An automatic material handling mechanism for cable processing equipment

Abstract

This invention relates to the field of cable processing technology, specifically to an automatic material handling mechanism for cable processing equipment, comprising: an incoming material aligning and conveying unit for flattening scattered cables and conveying them via a conveyor belt synchronized with an encoder; an industrial vision or AI vision unit for visually inspecting and measuring the cables during conveying; a material handling execution unit, including an actuator with X-Y-θ and Z degrees of freedom and a dual-modal compliant gripper; an attitude correction and guiding unit for correcting the end direction of the grasped cables and standardizing the loading attitude; and a control and synchronization unit for spatiotemporal synchronization with the conveyor encoder. The purpose of this invention is to provide an automatic material handling mechanism for cable processing equipment, which, by combining the incoming material aligning and conveying unit with an industrial vision or AI vision unit, effectively overcomes the shortcomings of low efficiency and easy omissions in separating flexible cables using vibration-stop photography or general handling systems.

Description

Technical Field

[0001] This invention relates to the field of cable processing technology, and in particular to an automatic material handling mechanism for cable processing equipment. Background Technology

[0002] In existing automated production, vision-based pick-and-place systems can perform high-speed pick-and-place operations on workpieces on the production line using camera-based positioning. However, these systems are mostly applied to regular workpieces and are not well-suited for slender, easily tangled, and easily deformable targets such as flexible cables. For example, Chinese patent CN114538088A, entitled "High-speed and high-precision pick-and-place method based on camera-based imaging," demonstrates the efficiency advantage of "camera-based imaging" in production lines by using a camera to take high-speed pictures along a set path and achieve high-speed, wait-free pick-and-place operations. However, it does not propose specific solutions for end recognition, tangling judgment, and compliant grasping of flexible cables.

[0003] In addition, existing vision-guided material handling modules generally adopt an architecture of "vision system + electric screw / robotic arm + material handling head", which can achieve precise positioning and gripping of workpieces on the production line. However, the attitude estimation, single-piece separation and dynamic gripping robustness of slender and flexible targets are still limited, and problems such as secondary driving, mis-grabbing and damage are prone to occur (e.g., CN109319477A). For flexible feeders with scattered feeding, existing solutions use the first vision device to stop vibration and take pictures to locate materials that meet the gripping posture and drive the robotic arm to grab them, which solves the problem of scattered picking to a certain extent. However, there is still a lack of systematic solutions for "cable end positioning, entanglement recognition and online flying shooting synchronous gripping" (e.g., CN113104531A).

[0004] In summary, cable processing scenarios place higher demands on "single wire separation, precise end positioning, winding / overlapping judgment, gripping that balances flexibility and stability, and real-time synchronization with the conveying cycle (flying camera)". Existing general vision pick-and-place equipment is still difficult to stably adapt to flexible cables under high cycle time and high yield. Summary of the Invention

[0005] The purpose of this invention is to propose an automatic material handling mechanism for cable processing equipment, targeting short cable segments after cutting and incoming material segments before terminal crimping, etc., to achieve: end-level positioning based on industrial vision / AI vision, intelligent judgment of tangling / overlapping, dynamic flying and synchronous picking and placing, dual-modal compliant gripping and online posture correction, thereby improving the single-piece separation rate, picking and placing success rate and cycle time adaptability of flexible cables.

[0006] To achieve the above objectives, the present invention provides an automatic material handling mechanism for cable processing equipment, comprising: an incoming material aligning and conveying unit for flattening scattered cables and conveying them with a conveyor belt synchronized by an encoder;

[0007] Industrial vision or AI vision units are used to perform visual inspection and measurement of cables during the conveying process, outputting instance masks, end positions, spindle posture, tangling / overlapping judgment and single-piece gripping score for a single cable, and completing aerial photography and gripping pose prediction without stopping the cable.

[0008] The material handling execution unit includes an actuator with XY-θ and Z degrees of freedom and a dual-modal compliant gripper. The gripper includes a controllable negative pressure adsorption module and an adaptive finger clamp module, and has force / slip sensing capabilities.

[0009] The attitude correction and guidance unit is used to correct the end direction of the grasped cable and standardize the feeding posture.

