aiming at composite requirements of multi-task cooperation, high-risk environment intervention, dynamic target operation and the like in an industrial scene, functional modules such as a repairing agent or a flame retardant and the like are integrated in a standardized projectile shell through a ballistic carrier design normal form; dynamic launching is achieved through the launching module, one-machine multi-energy flexible deployment is achieved, a single robot can cross a traditional function boundary, cross-domain tasks such as detection, repairing and marking are synchronously executed in a complex scene, and the problems of resource redundancy and response lag caused by function splitting of traditional equipment are solved.

Description

Technical Field

[0001] This invention relates to the field of industrial equipment technology, and in particular to a self-calibrating industrial robot for complex scenarios. Background Technology

[0002] Industrial robots operate in structured, static environments. Traditional robot perception systems and fixed trajectory planning struggle to effectively handle moving production lines, workpieces on conveyor belts, or sudden dynamic targets (such as moving obstacles). Furthermore, AGVs or wheeled robots typically operate on flat surfaces, and in confined spaces, environments with obstacles, or situations requiring fine-tuning of positioning, their movement becomes inflexible, often necessitating complex path planning and multiple adjustments. Therefore, we propose a self-correcting industrial robot for complex scenarios. Summary of the Invention

[0003] The main objective of this invention is to provide a self-correcting industrial robot for complex scenarios, aiming to systematically solve the technical problem of poor adaptability of traditional industrial robots when dealing with complex, dynamic, and multi-tasking scenarios.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A self-calibrating industrial robot for complex scenarios includes a mainframe architecture and a launching module. An omnidirectional chassis drive module, a shock-absorbing module, a main body connecting frame, and a launching end vertical architecture ammunition supply chain are disposed between the mainframe architecture and the launching module. The omnidirectional chassis drive module is fixedly connected to the mainframe architecture via the shock-absorbing module. The main body connecting frame is fixedly connected to the launching end vertical architecture ammunition supply chain. The launching modules are fixedly connected to each other, and the launching end vertical architecture ammunition supply chain is fixedly connected to the shock-absorbing module. The transmitting module and the omnidirectional chassis transmission module achieve precise movement control through a motor drive module. The motor drive module consists of a brushless motor unit, a permanent magnet stepper motor, a PID control unit, and an FOC control unit. The brushless motor unit is connected to the omnidirectional chassis transmission module, the permanent magnet stepper motor is connected to the transmitting module, and the PID control unit and the FOC control unit are signal-connected to the brushless motor unit and the permanent magnet stepper motor. The transmitting module and the omnidirectional chassis transmission module achieve obstacle avoidance adjustment by setting an environmental perception module. The environmental perception module consists of an industrial camera unit, a lidar unit, a gyroscope unit, a CAN communication unit, and a computing unit. The industrial camera unit, lidar unit, and gyroscope unit are signal-connected to the computing unit, and the computing unit is signal-connected to the CAN communication unit. The computing unit is used to filter and process the signal parameters acquired by the industrial camera unit, lidar unit, and gyroscope unit.

[0005] A further improvement of the present invention is that the launching module comprises a projectile launching module, a launching calibration module, a launching data analysis unit, and a perception vision module.

[0006] A further improvement of the present invention is that both the launch calibration module and the perception vision module are signal-connected to the launch data analysis unit. The launch calibration module and the launch data analysis unit are used to calibrate the launch angle of the projectile launch module. The launch data analysis unit and the perception vision module are used to analyze whether the execution signal sent by the launch calibration module to the projectile launch module matches the content identified by the perception vision module.

[0007] A further improvement of the present invention is that the host architecture is provided with a main controller, the main controller is provided with a serial communication unit, the serial communication unit is connected to the industrial camera unit, the lidar unit and the gyroscope unit via an RS8 bus, and the serial communication unit is connected to the CAN communication unit.

