An artificial intelligence-based logistics robot control method and system

Abstract

The application provides a logistics robot control method and system based on artificial intelligence. After the logistics robot obtains the surface material type, three-dimensional geometric contour and surface stress distribution data of the package to be sorted, the hardness grade and friction coefficient interval of the package are determined, and the geometric model thereof is reconstructed to calculate the minimum circumscribed cube size and identify the fragile area. According to the information, the suction or clamping mode, the unfolding angle, the contact point and the trajectory of the grabbing action are adjusted, and the sorting strategy is generated by considering dynamic factors such as the sorting line flow rate. During the execution process, the action parameters and decision weights are updated through the artificial intelligence algorithm according to the actual effect, and the sorting strategy is continuously optimized. The technical scheme provided by the application can improve the efficiency and precision of logistics robot control.

Description

Technical Field

[0001] This application relates to the field of logistics robot control technology, and in particular to a logistics robot control method and system based on artificial intelligence. Background Technology

[0002] In modern logistics centers, with the rapid development of e-commerce, the volume of parcel sorting has increased dramatically, leading to ever-higher demands for logistics efficiency and accuracy. To ensure high efficiency and low error rates, logistics systems need the ability to automatically identify parcel attributes (such as material and shape), assess their physical characteristics, and dynamically adjust handling methods accordingly. This not only helps protect parcels from damage but also optimizes the operating modes of logistics robots, ensuring efficient operation. Furthermore, considering the potentially fragile areas of different parcels, the system also needs to be able to accurately plan gripping points and paths to avoid any potential damage.

[0003] Currently, a targeted technical solution involves using a combination of vision and weight sensors to acquire basic information about the package, including its size, shape, and approximate mass distribution. Based on this data, the system can roughly estimate the package's physical characteristics and appropriate handling methods. This solution utilizes computer vision technology to quickly scan and create a two-dimensional or simple three-dimensional model of the package, and then provides basic operational guidance to the robot according to preset rules, such as selecting the suction or gripping position.

[0004] However, this existing approach has significant limitations. First, relying solely on visual and weight information, it lacks precision in determining the type of package material, its hardness level, coefficient of friction, and other key physical properties, potentially leading the robot to adopt unsuitable grasping strategies. Second, this method struggles to effectively identify subtle features on the package surface, particularly the boundaries of fragile areas, increasing the risk of damage during handling. Furthermore, in the face of complex and ever-changing real-world working environments, a decision-making mechanism based solely on fixed rules lacks sufficient flexibility and adaptability, failing to achieve true dynamic adjustment and optimization. Summary of the Invention

[0005] This application provides an artificial intelligence-based logistics robot control method and system to solve the problems of low control efficiency and poor accuracy of logistics robots in the prior art.

[0006] In a first aspect, embodiments of this application provide an artificial intelligence-based control method for logistics robots, comprising:

[0007] The surface material type data, three-dimensional geometric contour data, and surface stress distribution data of the packages to be sorted are obtained through logistics robots.

[0008] The surface material type data is used to determine the material hardness level and friction coefficient range of the packages to be sorted.

[0009] The geometric model of the package to be sorted is reconstructed based on the three-dimensional geometric contour data and the minimum outer cube size of the package is calculated. The fragile area boundary of the surface of the package to be sorted is identified based on the surface stress distribution data.

[0010] The adsorption or gripping mode of the logistics robot is adjusted according to the material hardness level and friction coefficient range. The unfolding angle of the logistics robot is adjusted according to the minimum package outer cube size. At the same time, the contact point distribution and movement trajectory of the logistics robot's grasping action are constrained according to the fragile area boundary. A dynamic sorting strategy is generated by combining the current sorting line flow rate, target sorting area capacity and robot joint load status.

[0011] Based on the execution results of the dynamic sorting strategy, the motion parameters of the logistics robot and the priority weights of the dynamic sorting decision are updated through artificial intelligence algorithms to update the dynamic sorting strategy of the logistics robot.

[0012] Optionally, the step of generating a dynamic sorting strategy by constraining the contact point distribution and motion trajectory of the logistics robot's grasping action according to the boundary of the fragile area, and combining the current sorting line flow rate, the target sorting area capacity, and the robot joint load state, includes:

[0013] Based on the boundary of the fragile region, candidate contact point regions are delineated on the surface of the geometric model, and coordinate points at the safety distance outside the boundary of the fragile region are excluded to generate an initial set of contact points.

[0014] From the initial set of contact points, candidate contact point pairs that satisfy the condition of coincidence between the direction of the grasping force and the projection of the center of gravity of the package are selected, and the force closure stability score of the candidate contact point pairs is calculated.

[0015] The grasping motion trajectory in the geometric model coordinate system is decomposed into a lateral translation component and a longitudinal following component. The velocity compensation of the lateral translation component is calculated based on the sorting line flow rate, and the timing parameters of the longitudinal following component are generated.

[0016] Based on the reciprocal of the ratio of the real-time joint current data of the logistics robot to the historical load, the joint torque balance factor of the candidate contact point pair is calculated. The force closure stability score, timing parameters and joint torque balance factor are fused to generate a multi-objective optimization weight set. Based on the multi-objective optimization weight set, a dynamic sorting strategy is output.

[0017] Optionally, the step of fusing the force closure stability score, time-series parameters, and joint torque balance factors to generate a multi-objective optimization weight set includes:

[0018] The initial weight allocation of the force closure stability score, timing parameters, and joint torque balance factor is determined based on the material hardness grade and the sorting line flow rate. When the material hardness grade is higher than the threshold, the score weight is increased.

[0019] The force closure stability score is converted into a standardized interval, the time series parameters are decomposed into synchronization error tolerance and time window matching degree, and a multi-dimensional parameter vector is generated by combining the joint torque balance factor.

[0020] The initial weight allocation is corrected based on the rate of change of the target sorting area capacity, and the corrected initial weight allocation is combined with the multi-dimensional parameter vector to generate a comprehensive score for the candidate scheme.

[0021] The joint torque balancing factor is truncated based on the difference between the real-time peak current of the logistics robot's joints and the historical load threshold, and a multi-objective optimization weight set is generated based on the comprehensive score of the candidate schemes.

[0022] Optionally, the step of converting the force closure stability score into a standardized interval, decomposing the time-series parameters into synchronization error tolerance and time window matching degree, and generating a multi-dimensional parameter vector by combining the joint torque equalization factor includes:

[0023] Based on the scoring distribution range of the historical crawled case library, the force closure stability score is dynamically mapped to a standardized interval;

[0024] When decomposing the timing parameters, the synchronization error tolerance is calculated based on the ratio of the sorting line flow rate to the size of the outer cube of the package, and the time window matching degree is generated based on the theoretical motion time from the center of gravity of the geometric model to the opening of the sorting container.

[0025] The joint torque equalization factor is dynamically adjusted based on the fluctuation variance between the real-time joint current data and the historical average current.

[0026] The standardized interval, synchronization error tolerance, time window matching degree, and adjusted joint torque balance factor are aligned according to the priority rules of logistics scenarios to generate a multi-dimensional parameter vector that includes mechanical stability, temporal synchronization, and mechanical safety.

[0027] Optionally, adjusting the adsorption or clamping mode of the logistics robot according to the material hardness grade and friction coefficient range includes:

[0028] The adsorption force safety threshold is set based on the material hardness level. When the material hardness level is lower than the flexibility threshold, the adsorption mode is activated, and the distribution density of the adsorption pore array is dynamically adjusted according to the friction coefficient range.

[0029] When the material hardness level is higher than the flexibility threshold, switch to clamping mode. Calculate the upper limit of the contact pressure of the clamping phalanx based on the friction coefficient range and constrain the initial closing angle of the phalanx. In the clamping mode, adjust the phalanx force closing direction based on the angle between the geometric model centroid projection and the clamping force line of action.

[0030] In the adsorption mode, the adsorption contact surface texture is matched according to the lower limit of the friction coefficient range, and the number of auxiliary adsorption pores activated is increased in local low friction areas.

[0031] When the friction coefficient range crosses a critical value, the mixing mode is activated. The adsorption is maintained in the central area of ​​the adsorption pore, and the clamping knuckles are deployed in the edge area. The initial closing angle of the knuckles is positively correlated with the diagonal length of the circumscribed cube.

[0032] Optionally, the step of calculating the upper limit of the contact pressure of the clamping phalanx based on the friction coefficient range and constraining the initial closing angle of the phalanx includes:

[0033] The maximum contact pressure threshold for clamping the knuckle is set according to the upper limit of the friction coefficient range. When the friction coefficient crosses the high and medium friction ranges, a segmented increasing strategy is adopted to increase the pressure threshold growth rate.

[0034] The safety margin of the initial closure angle of the knuckle is calculated based on the distribution density of the region with the maximum curvature in the geometric model. The safety margin is positively correlated with the product of the number of curvature extreme points and the lower limit of the friction coefficient.

[0035] Based on the shortest distance from the boundary of the fragile area to the clamping area in the surface stress distribution data, the maximum contact pressure threshold is dynamically attenuated, and the attenuation coefficient is inversely proportional to the square of the shortest distance.

[0036] Based on the aforementioned safety margin, when the upper limit of the contact pressure of the clamping phalanx reaches the threshold warning range, a high-frequency micro-amplitude vibration of the contact surface is triggered, and the pressure gradient distribution of the clamping phalanx is monitored in real time. If the pressure concentration area is detected to overlap with the area with the maximum curvature in the geometric model, the opening compensation angle of the adjacent phalanx is increased to disperse the stress.

[0037] Optionally, the step of acquiring surface material type data, three-dimensional geometric contour data, and surface stress distribution data of the packages to be sorted through a logistics robot includes:

[0038] The multispectral sensor of the logistics robot collects the reflectance spectrum of the package surface, extracts the multi-band spectral response difference to match the surface material type data, and enables polarization filtering to suppress interference in high reflective areas.