[0010] The control and synchronization unit synchronizes with the conveyor encoder in time and space to complete the flying camera trigger, trajectory planning, target allocation, and abnormal component diversion.

[0011] The industrial vision or AI vision unit works in conjunction with the control and synchronization unit to enable the actuator to dynamically grasp and place the cable without stopping the conveyor belt.

[0012] Preferably, the industrial vision or AI vision unit includes a top-mounted high-speed camera and a strobe light source, which performs aerial photography with short exposures to suppress motion blur.

[0013] Preferably, the industrial vision or AI vision unit further includes a 3D depth camera for performing height verification of cable overlap and multi-layer stacking, and adjusting the individual piece graspability score accordingly.

[0014] Preferably, the end detection is obtained by skeletonization after instance segmentation and extraction of high curvature end features, and the principal axis angle is calculated using PCA.

[0015] Preferably, the dual-modal compliant gripper switches to finger clamp closure after the end is lifted by adsorption to achieve stable transport, and triggers secondary clamping or switches to mid-section gripping based on slip sensing.

[0016] Preferably, the control and synchronization unit calculates the target displacement Δx=v·Δt based on the conveyor belt speed v and the expected gripping time Δt, and performs micro-position correction within 10–20 ms before the end reaches the target position.

[0017] Preferably, the entanglement / overlap determination is based on the fusion determination of the closed-loop / multi-intersection topology of the 2D instance skeleton and the 3D local height statistics. When the confidence threshold is exceeded, a decoupling action is performed, and if it fails, it is diverted to the return column.

[0018] Preferably, the attitude correction and guiding unit includes a guide groove and a passively rotating guide wheel to achieve secondary correction of the end direction and axial alignment.

[0019] Preferably, the light source includes switchable visible and near-infrared light sources to adapt to black or low-contrast cable surfaces.

[0020] Preferably, the target allocation uses a minimum cost matching algorithm to maximize throughput per unit time, and the cost function integrates target arrival time, executor reachability, and grasping confidence.

[0021] The beneficial effects of this invention are:

[0022] 1. By setting up a material sorting and conveying unit combined with an industrial vision or AI vision unit, and using a servo-driven conveyor belt synchronized with an encoder and a top-mounted high-speed camera and strobe light source for aerial photography acquisition system, continuous sorting, high-speed visual perception and online accurate detection of scattered cables can be achieved. This effectively overcomes the shortcomings of low efficiency and easy omission in separating flexible cables by vibration-stop photography or general pick-and-place systems. It realizes cable instance segmentation, end positioning, attitude estimation and tangling identification under high cycle time, which significantly improves the single-piece separation rate and production rhythm.

[0023] 2. By setting up a dual-modal compliant gripper with force / slip sensing and integrating an adaptive impedance control algorithm, combined with entanglement criteria based on 2D instance skeleton topology and 3D local height statistics, the system achieves compliant gripping, slip monitoring, and adaptive adjustment of the gripping strategy for cables. This solves the problem that rigid gripping can easily lead to cable damage, secondary driving, or mis-gripping. It ensures gripping reliability and avoids crushing or scratching the cable surface, significantly improving the gripping success rate and product quality.

[0024] 3. By setting the control and synchronization unit to synchronize with the encoder in time and space, integrating Kalman filter pose prediction, timing micro-correction and minimum cost target allocation algorithms, and cooperating with the V-groove and passive rotary guide wheel in the attitude correction unit, dynamic cable tracking and grasping, end orientation correction and rapid attitude standardization are achieved. This overcomes the problems of poor system cycle time adaptability and positioning being easily interfered with, and realizes high-precision grasping, unloading and abnormal diversion under non-stop conditions, comprehensively improving system throughput and overall automation level. Attached Figure Description

[0025] Figure 1 System overall structure diagram (conveying unit, vision unit, material handling execution unit, attitude correction unit, control unit);

[0026] Figure 2 Visual perception and grasping planning flowchart;

[0027] Figure 3 Schematic diagram of a dual-modal compliant gripper (adsorption module, finger clamp module, force / slip sensor);

[0028] Figure 4Schematic diagram of aerial photography timing and pose prediction (synchronized with encoder);

[0029] Figure 5 Schematic diagram of entanglement / overlap determination and single-piece separation strategy. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0031] This invention provides, for example Figures 1 to 5 An automatic feeding mechanism for a cable processing equipment is shown, comprising:

[0032] The incoming material alignment and conveying unit is used to flatten loose cables and convey them with an encoder-synchronized conveyor belt.