[0008] A further improvement of the present invention is that the arithmetic processing filtering method of the arithmetic unit includes: Step S1: First, perform a first-order Taylor expansion of the nonlinear system function, ignoring higher-order terms; ; ; in, for time right Jacobian matrix, for time right Jacobian matrix, for time right Jacobian matrix, for time right The Jacobian matrix, as shown above, is obtained by the following formula: ; ; ; ; Step S2: Prediction, prior state estimation: ; Step S3, verify the error covariance: ; Step S4, Correction, Kalman gain: ; Step S5: Update, update the posterior state estimate: Update the posterior error covariance: .

[0009] Compared with the prior art, the beneficial effects of the present invention are: 1. To address the complex needs of multi-task collaboration, intervention in high-risk environments, and dynamic target operation in industrial scenarios, this invention integrates functional modules such as repair agents or flame retardants into a standardized projectile shell through a ballistic carrier design paradigm. Dynamic launch is achieved through a launch module, enabling flexible deployment of a single robot with multiple functions. This allows a single robot to transcend traditional functional boundaries and simultaneously perform cross-domain tasks such as detection, repair, and marking in complex scenarios, solving the problems of resource redundancy and response lag caused by functional fragmentation in traditional equipment.

[0010] 2. To address the perception distortion and execution deviations caused by dynamic and static obstacles and multi-source interference in industrial environments, this invention constructs an environmental physical characteristic analysis network (3), combines it with a visual model detection framework (2) and an extended Kalman filter-optimized ballistic trajectory prediction model (2), and uses the YOLOv8s model to identify object weld points and the YOLOv8n model to identify objects, along with OpenCV to calculate the distance and size of target objects, and analyzes the texture, deformation, and mechanical properties of target objects in real time. The YOLOv8 visual model uses a real-time online detection visual algorithm based on neural networks and an extended Kalman filter-optimized trajectory prediction to achieve coordinated optimization of dynamic target tracking and obstacle avoidance strategies. The execution layer relies on high-precision servo control and a PID closed-loop algorithm to ensure that the projectile delivery accuracy reaches the sub-millimeter level, meeting the needs of complex industrial scenarios. Attached Figure Description

[0011] Figure 1 This is a diagram illustrating the composition of a self-calibrating industrial robot for complex scenarios according to the present invention.

[0012] Figure 2 This is a schematic diagram of the structure of a vibration damping module in a self-calibrating industrial robot for complex scenarios according to the present invention.

[0013] Figure 3 This is a schematic diagram of the launching module in a self-calibrating industrial robot for complex scenarios according to the present invention.

[0014] Figure 4 This is a schematic diagram of the omnidirectional chassis transmission module in a self-calibrating industrial robot for complex scenarios according to the present invention.

[0015] Figure 5 This is a structural schematic diagram of an omnidirectional chassis transmission module and a shock absorption module in a self-calibrating industrial robot for complex scenarios according to the present invention.

[0016] Figure 6 This is a schematic diagram of the main body connection frame in a self-calibrating industrial robot for complex scenarios according to the present invention.

[0017] Figure 7 This is a schematic diagram of the vertical structure ammunition supply chain in a self-calibrating industrial robot for complex scenarios according to the present invention.

[0018] Figure 8 This invention provides a control block for the FOC control module in a self-calibrating industrial robot for complex scenarios. Figure 1 .

[0019] Figure 9 This invention provides a control block for the FOC control module in a self-calibrating industrial robot for complex scenarios. Figure 2 .

[0020] In the diagram: 1. Mainframe architecture; 2. Launch module; 3. Omnidirectional chassis drive module; 4. Shock absorption module; 5. Main connecting frame; 6. Launcher vertical architecture ammunition supply chain. Detailed Implementation