[0039] A three-dimensional point cloud is reconstructed by fusing structured light and stereo vision, curvature features are extracted to generate three-dimensional geometric contour data, and multi-angle scanning is initiated to complete the point cloud in blind areas.

[0040] By applying a pressure gradient through a flexible tactile array, surface stress distribution data is generated by recording the deformation rate and stress relaxation curve. When a stress change is detected, high-frequency sampling is triggered to obtain high-frequency sampled stress data.

[0041] The multi-band spectral response differences, multi-angle scan-completed point cloud data, and high-frequency sampled stress data are synchronized to the three-dimensional coordinate system to generate a cross-modal correlation mapping table.

[0042] Secondly, embodiments of this application provide an artificial intelligence-based logistics robot control system, comprising:

[0043] The acquisition module obtains surface material type data, three-dimensional geometric contour data, and surface stress distribution data of the packages to be sorted through the logistics robot.

[0044] The determination module determines the material hardness level and friction coefficient range of the package to be sorted based on the surface material type data.

[0045] The identification module reconstructs the geometric model of the package to be sorted based on the three-dimensional geometric contour data and calculates the minimum outer cube size of the package. It also identifies the fragile area boundary of the surface of the package to be sorted based on the surface stress distribution data.

[0046] The generation module adjusts the adsorption or clamping mode of the logistics robot according to the material hardness level and friction coefficient range, adjusts the unfolding angle of the logistics robot according to the minimum package outer cube size, and constrains the contact point distribution and movement trajectory of the logistics robot's grasping action according to the fragile area boundary, and generates a dynamic sorting strategy by combining the current sorting line flow rate, target sorting area capacity and robot joint load status.

[0047] The update module updates the dynamic sorting strategy of the logistics robot by using artificial intelligence algorithms based on the execution results of the dynamic sorting strategy.

[0048] Thirdly, embodiments of this application provide a computing device, including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are invoked and executed by the processing component to implement an artificial intelligence-based logistics robot control method as described in the first aspect above.

[0049] Fourthly, embodiments of this application provide a computer storage medium storing a computer program, which, when executed by a computer, implements an artificial intelligence-based logistics robot control method as described in the first aspect.

[0050] In this embodiment, a logistics robot acquires surface material type data, three-dimensional geometric contour data, and surface stress distribution data of packages to be sorted. Based on the surface material type data, the material hardness level and friction coefficient range of the packages to be sorted are determined. Based on the three-dimensional geometric contour data, the geometric model of the packages to be sorted is reconstructed, and the minimum circumscribed cube size of the packages is calculated. Based on the surface stress distribution data, the fragile area boundaries of the surface of the packages to be sorted are identified. The adsorption or gripping mode of the logistics robot is adjusted according to the material hardness level and friction coefficient range. The unfolding angle of the logistics robot is adjusted in conjunction with the minimum circumscribed cube size of the packages. At the same time, the contact point distribution and motion trajectory of the logistics robot's grasping action are constrained according to the fragile area boundaries. Combined with the current sorting line flow rate, target sorting area capacity, and robot joint load status, a dynamic sorting strategy is generated. Based on the execution result of the dynamic sorting strategy, the motion parameters of the logistics robot and the priority weight of the dynamic sorting decision are updated through an artificial intelligence algorithm to update the dynamic sorting strategy of the logistics robot.

[0051] The technical solution of this application has the following beneficial effects:

[0052] This application utilizes a logistics robot to acquire data on the surface material type, three-dimensional geometric contour, and surface stress distribution of packages, ensuring a comprehensive understanding of the packages' physical properties. Based on the surface material type data, the hardness level and friction coefficient range of the packages are accurately determined, providing fundamental parameters for subsequent operations. The package's geometric model is reconstructed using the three-dimensional geometric contour data, and the minimum circumscribed cube size is calculated. Simultaneously, the boundaries of fragile areas are accurately identified based on the surface stress distribution data, helping to protect the package's integrity. By combining information on material hardness level, friction coefficient range, minimum circumscribed cube size, and fragile area boundaries, the robot's adsorption or gripping mode, unfolding angle, and the contact point distribution and motion trajectory of the grasping action are dynamically adjusted, improving the safety and efficiency of package handling. A dynamic sorting strategy is generated by comprehensively considering factors such as the current sorting line flow rate, the target sorting area capacity, and the robot joint load status, and the strategy is continuously optimized through feedback from the execution results.

[0053] Furthermore, this method first delineates candidate contact point regions for safe grasping based on the boundaries of fragile areas in the geometric model, selects contact points that satisfy the condition of coincidence between the force direction and the center of gravity projection, and calculates their stability scores. Next, it decomposes the grasping motion trajectory into lateral and longitudinal components, adjusts lateral velocity compensation according to the sorting line flow rate, and plans longitudinal following timing parameters. Finally, it integrates joint torque balance factors, force closure stability scores, and timing parameters to generate a multi-objective optimization weight set to output the final dynamic sorting strategy. This process effectively improves the safety, stability, and adaptability of the grasping action, significantly reduces the risk of package damage during handling, and simultaneously improves the working efficiency and flexibility of the logistics robot.

[0054] These or other aspects of this application will become more apparent in the following description of the embodiments. Attached Figure Description

[0055] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0056] Figure 1 A flowchart of an artificial intelligence-based logistics robot control method provided in this application is shown;

[0057] Figure 2 A schematic diagram of the structure of an artificial intelligence-based logistics robot control system provided in this application is shown;

[0058] Figure 3 A schematic diagram of the structure of a computing device provided in this application is shown. Detailed Implementation

[0059] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0060] In some of the processes described in the specification, claims, and accompanying drawings of this application, multiple operations appearing in a specific order are included. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or may be executed in parallel. The operation numbers, such as 101, 102, etc., are merely used to distinguish different operations and do not themselves represent any execution order. Furthermore, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should be noted that the descriptions such as "first," "second," etc., in this document are used to distinguish different messages, devices, modules, etc., and do not represent a chronological order, nor do they limit "first" and "second" to different types.

[0061] This solution utilizes logistics robots to acquire surface material type data, 3D geometric contour data, and surface stress distribution data of packages to be sorted. It aims to use multispectral sensors to collect surface reflectance maps of the packages and reconstruct their 3D models using structured light and stereo vision technologies. Simultaneously, a flexible tactile array is employed to record the pressure response characteristics of the package surface, thereby achieving a comprehensive perception of the package's physical properties. This method emphasizes cross-modal data fusion to improve the accuracy and completeness of information acquisition.

[0062] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0063] Figure 1 A flowchart of an artificial intelligence-based logistics robot control method is provided as an embodiment of this application, such as... Figure 1 As shown, the method includes:

[0064] 101. Obtain surface material type data, three-dimensional geometric contour data, and surface stress distribution data of packages to be sorted through logistics robots;

[0065] In this step, surface material type data refers to data that determines the physical properties of the outer material by analyzing its composition, including paper, plastic, or fabric. This information is used to select the appropriate processing method later.

[0066] Three-dimensional geometric contour data refers to all dimensional information describing the shape of a package and its variations. It is used to reconstruct a digital model of the package to help identify its shape and size.

[0067] Surface stress distribution data refers to the pressure values ​​at various points on the surface of a package. It helps identify potentially vulnerable areas on the package and ensures that excessive force is avoided on these fragile parts during handling.

[0068] In this embodiment, firstly, spectral analysis technology is used to identify the surface material type of the package. This is achieved by irradiating the package with light of a specific wavelength and analyzing the reflectance spectrum to determine the material type. Next, a laser scanner is used to capture the package's shape features, generating a high-precision digital model. Finally, an embedded pressure sensor is used to measure the pressure values ​​at different locations on the package's surface, obtaining a surface stress distribution map. By combining these three types of data, the logistics robot can comprehensively understand the specific characteristics of each package, thereby formulating corresponding processing strategies.

[0069] In a busy logistics center, an intelligent logistics robot first performs spectral analysis on a cardboard box, quickly determining its material to be corrugated cardboard. Then, the robot uses a laser scanner to create a detailed model of the box, including parameters such as its length, width, and height. Simultaneously, it records the pressure distribution on the box's surface using built-in pressure-sensing patches, paying particular attention to areas that may have high stress, preparing for subsequent operations.

[0070] 102. Determine the material hardness grade and friction coefficient range of the package to be sorted based on the surface material type data;

[0071] In this step, the material hardness rating is an indicator of the package's resistance to deformation, determined based on surface material type data. This rating helps in selecting the appropriate suction or clamping method to prevent the package from being damaged by excessive compression.

[0072] The coefficient of friction range reflects the degree to which the package may slip between itself and the gripping device. Calculated based on surface material type data, it guides the selection of the correct operating mode to ensure the package does not slip due to insufficient friction.

[0073] In this embodiment, a pre-trained machine learning model is used, taking the surface material type data obtained in step 101 as input, and calculating the hardness level and friction coefficient range of the encapsulated material by comparing it with standard samples in the database. This process involves using classification algorithms, such as support vector machines or decision tree algorithms, trained on a large dataset of known materials. The final result, based on the closest match, provides estimates of hardness and friction characteristics, guiding the selection of subsequent operations.

[0074] Based on the preceding analysis, the logistics robot consulted its internal database and analyzed the data through a machine learning model, concluding that the corrugated cardboard has a moderate hardness level and a high coefficient of friction range. This indicates that it is not easily damaged by external pressure, nor is it easy for it to slip off the gripper, providing an important reference for the next step of the operation.

[0075] 103. Reconstruct the geometric model of the package to be sorted based on the three-dimensional geometric contour data and calculate the minimum outer cube size of the package; identify the fragile area boundary of the surface of the package to be sorted based on the surface stress distribution data.