[0033] Industrial vision or AI vision units are used to perform visual inspection and measurement of cables during the conveying process, outputting instance masks, end positions, spindle posture, tangling / overlapping judgment and single-piece gripping score for a single cable, and completing aerial photography and gripping pose prediction without stopping the cable.

[0034] The material handling unit includes an actuator with XY-θ and Z degrees of freedom and a dual-modal compliant gripper. The gripper includes a controllable negative pressure adsorption module and an adaptive finger clamp module, and is equipped with force / slip sensing.

[0035] The attitude correction and guidance unit is used to correct the end direction of the grasped cable and standardize the feeding posture.

[0036] The control and synchronization unit synchronizes with the conveyor encoder in time and space to complete the flying camera trigger, trajectory planning, target allocation, and abnormal component diversion.

[0037] Among them, the industrial vision or AI vision unit works in conjunction with the control and synchronization unit to enable the actuator to dynamically grasp and place the cable without stopping the conveyor belt.

[0038] The incoming material alignment and conveying unit uses a set of differential rollers to initially flatten and disperse the stacked, scattered cables. These cables are then conveyed by a synchronous belt driven by a servo motor, whose encoder signal is connected to the control system. The industrial vision unit uses a 2-megapixel high-speed CMOS camera with an LED strobe light source to continuously photograph the cables while the conveyor belt is running. The acquired images are segmented using a built-in deep learning algorithm (such as Mask R-CNN) to generate pixel-level masks for each cable. Then, through skeleton extraction and end curvature analysis, the end coordinates and spindle angle are determined, and the deviation from the ideal gripping posture and whether it is entangled with other cables are calculated. The material handling execution unit uses a four-axis SCARA robot with a dual-modal gripper at its end. This gripper integrates a vacuum suction cup (controlled by a solenoid valve) and a pair of compliant pneumatic fingers. The finger gripper module incorporates a piezoresistive force sensor and a vibration sensor-based slip detection module at the fingertip. The posture correction and guidance unit, located at the end of the conveyor belt, consists of a V-shaped guide groove and a pair of freely rotating rubber wheels. When the robot arm feeds the cable end into this area, the cable automatically centers and corrects its axial posture under the constraint of the V-shaped groove and the rollers. The control and synchronization unit uses an industrial PC that communicates with the robot controller, vision system, and conveyor encoder via a real-time Ethernet bus. The position signal provided by the encoder is used to trigger the fly-through and predict the cable's trajectory. The industrial PC integrates the gripping point pose, the robot's current state, and the conveyor belt speed output by the vision system to perform real-time motion planning, directing the robot to dynamically track and grip the cable without stopping the conveyor belt. For abnormal cables that are determined by the vision system to be tangled or ungripable, a planned path is used to throw them into the bypass diversion box.

[0039] Furthermore, the industrial vision or AI vision unit includes a top-mounted high-speed camera and a strobe light source, which captures images with short exposures to suppress motion blur. The top-mounted high-speed camera uses an exposure time of 1 / 10000 second, and the LED strobe light source, which is synchronized with the camera's frame rate, provides high-intensity illumination at the moment of exposure, thereby completely suppressing motion blur generated by the cable during high-speed transport and ensuring that clear images are obtained for high-precision visual measurement and recognition.

[0040] Furthermore, the industrial vision or AI vision unit also includes a 3D depth camera, which is used to perform height verification of cable overlap and multi-layer stacking, and adjust the individual piece graspability score accordingly. After the initial 2D vision localization, the camera scans the cable stacking area and calculates the height variance of the local area of ​​the cable through point cloud data. If a significant height difference is detected, it is determined to be multi-layer stacking, thereby lowering the original high confidence "graspable" score based on 2D images, or triggering the robot to try different grasping strategies (such as increasing the suction force or prioritizing the grasping of the top layer cable).