[0021] The present invention will be further described below with reference to specific embodiments. The accompanying drawings are for illustrative purposes only and are schematic diagrams, not actual pictures. They should not be construed as limiting the present invention. In order to better illustrate the specific embodiments of the present invention, some parts in the drawings may be omitted, enlarged or reduced, and do not represent the actual product size. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings. Based on the specific embodiments of the present invention, all other specific embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Please see Figures 1-9A self-calibrating industrial robot for complex scenarios includes a mainframe architecture 1 and a launching module 2. An omnidirectional chassis transmission module 3, a shock-absorbing module 4, a main body connecting frame 5, and a launching end vertical architecture ammunition supply chain 6 are arranged between the mainframe architecture 1 and the launching module 2. The omnidirectional chassis transmission module 3 is fixedly connected to the mainframe architecture 1 through the shock-absorbing module 4. The main body connecting frame 5 is fixedly connected to the launching end vertical architecture ammunition supply chain 6. The launching modules 2 are fixedly connected to each other. The launching end vertical architecture ammunition supply chain 6 is fixedly connected to the shock-absorbing module 4. In one optional embodiment of this invention, the transmitting module 2 and the omnidirectional chassis transmission module 3 achieve precise movement control by setting a motor drive module. The motor drive module comprises a brushless motor unit, a permanent magnet stepper motor, a PID control unit, and a FOC control unit. The brushless motor unit is connected to the omnidirectional chassis transmission module 3, the permanent magnet stepper motor is connected to the transmitting module 2, and the PID control unit and the FOC control unit are signal-connected to the brushless motor unit and the permanent magnet stepper motor. Please refer to the accompanying drawings in the instruction manual. Figure 8 Among them, FOC control features high control precision, enabling rapid dynamic response and efficient torque control. This precision allows FOC to maintain stable performance under large load variations and high-frequency operation. Figure 8 In this context, Iq_Ref is the q-axis (quadrature axis) current setting value, and Id_Ref is the d-axis (direct axis) current setting value. Ia, Ib, and Ic are the sampled currents of phases A, B, and C, respectively, which can be directly obtained through AD sampling. Usually, two phases are sampled directly, and the third phase is calculated using the formula Ia+Ib+Ic=0. The electrical angle θ can be calculated by reading the value of the magnetic encoder in real time.

[0023] After obtaining the three-phase currents and electrical angles, the current loop can be executed: the three-phase currents Ia, Ib, and Ic are transformed by Clark to obtain Iα and Iβ; then transformed by Park to obtain Iq and Id; then the error values ​​are calculated with their set values ​​Iq_Ref and Id_Ref respectively; then the q-axis current error value is substituted into the q-axis current PI loop to calculate Vq, and the d-axis current error value is substituted into the d-axis current PI loop to calculate Vd; then Vq and Vd are transformed by the inverse Park to obtain Vα and Vβ; then Va, Vb, and Vc are obtained through the SVPWM algorithm, and finally input to the three phases of the motor. This completes one current loop control cycle.

[0024] When performing speed control on a PMSM, a speed loop needs to be added outside the current loop. Please refer to the attached diagram in the instruction manual. Figure 9The control block diagram shows that Speed_Ref is the speed setpoint and ω is the motor speed feedback, which can be calculated by the motor encoder.

[0025] The calculated motor speed ω is compared with the speed setpoint Speed_Ref to calculate the error value, which is then substituted into the speed PI loop. The result is used as the input to the current loop. Figure 2 and Figure 1 The current loop section can be found Figure 2 The d-axis current is set to zero (Id_Ref=0) because the d-axis current does not generate output force for the rotation of the drive motor, so the d-axis current is usually set to zero (but not always). When Id_Ref=0, Iq_Ref is equal to the output of the speed loop. Combined with the current loop above, the dual closed-loop control of speed and current is realized.