[0076] In this step, the geometric model is a digital representation of the package reconstructed using three-dimensional geometric contour data, used to accurately describe the shape and size of the package.

[0077] The minimum enclosing cube size refers to the space required to completely contain the package. It is calculated based on a geometric model and helps optimize storage space utilization.

[0078] The fragile area boundary is an area with high stress on the package surface identified based on surface stress distribution data. These areas are considered more vulnerable and require special protection during handling.

[0079] In this embodiment, computer-aided design software reconstructs the geometric model of the package based on three-dimensional geometric contour data. Then, an algorithm is applied to calculate the minimum circumscribed cube size of the package that can completely contain the model. For the identification of fragile area boundaries, a threshold is set for analyzing surface stress distribution data, marking areas with abnormally high stress as potentially vulnerable parts. This process combines image processing techniques and mathematical modeling methods to ensure accuracy and reliability.

[0080] Following the previous step, the logistics robot not only determined the optimal storage size for the cardboard box but also discovered a higher stress level on one side where the printed pattern was located, indicating it might be a vulnerable area. Therefore, when planning the handling route, this area was deliberately avoided to prevent accidental damage and ensure the package arrived safely at its destination.

[0081] 104. Adjust the adsorption mode or clamping mode of the logistics robot according to the material hardness level and friction coefficient range, adjust the unfolding angle of the logistics robot according to the minimum package outer cube size, and constrain the contact point distribution and movement trajectory of the logistics robot's grasping action according to the fragile area boundary, and generate a dynamic sorting strategy by combining the current sorting line flow rate, target sorting area capacity and robot joint load status.

[0082] In this step, the suction mode or gripping mode refers to the method used by the logistics robot to grasp the package, including using a vacuum suction cup for suction or a mechanical gripper for gripping. The most suitable mode is selected based on the package's material hardness level and friction coefficient range to ensure that the package is neither damaged nor slips during handling.

[0083] The sorting line flow rate is the speed at which packages flow along the sorting line, affecting the operating rhythm and efficiency of logistics robots.

[0084] The target sorting area capacity refers to the maximum number or volume of packages that the target sorting area can hold, which is crucial for planning the placement of logistics robots.

[0085] Robot joint load status describes the current load on each joint of the robot, helping to optimize gripping and movement strategies and prevent equipment damage caused by overload.

[0086] In this embodiment, firstly, the most suitable adsorption or gripping method, such as a vacuum suction cup or a robotic gripper, is selected based on the material hardness grade and friction coefficient range. Next, the appropriate robotic arm deployment angle is calculated using the minimum circumscribed cube size of the package to ensure a stable grip without damage. For fragile areas, a path planning algorithm determines the optimal gripping path to avoid direct contact with these areas. Furthermore, considering factors such as sorting line flow rate, target sorting area capacity, and robot joint load status, an optimization algorithm is used to formulate the most effective sorting strategy.

[0087] Based on previous data analysis, the logistics robot selected a gripping method suitable for the characteristics of the corrugated cardboard and adjusted the angle of the robotic arm for safe handling. When planning the transport path, special attention was paid to avoiding identified vulnerable areas to ensure that packages were not damaged due to improper handling. Furthermore, considering the current sorting line speed and the spatial constraints of the target area, the robot's motion sequence was optimized, improving overall efficiency.

[0088] 105. Based on the execution result of the dynamic sorting strategy, update the motion parameters of the logistics robot and the priority weight of the dynamic sorting decision through artificial intelligence algorithm, so as to update the dynamic sorting strategy of the logistics robot.

[0089] In this step, the execution result of the dynamic sorting strategy refers to the actual effect of the logistics robot after performing sorting tasks according to the predetermined strategy, such as completion time and package damage rate, which is used to evaluate the effectiveness of the strategy.

[0090] Artificial intelligence algorithms are a series of algorithms used to analyze the execution results of dynamic sorting strategies and adjust the motion parameters and decision priority weights of logistics robots accordingly, such as reinforcement learning algorithms.

[0091] Action parameters refer to specific settings such as the magnitude of suction force, clamping force, and moving speed, which directly affect the performance of logistics robots when performing tasks.

[0092] Decision priority weights are standard weight assignments that make the best choice among multiple possible operations, such as maximizing safety while ensuring speed. By adjusting these weights, sorting strategies can be optimized.

[0093] In this embodiment, a reinforcement learning algorithm is used to evaluate the effectiveness of each operation, collect feedback information, and adjust the strategy parameters for the next action accordingly. Specifically, the result of each task execution is used as input, evaluated by a series of predefined evaluation criteria (such as the time to successfully complete the task, the potential damage to the package, etc.), and then the corresponding parameter values ​​are adjusted by the algorithm. This process is a closed-loop control mechanism that gradually improves the system's performance through continuous iterative learning.

[0094] After numerous practical trials, the logistics robot adjusted its suction and gripping force settings and re-prioritized decisions based on collected data and feedback. For example, when handling similar types of packages, it learned to be more careful with printed areas, reducing the error rate. Over time, this optimization capability has made the entire logistics sorting system more efficient and reliable.

[0095] In summary, steps 101 to 105, integrating advanced sensing technologies and intelligent algorithms, enable highly personalized handling of packages to be sorted, significantly improving the efficiency and safety of logistics sorting. By accurately identifying package characteristics and adjusting operating modes accordingly, combined with real-time data analysis and self-optimization mechanisms, logistics robots can flexibly cope with various challenges in complex and ever-changing working environments. This not only improves the success rate of individual tasks but also promotes the continuous improvement and development of the entire logistics process.

[0096] To address the issue of optimizing gripping strategies based on the fragile areas of packages, this solution determines the material hardness level and friction coefficient range of the packages to be sorted based on surface material type data. The aim is to identify the specific material properties of the packages by analyzing their surface reflectance spectra, and thereby assess their hardness and friction characteristics. This process utilizes the correlation between material properties and optical response, providing necessary physical parameter support for subsequent operations. This step is a crucial step in achieving automated processing, ensuring the effective adjustment of subsequent gripping and handling strategies. In some embodiments, step 104, which constrains the contact point distribution and motion trajectory of the logistics robot's gripping action based on the boundary of the fragile area, and combines this with the current sorting line flow rate, target sorting area capacity, and robot joint load status to generate a dynamic sorting strategy, includes:

[0097] 201. Based on the boundary of the fragile region, define the candidate region of contact points on the surface of the geometric model, exclude the coordinate points of the safe distance outside the boundary of the fragile region, and generate an initial set of contact points;

[0098] In step 201, the fragile area boundary refers to the area with high stress on the package surface. It's an additional safety interval zone established to avoid applying pressure to the fragile area, ensuring that the gripping action doesn't directly affect the vulnerable part. The safety distance is an additional range established to protect the fragile area from direct contact. The initial contact point set is a series of locations selected based on these conditions that can be used for gripping operations. The contact point candidate area consists of possible locations for robot gripping other than these fragile areas.

[0099] In this embodiment, firstly, fragile areas on the package are identified by analyzing the stress distribution in the three-dimensional geometric contour data, and an appropriate safety distance is set around them as a buffer. Image processing techniques are then used to mark the coordinates of all non-fragile areas. Finally, all possible safe gripping points are marked on the remaining surface, forming an initial set of contact points. This process combines computer vision algorithms with mathematical modeling to ensure accuracy.

[0100] 202. Select candidate contact point pairs from the initial contact point set that satisfy the condition of coincidence between the gripping force direction and the projection of the package's center of gravity, and calculate the force closure stability score of the candidate contact point pairs.

[0101] In step 202, the coincidence condition between the gripping force direction and the package's center of gravity projection refers to the desire, when selecting the gripping point, for the gripping force direction to pass as close as possible to the package's center of gravity to reduce rotational effects. This considers whether the gripping force direction is close to the projection of the package's center of gravity onto the horizontal plane; a high coincidence indicates better stability. The force closure stability score assesses the likelihood of the package remaining stable during gripping; a high score indicates even greater stability.

[0102] In this embodiment, physical simulation software is used to calculate the distance and angle between each contact point pair and the projection of the center of gravity of the package, and to select contact point pairs whose force direction coincides with the projection of the center of gravity as much as possible. For each selected pair of contact points, a force closure stability score is calculated based on its geometric relationship and mechanical principles. Finally, the most suitable contact point pair is selected according to the score results to ensure the best gripping effect.

[0103] 203. Decompose the grasping motion trajectory in the geometric model coordinate system into a lateral translation component and a longitudinal following component. Calculate the velocity compensation amount of the lateral translation component based on the sorting line flow rate, and generate the timing parameters of the longitudinal following component.

[0104] In step 203, the lateral translation component and the longitudinal following component refer to the two main directions of the robot arm's movement during the grasping process. The former is used to adjust the robot's speed compensation to match the speed of the sorting line, adjusting the change in the robot's lateral movement speed according to the sorting line flow rate to ensure accurate package placement. The latter describes the time-series parameters of the robot following the target path in the vertical direction, ensuring a smooth transition.

[0105] In this embodiment, a path planning algorithm is used to decompose the grasping motion trajectory into two dimensions: horizontal and vertical. Based on the current speed of the sorting line, the amount of horizontal speed compensation that needs to be increased or decreased is calculated to synchronize the robot's movement speed with the sorting line. Simultaneously, time-series parameters for vertical movement are determined to ensure that the package can reach the target location smoothly and quickly.

[0106] 204. Calculate the joint torque balance factor of the candidate contact point pair based on the reciprocal of the ratio of the real-time joint current data of the logistics robot to the historical load. Integrate the force closure stability score, timing parameters and joint torque balance factor to generate a multi-objective optimization weight set. Output a dynamic sorting strategy based on the multi-objective optimization weight set.