[0041] Furthermore, end detection is obtained through skeletonization and high curvature end feature extraction after instance segmentation, and the principal axis angle is calculated using PCA. First, a binary mask image of the cable is obtained using a trained instance segmentation model. Then, morphological thinning is performed on the mask to obtain a skeleton line with a width of one pixel. The curvature is then calculated along the skeleton line, and the point with the maximum curvature is identified as the end point of the cable. Finally, principal component analysis (PCA) is performed on the mask region, and the direction of the first principal component is regarded as the principal axis angle of the cable for subsequent grasping posture planning.

[0042] Furthermore, the dual-modal compliant gripper switches to finger clamp closure after lifting the cable end to achieve stable handling. Based on slip detection, it triggers secondary clamping or switches to mid-section gripping. The robot first positions itself above the cable end, descends in the Z direction, and brings the negative pressure adsorption module (vacuum suction cup) into contact with the cable surface. After activating the vacuum generator to suck up the cable, it lifts it up, which aims to separate any overlapping cables. After lifting to a certain height, the pneumatic fingers of the adaptive finger clamp module close under air pressure, gripping the middle of the cable from both sides. During handling, if the force sensor on the finger clamp detects a sudden drop in gripping force or the vibration sensor detects high-frequency micro-movements, slippage is determined to have occurred. The control system immediately instructs the finger clamp to increase the gripping force or triggers the robot to stop the current action and move to the middle of the cable for secondary gripping.

[0043] Furthermore, the control and synchronization unit calculates the target displacement Δx = v·Δt based on the conveyor belt speed v and the expected gripping time Δt, and performs micro-position correction within 10–20 ms before the end effector reaches its position. The system continuously monitors the conveyor belt encoder signal and calculates the position of the target point on the cable in real time. Assuming the conveyor belt speed is 500 mm / s and the vision processing and robot motion planning take 100 ms, the predicted displacement of the target point Δx = 50 mm. During the tracking motion, the robot continuously receives visual feedback. In the last 15 ms before the end effector is about to contact the cable, it performs XY translation within one millimeter or θ rotation correction within degrees based on the small deviation between the actual and predicted positions of the cable calculated by the latest vision, to ensure accurate gripping.

[0044] Furthermore, the entanglement / overlap determination is based on the fusion of the closed-loop / multi-intersection topology of the 2D instance skeleton and the 3D local height statistics. When the confidence threshold is exceeded, a decoupling action is performed. If the decoupling fails, the component is diverted to the rework unit. In the 2D instance skeleton analysis, if a skeleton is found to form a closed loop or has more than two intersections with other skeletons, it is marked as a suspected entanglement. At the same time, the 3D depth camera samples the height of the area. If multiple high points are displayed and the height difference is greater than the cable diameter (e.g., 2mm), overlap is confirmed. When the weighted sum of the confidence of the two criteria exceeds the threshold (e.g., 0.9), it is determined to be an entanglement / overlap. Subsequently, the robot attempts to perform a predefined decoupling action such as pullback or lifting. If the component is still not separated after three attempts, it is marked as an abnormal component and placed by the robot into the rework box and sent back to the incoming material rework unit.

[0045] Furthermore, the attitude correction and guidance unit includes a guide groove and passive rotating guide wheels to achieve secondary correction of the end direction and axial alignment. A fixedly installed hard anodized aluminum V-groove serves as the main guiding mechanism, with a polyurethane passive rotating wheel that can swing within a small angle installed on each side of its outlet end. The robot arm feeds the gripped cable end into the V-groove in a certain posture. As the cable slides down the V-groove, its end is mechanically aligned. Subsequently, its main body passes between the pair of passive rotating wheels. The rotation and swing of the wheels further eliminate the lateral bending and twisting of the cable, ensuring that it enters the subsequent processing equipment in a standard straight posture.

[0046] Furthermore, the light source includes switchable visible and near-infrared light sources to adapt to black or low-contrast cable surfaces. The vision system is equipped with a composite light source module with two sets of LEDs embedded in it: one set of blue visible light with a wavelength of 460nm and the other set of near-infrared light with a wavelength of 850nm. When processing black rubber cables or cables with strong surface reflection, the system switches to the near-infrared light source. Since the cable material and the background have a greater difference in absorption / reflectivity of near-infrared light, the image feature contrast can be significantly enhanced, thereby obtaining a clearer image and improving detection accuracy.