[0026] In an optional embodiment of this invention, the transmitting module 2 and the omnidirectional chassis transmission module 3 achieve obstacle avoidance adjustment by setting an environmental perception module. The environmental perception module comprises an industrial camera unit, a lidar unit, a gyroscope unit, a CAN communication unit, and a computing unit. The industrial camera unit, lidar unit, and gyroscope unit are signal-connected to the computing unit, and the computing unit is signal-connected to the CAN communication unit. The computing unit is used to filter and process the signal parameters acquired by the industrial camera unit, lidar unit, and gyroscope unit. The environmental perception module in this invention is equipped with a high-resolution industrial camera to acquire visual information about the environment. The images captured by the camera are analyzed by image processing algorithms to identify target objects (such as obstacles, markers, other robots, etc.), extract image features, and perform visual positioning. To improve adaptability under different lighting conditions, the camera may also be equipped with automatic exposure and white balance functions. Furthermore, multi-camera systems (such as binocular vision) can be used to acquire depth information and enhance environmental perception. Based on this data, the present invention can quickly construct an environmental model through a computing unit and calculate an optimal or near-optimal path from the current position to the target position using any of the path planning algorithms, such as A*, Dijkstra's algorithm, or RRT (Rapidly-exploring Random Tree). The path planning algorithm considers the robot's size, mobility, and various constraints in the environment, such as collision avoidance, obstacle bypass, and adherence to specific driving rules. The lidar unit provides high-frequency real-time environmental data updates, enabling the present invention to dynamically adjust its path planning to adapt to constantly changing conditions in industrial environments. For example, when a moving obstacle suddenly appears in the robot's path, the lidar can quickly detect the obstacle and feed the information back to the path planning module, allowing the robot to adjust its route in time to avoid collisions. The lidar unit is another important environmental perception sensor; it acquires distance information about the surrounding environment by emitting a laser beam and measuring the time it takes for the beam to reflect back. LiDAR can generate high-precision 3D point cloud maps for building environmental models, obstacle detection, and path planning. Industrial environments often contain various moving or stationary obstacles, such as material boxes, shelves, workers, and dynamically changing production line layouts. In this invention, the computing unit utilizes LiDAR to quickly scan the surrounding environment, generate real-time 3D point cloud maps, and accurately identify the location and size of these obstacles.

[0027] In an optional implementation of this embodiment, after the robot of the present invention circles an unfamiliar map, it can automatically generate a global map by scanning and collecting information through LiDAR. After indicating the destination reached by the robot on the map, it will automatically generate the optimal path, and during the movement, it will achieve automatic obstacle avoidance through a local path planning algorithm. At the same time, based on the LiDAR information, the A* algorithm is used to set the position coordinates (x, y) of the grid cells, initialize the evaluation function f(n) to zero, generate the Open table and Close table, and select the node with the lowest cost for expansion by comprehensively evaluating the cost of each expansion node so that it can quickly guide the robot to the target point.

[0028] f(n) = g(n) + h(n) Where f(n) is the evaluation function, g(n) represents the actual cost incurred from the starting point to the current point, and h(n) represents the estimated cost incurred from the current point to the target point. In the A* algorithm, the selection of the heuristic function is crucial. Given a grid-based map environment and the presence of obstacles, the Manhattan distance is used as the heuristic function. f(n) = g(n) + (dx - nx) + (dy - ny) The choice of h(n) is extremely important when g(n) is fixed. The closer the current expansion point is to the target point, the smaller h is, and the smaller f is. This results in more nodes being searched and lower efficiency. Conversely, the larger f is, the higher the efficiency, but the harder it is to find the optimal path.

[0029] In this embodiment, the main mechanical structure of the invention consists of a main frame 1, an omnidirectional chassis transmission module 3, and a shock absorption module 4. The omnidirectional chassis transmission module 3 uses Mecanum wheel sets as the carrier for direct relative motion with the ground. A total of four 3508 motors are installed in the omnidirectional chassis transmission module 3 for driving, and each wheel set is equipped with an independent shock absorption module 4. Four D154-16 type embedded Mecanum wheels serve as the omnidirectional drive carrier. The rollers are arranged in a compact 45° angled layout to achieve multi-directional friction decomposition. Combined with modular independent drive units (each wheel equipped with one 3508 motor), it fully adapts to industrial scenario requirements. The 3508 motor has a built-in 19:1 reduction gearbox, providing a continuous stable torque output of 2.8 Nm (peak 15 Nm). The main frame 1 is constructed using 6061 aluminum tubing in a grid pattern and is integrally connected to the Mecanum wheel sets in the omnidirectional chassis transmission module 3 to ensure the integrity and stability of the chassis. Launch module 2 consists of three mechanisms: a feeding system, a multi-degree-of-freedom gimbal driven by yaw and pitch axes, and a projectile launching module. The projectile launching module has a magazine at its bottom, mounted on the chassis frame. A feeding dial is located at the bottom of the magazine and is driven by a 3508 motor to feed the projectile. The feeding dial has a vertical feeding chain composed of bearings, ensuring unobstructed projectile delivery to the launch port of the projectile launching module. This constitutes a complete feeding system. The yaw axis uses a Unisoc GO motor in conjunction with a synchronous belt module to achieve 360-degree omnidirectional rotation of the gimbal. The pitch axis uses a servo motor to drive a lead screw to control the pitch angle of the gimbal, and based on the characteristics of the lead screw, it can be locked at any specific angle. Both work together to drive the gimbal to perform multi-angle, multi-directional operations at the same point.