[0107] In step 204, the joint torque balance factor reflects the load balance state of each joint of the robot during task execution, and is calculated by real-time monitoring of joint current data. The force closure stability score is a quantitative standard for evaluating whether each candidate gripping point can stably keep the object from falling. The multi-objective optimization weight set is the optimal parameter combination determined after comprehensively considering multiple factors such as force closure stability, motion trajectory, and timing parameters, used to guide the robot to complete the optimal operation strategy.

[0108] In this embodiment, a joint torque balance factor is calculated by real-time monitoring of robot joint current data and combining it with historical load ratios. This factor is then combined with previously obtained force closure stability scores, timing parameters, etc., to form a multi-objective optimization weight set. This process involves data analysis and the application of optimization algorithms. Based on this weight set, the most suitable dynamic sorting strategy for the current environment is formulated.

[0109] Here's a concrete example: In a busy logistics center, an intelligent logistics robot first processes cardboard boxes. It identifies areas with printed patterns on one side of the box as having high stress levels, marks these as fragile areas, and establishes a safe distance around them. Next, the robot finds multiple potential safe gripping points on the remaining surface and selects several points closest to the package's center of gravity projection. Then, the robot calculates the force closure stability score for these points and adjusts the speed compensation for gripping and placing actions based on the sorting line flow. Finally, considering the joint load, the robot optimizes the gripping force distribution, ensuring both operational safety and efficiency.

[0110] In summary, steps 201 to 204 significantly improve the accuracy and efficiency of the logistics sorting process while reducing the risk of package damage. By accurately identifying package characteristics and adjusting operating modes accordingly, combined with real-time data analysis and self-optimization mechanisms, logistics robots can flexibly cope with various challenges in complex working environments, thereby improving the performance and reliability of the entire logistics process.

[0111] To address the sorting efficiency and safety issues of intelligent logistics robots handling fragile items in logistics centers, this solution reconstructs the geometric model of the packages to be sorted based on 3D geometric contour data and calculates the minimum outer cube size of the package. Simultaneously, it identifies the boundaries of fragile areas based on surface stress distribution data. The main objective is to accurately reconstruct the package's shape and conduct a detailed safety assessment based on this. Through high-precision 3D modeling technology, not only can the size and shape of the package be obtained, but also potentially vulnerable parts can be accurately located, laying the foundation for developing a safe handling plan. In some embodiments, step 204, which involves fusing the force closure stability score, temporal parameters, and joint torque balance factors to generate a multi-objective optimization weight set, includes:

[0112] 301. Determine the initial weight allocation of the force closure stability score, timing parameters, and joint torque balance factor based on the material hardness grade and the sorting line flow rate. When the material hardness grade is higher than the threshold, increase the score weight.

[0113] In step 301, the material hardness rating represents the level of the packaging material's resistance to deformation or breakage. This is obtained through physical testing or packaging information. The initial weight allocation for the force closure stability score adjusts the emphasis on gripping stability based on material hardness; harder materials can have their stability requirements appropriately reduced. The initial weight allocation for timing parameters is a process of determining the motion timing based on the sorting line flow rate; high-speed production lines require more precise timing control. The initial weight allocation for the joint torque balance factor must consider the robot joint load balance to avoid mechanical damage caused by uneven load.

[0114] In this embodiment, the material hardness of the cardboard box is first analyzed, and the initial weights of various indicators are set based on the real-time flow rate of the sorting line. If the material hardness of the cardboard box is detected to be higher than a preset threshold, the weight of the force closure stability score is increased accordingly, reducing the requirement for synchronization error tolerance. This step uses sensor data to compare and analyze with historical databases to obtain the corresponding weight values, thereby providing a basic reference for subsequent steps.

[0115] 302. Convert the force closure stability score into a standardized interval, decompose the time series parameters into synchronization error tolerance and time window matching degree, and generate a multi-dimensional parameter vector by combining the joint torque equalization factor;

[0116] In step 302, the standardization interval transforms the force closure stability score onto a common comparison scale, facilitating comparisons between different solutions. Synchronization error tolerance refers to the allowable range of time deviations during execution, ensuring that even slight time differences will not affect the overall task completion. Time window matching measures whether the actual operation conforms to the predetermined time plan, ensuring close coordination between each stage. The multidimensional parameter vector contains all standardized key parameters, used to comprehensively evaluate the overall performance of candidate solutions.

[0117] In this embodiment, a mathematical transformation method is used to map the force closure stability score to a standardized interval, and the time-series parameters are decomposed into two dimensions: synchronization error tolerance and time window matching degree. Then, a multi-dimensional parameter vector is formed by combining the joint torque balance factor. This process involves statistical methods and data analysis techniques, ultimately generating quantitative indicators reflecting the advantages and disadvantages of different schemes, providing a basis for further adjustments.

[0118] 303. Adjust the initial weight allocation according to the rate of change of the target sorting area capacity, and combine the adjusted initial weight allocation with the multi-dimensional parameter vector to generate a comprehensive score for candidate schemes;

[0119] In step 303, the target sorting area capacity change rate describes the rate of change in sorting area space utilization, affecting the frequency of strategy adjustments. The revised initial weight allocation dynamically adjusts the importance ratio of each indicator based on the sorting area status. The comprehensive score of candidate solutions refers to obtaining an overall evaluation score for each candidate solution by superimposing the revised weights and multi-dimensional parameter vectors, guiding the selection of the optimal path.

[0120] In this embodiment, the capacity changes of the target sorting area are monitored in real time, and the previously set weight allocation is adjusted accordingly. The adjusted weights are combined with the previously constructed multi-dimensional parameter vector to calculate the comprehensive score of each candidate solution. An adaptive algorithm is used here to respond to environmental changes, ensuring that the strategy is always in an optimal state to maximize efficiency and minimize risk.

[0121] 304. The joint torque balance factor is truncated based on the difference between the real-time peak current of the logistics robot's joints and the historical load threshold, and a multi-objective optimization weight set is generated based on the comprehensive score of the candidate schemes.

[0122] In step 304, the difference between the real-time peak current of the joint and the historical load threshold is used to determine whether the current load is close to the limit, protecting the machine from overload damage. The truncation of the joint torque balancing factor refers to reducing the influence of this factor when approaching the limit, preventing over-reliance on operations that could lead to mechanical failure. The multi-objective optimization weight set is a set of optimization strategies generated based on all the above information to ultimately guide the robot's behavior, ensuring efficient and safe operation.

[0123] In this embodiment, the real-time peak current of the robot's joints is monitored and compared with historical load thresholds. If an impending overload is detected, the influence of the joint torque balance factor is reduced accordingly. Based on this adjustment and the comprehensive scoring of candidate solutions, a new multi-objective optimization weight set is generated. This process employs a predictive model and feedback mechanism to ensure that the robot can operate efficiently within a safe range, achieving both safety and accuracy in automated operations.

[0124] Here is a specific example:

[0125] In a busy logistics center scenario, an intelligent logistics robot first performs a detailed scan of the cardboard boxes entering the sorting line, identifying their material hardness and avoiding fragile areas. Next, it determines initial weight allocation based on the cardboard box material and the sorting line flow rate, and then processes key parameters through a series of mathematical transformations to form a multi-dimensional parameter vector. Simultaneously, the system dynamically adjusts these weights based on the state of the sorting area and calculates a comprehensive score for each candidate solution. Finally, considering the safe load on the robot's joints, the overall strategy is optimized to ensure that every package is handled quickly and safely.

[0126] In summary, steps 301 to 304 significantly enhance the ability of intelligent logistics robots in the logistics center to handle complex tasks, not only improving sorting efficiency but also effectively reducing the risk of item damage, thus achieving safety and accuracy in automated operations.

[0127] To address the challenge of optimizing grasping strategies for intelligent logistics robots in dynamic environments, this solution adjusts the robot's adsorption or clamping modes based on material hardness levels and friction coefficient ranges. It also adjusts the unfolding angle according to the minimum package circumscribed cube size and constrains the contact point distribution and trajectory of the grasping action based on fragile area boundaries. The key focus is on dynamically optimizing the robot's operation based on the package's physical characteristics. By intelligently adjusting the adsorption or clamping force and position, the solution ensures the package is both stable and safe during handling, minimizing the risk of damage. In some embodiments, step 302 involves converting the force closure stability score into a standardized interval, decomposing the timing parameters into synchronization error tolerance and time window matching degree, and generating a multi-dimensional parameter vector based on the joint torque equalization factor, including:

[0128] 401. Based on the score distribution range of the historical crawled case library, the force closure stability score is dynamically mapped to a standardized interval;

[0129] In step 401, the score distribution range of the historical grabbing case library is based on the set of force closure stability score data from previous operations, used to determine the boundaries of the standardized interval. The force closure stability score is a quantitative standard for evaluating whether each candidate grabbing point can stably keep an object from falling. The standardized interval transforms the force closure stability scores under different conditions onto a common comparative scale, facilitating comparative analysis between different solutions.

[0130] In this embodiment, force closure stability score data from a historical crawling case library is first collected and analyzed to determine its distribution range. Then, using linear transformation or other mathematical methods (such as minimax normalization), the current score is dynamically adjusted to a preset standardized interval based on this distribution range. This step ensures that scores obtained under different conditions can be compared on the same benchmark, thus providing a reliable basis for subsequent steps. The final result is a standardized list of force closure stability scores.

[0131] 402. When decomposing the timing parameters, calculate the synchronization error tolerance based on the ratio of the sorting line flow rate to the size of the outer cube of the package, and generate the time window matching degree based on the theoretical motion time from the center of gravity of the geometric model to the opening of the sorting container.

[0132] In step 402, the synchronization error tolerance allows for a range of time deviations during execution to accommodate minor time differences in actual operation without affecting task completion. Time window matching measures whether the actual operation conforms to the predetermined time plan, ensuring close coordination between each stage and improving the overall process's coordination. The package's circumscribed cube size refers to the package's maximum outer dimensions, used to calculate the required operating space and time, helping to determine the optimal grasping timing.