[0047] Furthermore, the target allocation employs a minimum cost matching algorithm to maximize throughput per unit time. The cost function integrates target arrival time, actuator reachability, and grasping confidence. In the embodiment of the minimum cost matching algorithm for target allocation, the cost function is C = α·(T_arrival) + β·(D_reachable) + γ·(1-C_confidence), where T_arrival represents the time it takes for the target cable to reach the robot's graspable area, D_reachable represents the joint space distance required for the robot's end effector to move from its current position to the target point, C_confidence represents the grasping confidence score given by vision, and α, β, and γ are weighting coefficients adjusted according to actual production efficiency requirements. Every 100ms, the system calculates the cost of all graspable cables within the field of view and allocates the target with the minimum cost to the robot, thereby dynamically deciding the grasping order and maximizing system throughput.

[0048] An automatic material handling mechanism is described in a method for operating the automatic material handling mechanism of a cable processing equipment as described above, comprising:

[0049] S1: A visible tape stream is formed by synchronously conveying cables using an encoder;

[0050] S2: Perform aerial photography to collect data and output instance masks, key points at the ends, spindle angles, entanglement / overlap determination, and graspability scores;

[0051] S3: Pose prediction and trajectory planning based on the calibration relationship between the camera and the conveyor belt;

[0052] S4: The gripping is completed using a dual-modal compliant gripping strategy of tip adsorption lifting and finger clamp closing;

[0053] S5: The end direction and axis are aligned within the attitude correction and guidance unit and verified by secondary vision.

[0054] S6: Decouple tangled / overlapping samples first; if this fails, return to the whole column.

[0055] In S2, a composite detection method is used for the end face, which combines high curvature features and strong contrast features of the cut, and multi-layer stacking is eliminated using a 3D height threshold.

[0056] S3 uses Kalman filtering to denoise the target pose prediction and performs timing micro-corrections before the end effector reaches its target position.

[0057] The force control in S4 uses impedance control F=kx+b·v, and the clamping force is adaptively adjusted according to the wire diameter and material to avoid damage.

[0058] A computer-readable storage medium storing a program that, when executed by a processor, causes a device to perform the method steps described above.

[0059] In stage S1, the servo motor drives the conveyor belt to operate continuously, and its built-in encoder provides real-time position signals. These signals are collected by the control system and used as the time reference for the entire system, ensuring that the speed of the cable flow is stable and predictable, laying a synchronous foundation for subsequent dynamic visual inspection and grasping. In stage S2, while the conveyor belt is running continuously, the encoder signal triggers the top-mounted high-speed camera to take snapshots, freezing the images using short exposures and strobe illumination. The acquired images are immediately sent to the visual algorithm for processing. A pre-trained instance segmentation model generates a precise pixel-level mask for each cable. Based on the mask skeletonization, high curvature points are found, and combined with the features of the metal core or cuts (which form a strong color or texture contrast with the cable body) that are usually present at the cable end, composite localization is performed, which greatly improves the robustness of end detection. At the same time, 2D instance mask analysis (such as topological structure to determine entanglement) and height information collected by 3D depth camera are integrated. The height variance of the local area is calculated and a physical threshold is set. To reliably identify multi-layer overlap (e.g., exceeding 1.5 times the diameter of a single cable), the system ultimately outputs complete status information for each cable. In stage S3, the system first utilizes the transformation relationship between the camera coordinate system and the conveyor belt coordinate system (robot coordinate system) obtained through hand-eye calibration to convert the visually recognized cable image coordinates into world coordinates operable by the robot. Next, considering the uncertainty of the target pose caused by the continuous movement of the conveyor belt, a Kalman filter is used to predict and estimate the motion of this world coordinate. By fusing belt speed and historical position data, visual measurement noise and jitter are filtered out to obtain a smoother and more accurate future position prediction value. Based on this prediction value, the robot performs dynamic grasping trajectory planning. Within a very short time window (e.g., 10-20ms) before its end effector finally reaches the target point, it receives the latest actual position of the cable from the vision system and performs a real-time pose offset correction at the micrometer / milliradian level to ensure that the grasper can accurately align with the target.