[0030] In one optional embodiment of this example, the launching module 2 comprises a projectile launching module, a launching calibration module, a launching data analysis unit, and a perception vision module.

[0031] In one optional embodiment of this example, both the launch calibration module and the perception vision module are signal-connected to the launch data analysis unit. The launch calibration module and the launch data analysis unit are used to calibrate the launch angle of the projectile launch module. The launch data analysis unit and the perception vision module are used to analyze whether the execution signal sent by the current launch calibration module to the projectile launch module matches the content identified by the perception vision module.

[0032] In this embodiment, ensuring the projectile accurately hits the target during dynamic operations is crucial. Therefore, high-precision sensors are needed to monitor the robot and projectile's status in real time throughout the entire movement. The launch calibration module includes an accelerometer and a gyroscope to measure the acceleration and angular velocity changes experienced by the projectile after launch. By collecting and processing this real-time data, the launch data analysis unit uses a fuzzy PID algorithm to accurately calculate the projectile's motion, eliminating noise interference and updating the projectile's position and velocity estimates at each time step. Continuous analysis of the trajectory data identifies any external factors that may affect the trajectory, including wind speed, changes in direction, and even minor fluctuations in gravity. After calculating the deviation between the projectile's actual trajectory and the expected trajectory, the launch calibration module adjusts the launch angle and power settings of the projectile launch module to correct the trajectory. Furthermore, the dynamic self-correction mechanism improves the accuracy of projectile launch, ensuring that the invention can efficiently and quickly identify problems in industrial production during operation. The efficient execution of this process enables the robot to maintain target orientation and accurately hit the designated target in a changing environment.

[0033] In this embodiment, the vision scheme of the perception vision module is divided into two parts. One part is to train the dataset in the task recognition module that the robot needs to perform using the yolov8s model in the YOLOv8 model. The dataset adopts the selection of multiple samples, complexity and multiple sites, aiming to achieve high-precision recognition in different environments. Therefore, the training of the model adopts multi-round training recognition on datasets obtained from various industrial products in diverse scene training, such as exposure, highlight, natural light, low light source and dark environment.

[0034] The other part is the robot's auto-aiming vision program, which comprises two main components: first, the yolov8n model within the YOLOv8 framework; and second, OpenCV, used to identify and measure the distance between the target object and the robot. The yolov8n model provides high-precision recognition, while OpenCV provides high-precision distance measurement, ensuring the accuracy of the robot's auto-aiming. Robot auto-aiming requires real-time feedback, therefore, yolov8n's real-time feedback characteristics better meet the needs of the application scenario. The dataset used to train the yolov8n model uses images with multiple objects, artificially increasing noise to ensure better robustness of the yolov8n model under various environmental factors. OpenCV can measure the distance between the robot and the target object, as well as the distance between the target object and other objects, ensuring that the robot can target specific objects during operation. Furthermore, OpenCV can measure the area of ​​the target object, allowing for the consideration of applied forces during robot operation, making the robot's operation a dynamic process.

[0035] In one optional embodiment of this example, the host architecture 1 is provided with a main controller, which is provided with a serial communication unit. The serial communication unit is connected to the industrial camera unit, the lidar unit and the gyroscope unit via an RS485 bus, and is also connected to the CAN communication unit.

[0036] In this embodiment, the perception vision module visually integrates extended Kalman filtering, OpenCV measurements, and hardware error values. This extended Kalman filtering can be used to optimize and predict the position of the object being measured in the next moment. In certain special scenarios, where the object is in real-time motion, extended Kalman filtering allows the robot to complete the task at minimal cost.