[0133] In this embodiment, the synchronization error tolerance is determined by calculating the ratio of the sorting line velocity to the size of the outer cube of the package, reflecting the system's adaptability to speed changes. Simultaneously, the time window matching degree is generated based on the theoretical motion time from the geometric model's center of gravity to the sorting container opening, involving the application of path planning algorithms. This process ensures precise adjustment of the robot's movement rhythm, thereby achieving an efficient and accurate operation. The final results are specific values ​​for the synchronization error tolerance and the time window matching degree, providing a basis for optimization strategies.

[0134] 403. Based on the fluctuation variance between the real-time joint current data and the historical average current, dynamically adjust the gain of the joint torque equalization factor.

[0135] In step 403, the variance of the fluctuation between the real-time joint current data and the historical average current reflects the changes in the robot's joint load, assessing the health status of the mechanical system. The joint torque equalization factor reflects whether the load borne by each joint of the robot is balanced during task execution, and is calculated by monitoring the joint current data in real time. Dynamic gain adjustment can adjust the joint torque equalization factor in a timely manner according to the joint load condition, optimizing the safety and efficiency of mechanical operation.

[0136] In this embodiment, the real-time current of the robot joints is continuously monitored and compared with the historical average current to calculate the fluctuation variance. Based on this data, proportional-integral-derivative (PID) control or other adaptive control algorithms are used to dynamically adjust the joint torque balance factor. This process not only effectively prevents overload risks but also improves operational flexibility and response speed. The final result is an adjusted joint torque balance factor that ensures stable robot operation while improving operational efficiency.

[0137] 404. Align the standardized interval, synchronization error tolerance, time window matching degree, and adjusted joint torque balance factor according to the priority rules of logistics scenarios to generate a multi-dimensional parameter vector containing mechanical stability, temporal synchronization and mechanical safety.

[0138] In step 404, the logistics scenario priority rule is used to determine the importance ranking of each parameter based on the needs of the specific application scenario. For example, when handling fragile items, the force closure stability score may be more important than the synchronization error tolerance; while on a high-speed sorting line, the time window matching degree may have a higher priority. The multidimensional parameter vector is a vector containing quantitative indicators such as mechanical stability, temporal synchronization, and mechanical safety, used to comprehensively evaluate the overall performance of candidate solutions.

[0139] In this embodiment, the standardized interval, synchronization error tolerance, time window matching degree, and adjusted joint torque balance factor are dimensionally aligned according to the priority rules of the logistics scenario. This step involves complex weight allocation strategies and data analysis techniques, such as the analytic hierarchy process (AHP) or entropy weighting, to determine the weight of each parameter. This ultimately forms a multidimensional parameter vector that comprehensively reflects the state of each key indicator, providing a scientific basis for formulating the optimal strategy. The output is a multidimensional parameter vector containing all necessary information, guiding the robot to complete the sorting task efficiently and safely.

[0140] Here is a specific example:

[0141] In a busy logistics center scenario, when an intelligent logistics robot begins its work, it first performs a detailed scan of the cardboard boxes entering the sorting line, identifying their material hardness and avoiding fragile areas. Next, based on data from a historical grasping case library, it maps the force closure stability score to a standardized range. Simultaneously, it calculates the synchronization error tolerance based on the sorting line flow rate and package size, and generates a time window matching degree based on the distance from the geometric model's center of gravity to the sorting container opening. Furthermore, it monitors the real-time current of the robot's joints and adjusts the joint torque equalization factor based on historical data. Finally, combining all the information and prioritizing it according to rules, it constructs a multi-dimensional parameter vector to guide the robot to complete the sorting task efficiently and safely.

[0142] In summary, steps 401 to 404 significantly enhance the ability of intelligent logistics robots in the logistics center to handle complex tasks. This not only improves sorting efficiency but also effectively reduces the risk of item damage, achieving both safety and accuracy in automated operations. By dynamically adjusting strategies, the robots can flexibly respond to various situations, thereby greatly enhancing the flexibility and reliability of the entire logistics system.

[0143] To address the challenge of selecting the appropriate gripping mode for intelligent logistics robots when handling packages of different materials and friction coefficients, the solution generates a dynamic sorting strategy that comprehensively considers factors such as the current sorting line flow rate, the target sorting area capacity, and the robot joint load status. This aims to construct a highly adaptable and efficient sorting system. By monitoring and analyzing changes in the logistics environment in real time, the sorting strategy is adjusted promptly to ensure the smooth operation of the entire system. This flexibility enables the logistics robot to maintain efficient operation in complex and ever-changing working environments. In some embodiments, step 104, which involves adjusting the adsorption or gripping mode of the logistics robot based on the material hardness level and friction coefficient range, includes:

[0144] 501. Based on the material hardness level, set an adsorption force safety threshold. When the material hardness level is lower than the flexibility threshold, activate the adsorption mode and dynamically adjust the distribution density of the adsorption pore array according to the friction coefficient range.

[0145] In step 501, the material hardness rating indicates the level of the packaging material's ability to resist deformation or damage. This is obtained through physical testing or packaging information. The adsorption force safety threshold is the maximum permissible adsorption force set based on the material hardness to avoid damage to the package. The flexibility threshold is a preset hardness value; below this value, adsorption mode is used, and above this value, clamping mode is switched. The adsorption pore array distribution density refers to the density of the adsorption pores on the adsorption device, which is dynamically adjusted according to the coefficient of friction.

[0146] In this embodiment, the material hardness level of the cardboard box is first measured using a sensor, and a safe threshold for adsorption force is set accordingly. If the material hardness is detected to be lower than the flexibility threshold, the adsorption mode is activated. Next, image recognition technology is used to analyze the friction coefficient range of the package surface, and an algorithm (such as a linear regression model) is applied to dynamically calculate and adjust the distribution density of the adsorption pore array. For example, the number of adsorption pores is reduced in high-friction areas and increased in low-friction areas to adapt to different friction characteristics. The final result is an optimized adsorption pore layout scheme for the current package characteristics.

[0147] 502. When the material hardness level is higher than the flexibility threshold, switch to clamping mode, calculate the upper limit of the contact pressure of the clamping phalanx based on the friction coefficient range and constrain the initial closing angle of the phalanx. In the clamping mode, adjust the phalanx force closing direction based on the angle between the geometric model centroid projection and the clamping force line of action.

[0148] In step 502, the clamping mode is a gripping method activated when the material hardness exceeds the flexibility threshold, suitable for harder or regularly shaped items. This mode utilizes mechanical knuckles for a stable grip. The upper limit of contact pressure is the maximum permissible clamping pressure determined based on the coefficient of friction to prevent damage from excessive compression. A higher coefficient of friction allows for greater clamping pressure. The initial knuckle closing angle refers to the angle setting when the gripper begins to move, related to the circumscribed cube size, to ensure the optimal gripping position. The force closing direction refers to the direction adjusted according to the angle between the geometric model's center of gravity projection and the line of action of the clamping force, ensuring stable clamping.

[0149] In this embodiment, when the material hardness exceeds the flexibility threshold, the system switches to clamping mode. Using friction coefficient range data, the upper limit of the contact pressure on the clamping knuckles is calculated through mechanical analysis, and the initial closing angle of the knuckles is set to be proportional to the diagonal length of the outer cube dimension of the package. Simultaneously, the closing direction of the knuckle force is adjusted according to the angle between the geometric model's centroid projection and the line of action of the clamping force to ensure optimal gripping stability. This step involves the application of complex mechanical simulations and path planning algorithms. The final output is an optimized clamping strategy configuration, ensuring efficient and safe gripping operations.

[0150] 503. In the adsorption mode, the adsorption contact surface texture is matched according to the lower limit of the friction coefficient range, and the number of auxiliary adsorption pores activated is increased in local low friction areas.

[0151] In step 503, the lower limit of the friction coefficient range refers to the lowest friction coefficient value in the friction coefficient distribution of the wrapping surface. This value is used to determine the specific areas where enhanced adsorption is needed to ensure stable adsorption even under low-friction conditions. The number of auxiliary adsorption pores activated refers to the number of additional adsorption pores added in local low-friction areas. These pores can automatically open or close under specific conditions to improve the overall adsorption effect. Adsorption contact surface texture refers to the microstructural features of the wrapping surface; different textures may affect the adsorption effect. By analyzing these texture features, the design and configuration of the adsorption device can be optimized.

[0152] In this embodiment, during adsorption mode, the texture features of the adsorption contact surface are matched using image recognition technology based on the lower limit of the friction coefficient range. For detected localized low-friction areas, the number of auxiliary adsorption pores is increased to enhance adsorption force. This step involves the application of computer vision technology and data analysis methods, such as convolutional neural networks for texture recognition. Ultimately, a comprehensive adsorption layout scheme comprising main and auxiliary adsorption pores is formed, ensuring reliable adsorption even on complex surfaces.

[0153] 504. When the friction coefficient range crosses the critical value, the mixing mode is activated. The adsorption pore center area maintains adsorption, and the clamping knuckles are deployed in the edge area. The initial closing angle of the knuckles is positively correlated with the diagonal length of the outer cube size.

[0154] In step 504, the hybrid mode is a mode activated when the friction coefficient crosses a critical value. It combines the advantages of both adsorption and clamping methods, providing a more flexible gripping solution. Adsorption is maintained in the central area of ​​the adsorption pores in hybrid mode, providing basic adsorption force support. Clamping knuckles are deployed at the edges to reinforce the package, ensuring gripping stability and safety. The initial closure angle of the knuckles is positively correlated with the diagonal length of the circumscribed cube, meaning the initial angle of the knuckles is adjusted according to the package size to accommodate packages of different dimensions.