[0060] In stage S4, the material handling actuator moves to the cable end according to the planned trajectory. Its dual-modal gripper first activates the negative pressure adsorption module (vacuum suction cup), applying a certain contact force to the cable end and establishing negative pressure. It then lifts vertically upwards; this action aims to initially separate any slight overlap. After adsorption and lifting, the gripper switches modes, and the adaptive finger clamp module (usually a pneumatically driven, compliant finger) begins to close. This closing process is not simple position control but force control, specifically using the impedance control model F=kx+b·v, where F is the target contact force, x is the position deviation, v is the velocity, and k and b are the stiffness and damping coefficients, respectively. This model allows the finger to compliantly conform to the cable surface and apply a stable gripping force, like a spring-damped system, rather than a rigid impact. Simultaneously, the system's preset reference gripping force is determined based on the estimated cable diameter obtained from visual measurements and a cable material database (e.g., different friction coefficients for PVC, rubber, and silicone). The robot adaptively queries and adjusts the clamping force (including compressive strength), for example, using less force for thin or flexible cables and increasing the clamping force for thicker cables. The core objective is to ensure reliable gripping while absolutely avoiding damage to the cable. In stage S5, the successfully gripped cable is transferred by the robot to a fixed posture correction and guidance unit, which typically consists of a V-groove and a pair of freely rotatable guide wheels. The robot guides the cable end into the V-groove, automatically aligning it under the constraint of its inclined plane. Simultaneously, as the cable body passes through the guide wheels, its own movement drives the wheels to rotate, thereby eliminating axial torsion and achieving standardized alignment of the end direction with the overall axis. To ensure 100% correct loading posture, a fixed secondary verification vision sensor (such as a set of photoelectric sensors or a small camera) is usually set near the exit of the guidance unit to quickly confirm the final posture of the cable. Only cables that pass the verification are placed into the next processing station.

[0061] In stage S6, for cable samples identified as severely tangled or overlapping by the vision system in S2, the control system will prioritize processing them: the robot will execute a specific set of decoupling actions, such as using a gripper to pick up one end of the suspected tangled cable and performing a lifting, shaking, or pulling motion along a specific trajectory to attempt to separate it from surrounding cables; if after several (e.g., 2-3) predefined decoupling attempts, the vision system detects that the cable is still tangled or the gripper reports a failure to grasp it, then the sample is determined to be an undecoupling anomaly. The robot will then divert it to a dedicated rework box, where it will be manually or otherwise returned to the receiving area. The beginning of the entire unit is reorganized to ensure the purity and graspability of the material on the main conveyor. Finally, the entire workflow, from synchronous conveying, fly-by detection, pose prediction, compliant grasping, attitude correction to anomaly handling, and all its algorithm details (including Kalman filter prediction, impedance control parameters, visual recognition models, etc.), are written and packaged into an executable software program and stored in a computer-readable storage medium (such as an industrial PC's solid-state drive, EEPROM, or optical disc). When this program is executed by the processor in the device (such as an industrial CPU), it can drive the entire material handling mechanism to automatically perform all the above functions.

[0062] The dual-modal compliant gripping strategy described in S4 has the following specific implementation method for its force control algorithm: The force control of the adaptive finger clamp module adopts a position-based impedance control model, which indirectly achieves the desired contact force by adjusting the finger clamp's closing position. The controller calculates the target clamping force in real time. Actual contact force Deviation between This deviation is then mapped to a position correction amount for the finger joint through a virtual spring-damping system. :

[0063]

[0064] in, This is the virtual stiffness coefficient. This is the virtual damping coefficient. The current closing speed of the finger clamp. The final target position of the finger clamp. From the initial planned location The result obtained by adding this correction amount is:

[0065]

[0066] In this way, the finger clamp exhibits a compliant mechanical behavior when in contact with the cable, avoiding rigid impact.