[0037] In an optional embodiment of this example, the arithmetic processing filtering method of the arithmetic unit includes: Step S1: First, perform a first-order Taylor expansion of the nonlinear system function, ignoring higher-order terms; ; ; in, for time right Jacobian matrix, for time right Jacobian matrix, for time right Jacobian matrix, for time right The Jacobian matrix, as shown above, is obtained by the following formula: ; ; ; ; Step S2: Prediction, prior state estimation: ; Step S3, verify the error covariance: ; Step S4, Correction, Kalman gain: ; Step S5: Update, update the posterior state estimate: Update the posterior error covariance: .

[0038] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A self-correcting industrial robot for complex scenarios, comprising a main frame architecture (1) and a launch module (2), characterized in that, The host architecture (1) and the launch module (2) are provided with an omnidirectional chassis transmission module (3), a damping module (4), a main body connecting frame (5) and a launch end vertical architecture for a belt (6), the omnidirectional chassis transmission module (3) is fixedly connected with the host architecture (1) through the damping module (4), the main body connecting frame (5) is fixedly connected with the launch end vertical architecture for a belt (6), the launch module (2) is fixedly connected with the launch module (2), and the launch end vertical architecture for a belt (6) is fixedly connected with the damping module (4); The launch module (2) and the omnidirectional chassis transmission module (3) are precisely moved and controlled by being provided with a motor drive module, the motor drive module comprises a brushless motor unit, a permanent magnet type stepping motor, a PID control unit and a FOC control unit, wherein the brushless motor unit is connected with the omnidirectional chassis transmission module (3), the permanent magnet type stepping motor is connected with the launch module (2), and the PID control unit and the FOC control unit are signal connected with the brushless motor unit and the permanent magnet type stepping motor; The launch module (2) and the omnidirectional chassis transmission module (3) are adjusted to avoid obstacles by being provided with an environment sensing module, the environment sensing module comprises an industrial camera unit, a laser radar unit, a gyroscope unit, a CAN communication unit and an operation unit, the industrial camera unit, the laser radar unit and the gyroscope unit are signal connected with the operation unit, the operation unit is signal connected with the CAN communication unit, and the operation unit is used for filtering the signal parameters acquired by the industrial camera unit, the laser radar unit and the gyroscope unit.

2. A self-correcting industrial robot for complex scenarios as claimed in claim 1, wherein: The launch module (2) comprises a projectile launch module, a launch calibration module, a launch data analysis unit and a sensing vision module.

3. A self-correcting industrial robot for complex scenarios as claimed in claim 2 wherein: The launch calibration module and the sensing vision module are signal connected with the launch data analysis unit, the launch calibration module and the launch data analysis unit are used for calibrating the launch angle of the projectile launch module, and the launch data analysis unit and the sensing vision module are used for analyzing whether the execution signal emitted by the launch calibration module to the projectile launch module composes the identification content of the sensing vision module.

4. A self-correcting industrial robot for complex scenarios as claimed in claim 3 wherein: A main controller is arranged in the host architecture (1), the main controller is provided with a serial communication unit, the serial communication unit is signal connected with the industrial camera unit, the laser radar unit and the gyroscope unit through an RS485 bus, and the serial communication unit is signal connected with the CAN communication unit.

5. A self-correcting industrial robot for complex scenarios as claimed in claim 1, wherein: The operation processing filtering method of the operation unit comprises: Step S1, first, a first-order Taylor expansion of a nonlinear system function is carried out, and high-order terms are ignored; Step S2, the first-order Taylor expansion is differentiated to obtain a first-order derivative of the nonlinear system function; ; ； wherein is the time for the Jacobian matrix, is the time for the Jacobian matrix, is the time for the Jacobian matrix, is the time for the Jacobian matrix, the above matrix is obtained from the following calculation formula: ； ； ； ; Step S2, prediction, prior state estimate: ; Step S3, check error covariance: ; Step S4, correction, Kalman gain: ; Step S5, update, update the posterior state estimate: , update the posterior error covariance: .

Images