[0155] In this embodiment, when the coefficient of friction crosses a critical value, the system automatically enters a hybrid mode. At this time, the central area of ​​the adsorption pore continues to maintain its adsorption function, while the edge area deploys gripping knuckles. The initial closing angle of the knuckles is adjusted according to the diagonal length of the outer cubic dimension of the package to ensure stability and safety of the grip. This process requires precise sensor data and complex control algorithms, such as a controller for real-time adjustment of the gripping force. Ultimately, a gripping solution that balances the advantages of both adsorption and gripping is generated, ensuring an efficient and safe operating procedure.

[0156] Here is a specific example:

[0157] In a busy logistics center, when an intelligent logistics robot begins its work, it first performs a detailed scan of the cardboard boxes entering the sorting line, identifying their material hardness level and coefficient of friction range. If the material hardness is below the flexibility threshold, it activates the adsorption mode and dynamically adjusts the distribution density of the adsorption pore array according to the coefficient of friction range. If the material hardness exceeds the flexibility threshold, it switches to the gripping mode, calculates the upper limit of the contact pressure of the gripping knuckles, and sets the initial closing angle of the knuckles. For packages with complex coefficient of friction distributions, the robot may enter a hybrid mode, using adsorption in the central area and gripping knuckles in the edge areas. All these adjustments are based on real-time monitoring data and advanced algorithms, ensuring an efficient and safe operating process.

[0158] In summary, steps 501 to 504 significantly enhance the ability of intelligent logistics robots in the logistics center to handle different types of packages. This not only improves sorting efficiency but also effectively reduces the risk of damage to goods, achieving both safety and precision in automated operations. By dynamically adjusting the gripping mode, the robot can flexibly respond to various situations, greatly enhancing the flexibility and reliability of the entire logistics system. In particular, through detailed analysis of material hardness and friction coefficients, it ensures that each operation reaches its optimal state, thereby guaranteeing work efficiency while maximizing the protection of goods. This method, through meticulous data analysis and intelligent adjustment mechanisms, ensures the optimization of robot operation.

[0159] To further address the issue of selecting appropriate gripping modes for intelligent logistics robots when handling packages with different friction coefficients and geometric shapes, this solution updates the robot's motion parameters and the priority weights of dynamic sorting decisions using artificial intelligence algorithms. The aim is to continuously optimize the robot's behavior patterns, making it more adaptable to actual work requirements. Based on learning and analysis of historical data, the system can self-adjust, improving the accuracy and rationality of its decisions. This method significantly enhances the adaptability and work efficiency of the logistics robot. In some embodiments, step 502, which involves calculating the upper limit of the contact pressure of the gripping phalanx based on the friction coefficient range and constraining the initial closing angle of the phalanx, includes:

[0160] 601. Set the maximum contact pressure threshold for clamping the knuckles according to the upper limit of the friction coefficient range. When the friction coefficient crosses the high and medium friction ranges, adopt a segmented increasing strategy to increase the pressure threshold growth rate.

[0161] In step 601, the upper limit of the friction coefficient range represents the maximum value of the friction coefficient of the package surface, used to set the maximum contact pressure threshold for the gripping knuckles. A high friction coefficient allows for higher contact pressure without slippage. The maximum contact pressure threshold is the maximum pressure that the gripping knuckles can apply, determined based on the upper limit of the friction coefficient range, ensuring no damage to the package. This threshold is derived from experimental data or material properties. The segmented incremental strategy refers to a method of increasing the pressure threshold rate when the friction coefficient crosses high and medium friction domains to adapt to the gripping needs under different friction conditions. This strategy aims to optimize the gripping effect and avoid over- or under-pressure.

[0162] In this embodiment, the friction coefficient range of the cardboard box surface is first measured using sensors, and the maximum contact pressure threshold for the gripping knuckles is set based on its upper limit. Specifically, a machine learning model is used to analyze historical datasets and predict suitable pressure thresholds under different friction coefficient conditions. If the friction coefficient spans high and medium friction ranges, a segmented incremental strategy is adopted to gradually increase the rate of increase of the pressure threshold. The final result is an optimized maximum contact pressure threshold setting based on the current friction coefficient range, ensuring optimal gripping performance under various friction conditions.

[0163] 602. Calculate the safety margin of the initial closure angle of the knuckle based on the distribution density of the region with the maximum curvature in the geometric model. The safety margin is positively correlated with the product of the number of curvature extreme points and the lower limit of the friction coefficient.

[0164] In step 602, the distribution density of the region of maximum curvature refers to the number and distribution of regions with the greatest curvature change on the package surface, used to calculate the safety margin of the initial closure angle of the knuckle. High curvature regions may require more careful clamping. The allowable range of the initial closure angle of the knuckle with the safety margin is positively correlated with the product of the number of curvature extrema and the lower limit of the coefficient of friction, ensuring the safety of the clamping operation. The larger the safety margin, the more stable the clamping. The number of curvature extrema refers to the number of points on the package surface where the curvature change is most significant, used to assess the complexity of the package. More extrema mean that more precise clamping control is required.

[0165] In this embodiment, a scanning technique is used to acquire the geometric model of the package and analyze the distribution density of the region with the greatest curvature. Based on this data, the safety margin of the initial closure angle of the knuckles is calculated. The safety margin is proportional to the product of the number of curvature extrema and the lower limit of the friction coefficient. This step involves complex geometric analysis and mechanical simulation, using finite element analysis to simulate the stress distribution under different clamping angles, ensuring that the knuckles can close in the optimal position and reducing potential damage to the package. The final output is an optimized initial closure angle configuration scheme for the knuckles, ensuring efficient and safe clamping operation.

[0166] 603. Based on the shortest distance from the boundary of the fragile area to the clamping area in the surface stress distribution data, the maximum contact pressure threshold is dynamically attenuated, and the attenuation coefficient is inversely proportional to the square of the shortest distance;

[0167] In step 603, the surface stress distribution data refers to the stress distribution at various points on the package surface, used to identify fragile areas and their boundaries. This data helps adjust the clamping force to avoid damaging the package. The closest distance from the fragile area boundary to the clamping area refers to the minimum distance between the fragile area boundary and the actual clamping area, used to dynamically attenuate the maximum contact pressure threshold. The closer the distance, the lower the pressure should be. The attenuation coefficient is inversely proportional to the square of the closest distance, used to adjust the maximum contact pressure threshold to ensure that fragile areas are not subjected to excessive pressure. This proportional relationship ensures gradual pressure adjustment.

[0168] In this embodiment, stress distribution data of the package surface is acquired using stress sensors or simulation software, and fragile areas and their boundaries are identified. Based on this information, the shortest distance from the boundary of the fragile area to the clamping area is calculated, and the maximum contact pressure threshold is dynamically attenuated accordingly. The attenuation coefficient is inversely proportional to the square of the shortest distance, ensuring that pressure is appropriately reduced when approaching the fragile area to avoid damage. This step involves the application of data analysis and real-time monitoring technologies, using a controller for precise pressure adjustment. Ultimately, a pressure threshold adjustment scheme adapted to the current package characteristics is formed to ensure the safety of the fragile area during clamping.

[0169] 604. Based on the safety margin, when the upper limit of the contact pressure of the clamping phalanx reaches the threshold warning range, a high-frequency micro-amplitude vibration of the contact surface is triggered, and the pressure gradient distribution of the clamping phalanx is monitored in real time. If the pressure concentration area is detected to overlap with the area with the maximum curvature in the geometric model, the opening compensation angle of the adjacent phalanx is increased to disperse the stress.

[0170] In step 604, the threshold warning interval refers to the warning interval triggered when the upper limit of the contact pressure of the clamping phalanx approaches the set maximum value, used to initiate further safety measures. This interval provides early warning to prevent overpressure. High-frequency micro-amplitude vibration is a slight vibration method used to disperse local stress concentrations generated during clamping. Vibration helps to evenly distribute pressure and prevent local overload. The opening compensation angle refers to increasing the opening angle of adjacent phalanxes when the detected pressure concentration area overlaps with the area of ​​maximum curvature in the geometric model, in order to disperse stress. This method can effectively avoid damage caused by local stress concentration.

[0171] In this embodiment, when the upper limit of the contact pressure on the clamping knuckle reaches the threshold warning range, the system automatically triggers high-frequency micro-amplitude vibration of the contact surface. This process is achieved through a micro-vibration device installed on the knuckle. Simultaneously, the pressure gradient distribution of the clamping knuckle is monitored in real time. If the pressure concentration area overlaps with the area of ​​maximum curvature in the geometric model, the opening compensation angle of adjacent knuckles is immediately increased to disperse stress. This step involves the application of advanced sensing technology and feedback control mechanisms, such as using capacitive sensors for real-time pressure monitoring. The final output is an optimized clamping strategy configuration scheme, ensuring an efficient and safe operating procedure.

[0172] Here is a specific example:

[0173] In a busy logistics center, when an intelligent logistics robot begins its work, it first performs a detailed scan of the cardboard boxes entering the sorting line, identifying their material hardness level and coefficient of friction range. If the cardboard box's material hardness exceeds a flexibility threshold, the robot switches to gripping mode. Next, the robot sets the maximum contact pressure threshold for the gripping joints based on the upper limit of the coefficient of friction range, and employs a segmented incremental strategy to increase the pressure threshold growth rate. Then, it uses scanning technology to analyze the geometric features of the package surface and calculates the safety margin of the initial closure angle of the joints. For identified fragile areas, the maximum contact pressure threshold is dynamically reduced to ensure that these areas are not subjected to excessive pressure. Finally, during gripping, if the contact pressure approaches the threshold warning range, the robot triggers high-frequency micro-vibration and monitors the pressure distribution in real time, increasing the opening compensation angle of adjacent joints as needed to ensure an efficient and safe operating procedure.