[0067] The target clamping force It is not a fixed value, but rather an estimated wire diameter based on the cable instance mask output by the vision unit. And the cable material types pre-installed in the database It performs adaptive adjustments. The adjustment strategy is implemented through a predefined lookup table:

[0068]

[0069] in, and The empirical coefficients were obtained based on a large amount of experimental data. This is an adjustment factor related to the material's coefficient of friction and allowable pressure. For example, for smooth polyvinyl chloride (PVC) materials, Use a lower value to avoid crushing; for rough rubber materials, use a higher value to ensure gripping reliability.

[0070] During the finger clamp closure process, the built-in force sensor continuously monitors the actual contact force. The vibration sensor monitors high-frequency signals. If within the time window... Internal detection The sudden drop (satisfying) (or vibration energy exceeding the threshold) If this occurs, a sliding response strategy will be immediately triggered: First, the sliding response will proceed in a certain step size. Increase target clamping force If the increased slip signal persists, the current gripping point is deemed unreliable. The controller instructs the gripper to fully release and plans to move the actuator to the middle of the cable, switching to a larger default gripping force. It directly performs finger gripping, thereby achieving adaptive switching of gripping modes and redundancy protection.

[0071] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of the invention (including the claims) is limited to these examples; within the framework of the invention, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of the different aspects of the invention as described above, which are not provided in the details for the sake of brevity.

[0072] This invention is intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. An automatic material handling mechanism for cable processing equipment, characterized in that, include: The incoming material alignment and conveying unit is used to flatten the scattered cables and convey them through a conveyor belt driven by a servo motor. The encoder signal of the servo motor is connected to the control and synchronization unit. Industrial vision or AI vision units are used to perform visual inspection and measurement of cables during the conveying process, outputting instance masks, end positions, spindle posture, tangling / overlapping judgment and single-piece gripping score for a single cable, and completing aerial photography and gripping pose prediction without stopping the cable. The material handling execution unit includes an actuator with XY-θ and Z degrees of freedom and a dual-modal compliant gripper. The gripper includes a controllable negative pressure adsorption module and an adaptive finger clamp module, and has force / slip sensing capabilities. The attitude correction and guidance unit is used to correct the end direction of the grasped cable and standardize the feeding posture. The control and synchronization unit synchronizes with the conveyor encoder in time and space to complete the flying camera trigger, trajectory planning, target allocation and abnormal component diversion; The industrial vision or AI vision unit works in conjunction with the control and synchronization unit to enable the actuator to dynamically grasp and place the cable without stopping the conveyor belt. The industrial vision or AI vision unit includes a top-mounted high-speed camera and a strobe light source, which performs short-exposure aerial photography to suppress motion blur. The industrial vision or AI vision unit also includes a 3D depth camera, used to perform high-level verification of cable overlap and multi-layer stacking, and adjust the individual piece graspability score accordingly. The end detection is obtained by skeletonization and high curvature end feature extraction after instance segmentation, and the principal axis angle is calculated by PCA. The dual-modal compliant gripper switches to finger clamp closure after the end is lifted by adsorption to achieve stable transport, and triggers secondary clamping or changes to mid-section gripping based on slip sensing. The entanglement / overlap determination is based on the fusion determination of the closed-loop / multi-intersection topology of the 2D instance skeleton and the 3D local height statistics. When the confidence threshold is exceeded, a decoupling action is performed, and if it fails, the data is diverted to the whole column.

2. The automatic material handling mechanism for cable processing equipment according to claim 1, characterized in that, The control and synchronization unit calculates the target displacement Δx=v·Δt based on the conveyor belt speed v and the expected gripping time Δt, and performs micro-position correction within 10–20 ms before the end reaches the target position.

3. The automatic material handling mechanism for cable processing equipment according to claim 1, characterized in that, The attitude correction and guidance unit includes a guide groove and a passive rotating guide wheel to achieve secondary correction of the end direction and axial alignment.

4. The automatic material handling mechanism for cable processing equipment according to claim 1, characterized in that, The light source includes switchable visible and near-infrared light sources to adapt to black or low-contrast cable surfaces.

5. The automatic material handling mechanism for cable processing equipment according to claim 1, characterized in that, The target allocation uses a minimum cost matching algorithm to maximize throughput per unit time. The cost function integrates target arrival time, executor reachability, and grasp confidence.

Images