[0174] In summary, steps 601 to 604 significantly improve the gripping ability and safety of intelligent logistics robots when handling different types of packages. Through precise friction coefficient analysis, geometric feature recognition, and real-time monitoring and adjustment, the stability and reliability of the gripping operation are ensured. Especially when dealing with complex surfaces and fragile items, this method effectively reduces the risk of damage and improves the safety and accuracy of automated operations. Furthermore, by dynamically adjusting the contact pressure and knuckle angle, the robot can flexibly respond to various situations, greatly enhancing the flexibility and efficiency of the entire logistics system. This refined adjustment strategy not only ensures work efficiency but also maximizes the safety of goods.

[0175] To address the challenge of accurately acquiring the surface material type, three-dimensional geometric contour, and surface stress distribution of packages handled by intelligent logistics robots, this solution integrates force closure stability scores, temporal parameters, and joint torque balance factors to generate a multi-objective optimization weight set. The final output is a dynamic sorting strategy. The core idea is to formulate the optimal gripping plan by comprehensively considering factors such as mechanical stability, time synchronization, and mechanical safety. This method fully considers various variables during package handling, ensuring that each operation achieves optimal results and effectively improving the accuracy and safety of package processing. In some embodiments, step 101, which involves acquiring the surface material type data, three-dimensional geometric contour data, and surface stress distribution data of the package to be sorted by the logistics robot, includes:

[0176] 701. Collect the surface reflectance spectrum of the package using the multispectral sensor of the logistics robot, extract the multi-band spectral response difference to match the surface material type data, and enable polarization filtering to suppress interference in high reflective areas;

[0177] In step 701, a multispectral sensor is a device capable of acquiring the reflectivity of light at different wavelengths to identify the material of the package surface. Materials are distinguished by analyzing the differences in reflectivity across different wavelength bands. Surface material type data, extracted based on differences in multi-band spectral response, describes the type and characteristics of the package surface material. This data helps determine an appropriate grasping strategy. Polarization filtering is a technique used to suppress interference from highly reflective areas, improving the accuracy of material identification. By filtering out unnecessary reflected light, the identification of key information is enhanced.

[0178] In this embodiment, a multispectral sensor is first used to scan the packaged surface, acquiring reflectance spectra at different wavelengths. Machine learning algorithms are then employed to analyze these reflectance differences and match them to the surface material type. For highly reflective areas, polarization filtering technology is used to reduce interference from reflected light, ensuring accurate material identification. The final result is a precise dataset of surface material types, providing fundamental information for subsequent operations. This process involves complex optical analysis and the application of machine learning models, ensuring high-precision material identification.

[0179] 702. The three-dimensional point cloud is reconstructed by fusing structured light and stereo vision, and the curvature features are extracted to generate three-dimensional geometric contour data. Multi-angle scanning is initiated to complete the point cloud in the blind area.

[0180] In step 702, the structured light and stereo vision fusion reconstruction refers to combining structured light projection and stereo vision technology to generate a 3D point cloud model of the package. This method provides high-precision geometric information. The 3D geometric contour data is based on curvature features extracted from the 3D point cloud model, describing the shape and size of the package. This data helps plan the optimal grasping path. Blind spot completion refers to using multi-angle scanning methods to supplement the complete 3D point cloud data for areas that cannot be directly observed. This step ensures the integrity of the model.

[0181] In this embodiment, structured light projection technology is used to project a specific pattern onto the surface of the package, and images are captured by a stereo camera to generate an initial 3D point cloud model. Then, computer vision algorithms are used to extract curvature features from the model to generate 3D geometric contour data. For areas with blind spots, a multi-angle scanning strategy is initiated to supplement the point cloud data from different perspectives, ensuring model integrity. The final output is a 3D geometric contour dataset containing all details, providing accurate geometric information for clamping and handling. This process involves complex image processing and 3D reconstruction techniques to ensure high-precision geometric modeling.

[0182] 703. Apply pressure gradient through flexible tactile array, record deformation rate and stress relaxation curve to generate surface stress distribution data, trigger high-frequency sampling when stress change is detected, and obtain high-frequency sampled stress data.

[0183] In step 703, the flexible tactile array is a pressure sensor array mounted on the robotic gripper to detect pressure changes on the package surface. These sensors can record the pressure distribution in real time. Surface stress distribution data is used to record the deformation rate and stress relaxation curve, describing the stress state at various points on the package surface. This data helps optimize gripping force and avoid damage to the package. High-frequency sampling refers to triggering a high-frequency sampling mechanism when a sudden stress change is detected to capture more detailed pressure change information. This mechanism improves data resolution and reliability.

[0184] In this embodiment, a pressure gradient is applied using a flexible tactile array, and the deformation rate and stress relaxation curve of the wrapped surface are recorded in real time to generate surface stress distribution data. If a sudden stress change is detected, the system automatically switches to a high-frequency sampling mode, increasing the sampling frequency to capture more details. This step involves the application of advanced sensing technologies and data analysis methods, such as dynamic time warping algorithms used to process the stress curves. Ultimately, a detailed surface stress distribution dataset is formed, helping to avoid damage caused by excessive pressure. This process involves complex data acquisition and signal processing techniques to ensure high-precision stress monitoring.

[0185] 704. Synchronize the multi-band spectral response differences, multi-angle scan-completed point cloud, and high-frequency sampling stress data to the three-dimensional coordinate system to generate a cross-modal correlation mapping table.

[0186] In step 704, the cross-modal correlation mapping table is generated by synchronizing multi-band spectral response differences, multi-angle scan-completed point cloud data, and high-frequency sampled stress data to a unified three-dimensional coordinate system. This table provides comprehensive data integration, facilitating integrated analysis. The three-dimensional coordinate system refers to the spatial reference frame used to integrate various sensor data, ensuring data consistency and comparability. This coordinate system is the foundation of all data.

[0187] In this embodiment, the various data obtained in steps 701 to 703 are synchronized to a unified three-dimensional coordinate system. Specifically, coordinate transformation algorithms (such as homogeneous coordinate transformation) are used to align data from different sources and generate a cross-modal association mapping table. This process involves complex mathematical calculations and data fusion techniques to ensure that all information can be comprehensively analyzed within the same spatial reference frame. The final result is a comprehensive and consistent dataset, providing a solid foundation for subsequent decision-making and operations. This process involves advanced data fusion and coordinate transformation techniques, ensuring a high degree of data consistency.

[0188] Here is a specific example:

[0189] In a busy logistics center, when an intelligent logistics robot begins its work, it first performs a detailed scan of the cardboard boxes entering the sorting line. The robot uses multispectral sensors to collect the reflectivity spectrum of the package surface. By analyzing the differences in multi-band spectral responses, it applies polarization filtering to suppress interference in highly reflective areas. Next, it uses structured light and stereo vision fusion reconstruction technology to generate a 3D point cloud model, and initiates multi-angle scanning to complete the point cloud for blind spots. Subsequently, it applies pressure gradients through a flexible tactile array, records the deformation rate and stress relaxation curves to generate surface stress distribution data, and triggers high-frequency sampling when a stress mutation is detected. Finally, all collected data is synchronized to a 3D coordinate system to generate a cross-modal correlation mapping table, providing accurate data support for subsequent efficient sorting.

[0190] In summary, steps 701 to 704 significantly enhance the intelligent logistics robot's ability to acquire key information when handling different types of packages. This not only improves sorting efficiency but also effectively reduces the risk of damage to goods, achieving both safety and accuracy in automated operations. The application of advanced technologies such as multispectral sensors, structured light and stereo vision fusion reconstruction, and flexible tactile arrays ensures high precision and reliability in material recognition, geometric modeling, and stress monitoring. In particular, cross-modal data fusion enables the robot to flexibly handle various complex situations, greatly enhancing the flexibility and efficiency of the entire logistics system. This method not only guarantees work efficiency but also maximizes the protection of goods. This sophisticated data acquisition and processing strategy ensures that each package is handled safely and efficiently.

[0191] Figure 2 This application provides a schematic diagram of the structure of an artificial intelligence-based logistics robot control system, as shown in the embodiment. Figure 2 As shown, the system includes:

[0192] Module 21 acquires surface material type data, three-dimensional geometric contour data, and surface stress distribution data of the packages to be sorted through the logistics robot;

[0193] The determination module 22 determines the material hardness level and friction coefficient range of the package to be sorted based on the surface material type data.

[0194] The identification module 23 reconstructs the geometric model of the package to be sorted based on the three-dimensional geometric contour data and calculates the minimum outer cube size of the package. It also identifies the fragile area boundary of the surface of the package to be sorted based on the surface stress distribution data.

[0195] The generation module 24 adjusts the adsorption mode or clamping mode of the logistics robot according to the material hardness level and friction coefficient range, adjusts the unfolding angle of the logistics robot according to the minimum package outer cube size, and constrains the contact point distribution and movement trajectory of the logistics robot's grasping action according to the fragile area boundary, and generates a dynamic sorting strategy by combining the current sorting line flow rate, target sorting area capacity and robot joint load status.

[0196] The update module 25 updates the dynamic sorting strategy of the logistics robot by using an artificial intelligence algorithm based on the execution result of the dynamic sorting strategy.

[0197] Figure 2 The aforementioned AI-based logistics robot control system can execute... Figure 1The implementation principle and technical effects of the AI-based logistics robot control method described in the illustrated embodiment will not be repeated here. The specific methods by which each module and unit of the AI-based logistics robot control system in the above embodiments perform operations have been described in detail in the embodiments related to this method, and will not be elaborated upon here.

[0198] In one possible design, Figure 2 The AI-based logistics robot control system shown in the embodiment can be implemented as a computing device, such as... Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32;

[0199] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are invoked and executed by the processing component 32.

[0200] The processing component 32 is used for the above Figure 1 The embodiment describes an artificial intelligence-based control method for logistics robots.

[0201] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above-described method. Alternatively, the processing component may be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above-described method.

[0202] Storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0203] Of course, computing devices may also include other components, such as input / output interfaces, display components, communication components, etc.

[0204] Input / output interfaces provide interfaces between processing components and peripheral interface modules, which can be output devices, input devices, etc.

[0205] The communication components are configured to facilitate wired or wireless communication between computing devices and other devices.

[0206] The computing device can be a physical device or an elastic computing host provided by a cloud computing platform. In this case, the computing device can refer to a cloud server, and the aforementioned processing components, storage components, etc., can be basic server resources rented or purchased from the cloud computing platform.

[0207] This application also provides a computer storage medium storing a computer program, which, when executed by a computer, can perform the above-described functions. Figure 1 The embodiment shown illustrates an artificial intelligence-based control method for logistics robots.

[0208] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0209] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0210] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0211] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A control method for logistics robots based on artificial intelligence, characterized in that, The method comprises the following steps: acquiring surface material type data, three-dimensional geometric profile data, and surface stress distribution data of a package to be sorted by a logistics robot; determining the material hardness level and the friction coefficient interval of the package to be sorted based on the surface material type data; reconstructing a geometric model of the package to be sorted based on the three-dimensional geometric profile data and calculating the minimum package circumscribed cube size, and identifying the fragile area boundary of the surface of the package to be sorted based on the surface stress distribution data; adjusting the adsorption mode or clamping mode of the logistics robot according to the material hardness level and the friction coefficient interval, adjusting the unfolding angle of the logistics robot in combination with the minimum package circumscribed cube size, and simultaneously constraining the contact point distribution and motion trajectory of the grabbing action of the logistics robot according to the fragile area boundary, and generating a dynamic sorting strategy in combination with the current sorting line flow rate, the target sorting area capacity, and the joint load state of the robot; updating the action parameters of the logistics robot and the priority weight of the dynamic sorting strategy based on the execution result of the dynamic sorting strategy by an artificial intelligence algorithm to update the dynamic sorting strategy of the logistics robot; the step of generating a dynamic sorting strategy according to the fragile area boundary to constrain the contact point distribution and motion trajectory of the grabbing action of the logistics robot, and in combination with the current sorting line flow rate, the target sorting area capacity, and the joint load state of the robot, comprises the following steps: based on the fragile area boundary, demarcating a contact point candidate area on the surface of the geometric model, excluding coordinate points outside the fragile area boundary with an expanded safety distance to generate an initial contact point set; screening a candidate contact point pair that meets the grabbing force direction coincidence degree condition with the package gravity center projection from the initial contact point set, and calculating the force closure stability score of the candidate contact point pair; decomposing the grabbing motion trajectory in the geometric model coordinate system into a transverse translation component and a longitudinal following component, calculating the velocity compensation amount of the transverse translation component based on the sorting line flow rate, and generating the timing parameters of the longitudinal following component; calculating the joint torque balance factor of the candidate contact point pair according to the inverse of the real-time current data and the historical load ratio of the joint of the logistics robot, fusing the force closure stability score, the timing parameters, and the joint torque balance factor to generate a multi-objective optimization weight set, and outputting a dynamic sorting strategy based on the multi-objective optimization weight set.

2. The method of claim 1, wherein, the step of fusing the force closure stability score, the timing parameters, and the joint torque balance factor to generate a multi-objective optimization weight set, comprises the following steps: determining the initial weight distribution of the force closure stability score, the timing parameters, and the joint torque balance factor based on the material hardness level and the sorting line flow rate, and increasing the score weight when the material hardness level is higher than a threshold value; converting the force closure stability score into a standardized interval, decomposing the timing parameters into a synchronization error tolerance and a time window matching degree, and generating a multi-dimensional parameter vector in combination with the joint torque balance factor; correcting the initial weight distribution according to the target sorting area capacity change rate, superimposing the corrected initial weight distribution and the multi-dimensional parameter vector to generate a candidate scheme comprehensive score; and The joint torque balance factor is truncated according to the difference between the real-time current peak value of the joint of the logistics robot and the historical load threshold value, and a multi-objective optimization weight set is generated based on the candidate scheme comprehensive score.

3. The method of claim 2, wherein, The force closure stability score is converted into a standardized interval, the timing parameter is decomposed into synchronization error tolerance and time window matching degree, and a multi-dimensional parameter vector is generated in combination with the joint torque balance factor, which includes: The force closure stability score is dynamically mapped to a standardized interval based on the scoring distribution range of the historical picking case library; When decomposing the timing parameter, the synchronization error tolerance is calculated according to the ratio of the sorting line flow rate to the size of the package circumscribed cube, and the time window matching degree is generated based on the theoretical motion time of the geometric model gravity center to the sorting container opening; The joint torque balance factor is dynamically gain-adjusted according to the fluctuation variance of the joint real-time current data and the historical average current; The standardized interval, synchronization error tolerance, time window matching degree and adjusted joint torque balance factor are dimensionally aligned according to the logistics scene priority rules to generate a multi-dimensional parameter vector containing mechanical stability, timing synchronization and mechanical safety.

4. The method of claim 1, wherein, The adjustment of the absorption mode or the clamping mode of the logistics robot according to the material hardness grade and the friction coefficient interval includes: An absorption force safety threshold is set based on the material hardness grade, and when the material hardness grade is lower than the flexibility threshold, the absorption mode is activated, and the distribution density of the absorption hole array is dynamically adjusted according to the friction coefficient interval; When the material hardness grade is higher than the flexibility threshold, the clamping mode is switched to, the upper limit of the contact pressure of the clamping finger is calculated based on the friction coefficient interval, and the initial closing angle of the finger is constrained, and in the clamping mode, the finger force closure direction is adjusted based on the projection of the geometric model gravity center and the angle between the clamping force action line; In the absorption mode, the absorption contact surface texture is matched according to the lower limit value of the friction coefficient interval, and the number of activated auxiliary absorption holes in the local low-friction area is increased; When the friction coefficient interval crosses the critical value, the mixed mode is started, the absorption holes in the central area are kept absorbing, and the clamping fingers are deployed in the edge area, and the initial closing angle of the fingers is positively correlated with the diagonal length of the circumscribed cube.

5. The method of claim 4, wherein, The adjustment of the absorption mode or the clamping mode of the logistics robot according to the material hardness grade and the friction coefficient interval includes: The maximum contact pressure threshold of the clamping finger is set according to the upper limit value of the friction coefficient interval, and when the friction coefficient crosses the high and medium friction domain, a segmented incremental strategy is adopted to increase the pressure threshold growth rate; The safety margin of the initial closing angle of the finger is calculated based on the distribution density of the maximum curvature region in the geometric model, and the safety margin is positively correlated with the product of the number of curvature extreme points and the lower limit value of the friction coefficient; The maximum contact pressure threshold is dynamically attenuated based on the nearest distance from the fragile area boundary to the clamping area in the surface stress distribution data, and the attenuation coefficient is inversely proportional to the square of the nearest distance. Based on the safety margin, when the upper limit of the contact pressure of the clamping fingers reaches a threshold early warning interval, high-frequency micro-amplitude vibration of the contact surface is triggered to monitor the pressure gradient distribution of the clamping fingers in real time, and if it is detected that the pressure concentration area overlaps with the maximum curvature area in the geometric model, the opening compensation angle of the adjacent fingers is increased to disperse stress.

6. The method of claim 1, wherein, The surface material type data, three-dimensional geometric contour data, and surface stress distribution data of the package to be sorted are obtained by the logistics robot, including: The multispectral sensor of the logistics robot collects the reflectivity spectrum of the package surface, extracts the multi-band spectral response difference to match the surface material type data, and enables polarization filtering in the high-reflectivity area to suppress interference; Three-dimensional point cloud is reconstructed by fusing structured light and stereo vision, and three-dimensional geometric contour data is extracted by generating curvature features, and multi-angle scanning is started in the blind area to complete the point cloud; The surface stress distribution data is generated by recording the deformation rate and stress relaxation curve through the flexible tactile array loading pressure gradient, and high-frequency sampling is triggered when stress mutation is detected to obtain high-frequency sampling stress data; The multi-band spectral response difference, multi-angle scanning completed point cloud, and high-frequency sampling stress data are synchronized to the three-dimensional coordinate system to generate a cross-modal correlation mapping table.

7. An artificial intelligence-based logistics robot control system applied to the artificial intelligence-based logistics robot control method of any one of claims 1 to 6, characterized in that, It includes: An acquisition module obtains surface material type data, three-dimensional geometric contour data, and surface stress distribution data of a package to be sorted by a logistics robot; A determination module determines the material hardness grade and friction coefficient interval of the package to be sorted based on the surface material type data; An identification module reconstructs a geometric model of the package to be sorted based on the three-dimensional geometric contour data and calculates the minimum package circumscribed cube size, and identifies the fragile area boundary of the surface of the package to be sorted based on the surface stress distribution data; A generation module adjusts the adsorption mode or clamping mode of the logistics robot according to the material hardness grade and friction coefficient interval, adjusts the deployment angle of the logistics robot in combination with the minimum package circumscribed cube size, and simultaneously constrains the contact point distribution and motion trajectory of the gripping action of the logistics robot according to the fragile area boundary, and generates a dynamic sorting strategy in combination with the current sorting line flow rate, target sorting area capacity, and robot joint load state; An update module updates the action parameters of the logistics robot and the priority weight of the dynamic sorting strategy based on the execution result of the dynamic sorting strategy by an artificial intelligence algorithm to update the dynamic sorting strategy of the logistics robot.

8. A computing device, comprising: It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to realize the logistics robot control method based on artificial intelligence in any one of claims 1-6.

9. A computer storage medium, characterized in that A computer program is stored, and when the computer program is executed by a computer, a logistics robot control method based on artificial intelligence in any one of claims 1-6 is realized.

Images