Coffee robot mechanical arm control method and system based on knowledge graph

Abstract

The application discloses a coffee robot mechanical arm control method and system based on a knowledge graph, relates to the technical field of mechanical arm control, and aims to solve the technical problems that in the prior art, displacement deviation of mechanical joints and offset of matching points are prone to cause coffee spilling, and low coordination efficiency between an execution end and a controlled object further aggravates operation failure, specifically, overall system covers execution end stability monitoring, executed end matching degree analysis, double-end fault tolerance fusion judgment, and abnormal operation traceability regulation, from mechanical action deviation, container positioning error to material spilling and process abnormality, and through threshold comparison and trend analysis, early prediction and real-time intervention are realized, so that the probability of coffee production failure is significantly reduced; taking the knowledge graph as the core, not only can execution logic be updated based on real-time data, but also can environment and raw material parameter traceability abnormal inducement be combined to realize dynamic adaptation of "equipment aging adaptation-process adjustment compatibility-raw material change response".

Description

Technical Field

[0001] This invention relates to the field of robotic arm control technology, specifically to a knowledge graph-based control method and system for a coffee robot robotic arm. Background Technology

[0002] Currently, the control system of coffee robot robotic arms suffers from insufficient execution accuracy and coordination in complex production scenarios. The system mainly relies on preset programs to control the robotic arm's movements, lacking the ability to perceive and adapt to dynamic changes during execution. In continuous production, displacement deviations of mechanical joints and misalignments of mating points can easily lead to coffee spillage, while low coordination efficiency between the execution end and the controlled object can further exacerbate operational errors.

[0003] It is difficult to identify the unstable operating state of the robotic arm in real time, and it is also impossible to effectively judge the coordination efficiency between devices, resulting in poor fault tolerance of the system.

[0004] Meanwhile, the fixed control logic cannot be dynamically adjusted according to changes in the actual production environment. When there are abnormalities in the raw material status or fluctuations in environmental parameters, the system lacks an effective mechanism for anomaly identification and logic update.

[0005] Therefore, a solution is proposed to address the practical situation in the existing technology. Summary of the Invention

[0006] The purpose of this invention is to solve the problems mentioned above by proposing a knowledge graph-based control method and system for a coffee robot robotic arm.

[0007] The objective of this invention can be achieved through the following technical solution: a coffee robot robotic arm control system based on knowledge graph, which includes a coffee machine operation system and a mechanical control platform and a knowledge graph building terminal.

[0008] The mechanical control platform has the following communication connections:

[0009] The execution analysis unit performs execution analysis on the coffee making process and obtains the risky and safe running segments based on the execution analysis.

[0010] The execution end analysis unit analyzes the execution end and determines the efficient and inefficient cooperation stages based on the analysis.

[0011] The fusion analysis unit performs fusion analysis on risky runtime segments and inefficient coordination stages, and infers the fault tolerance performance of the execution end and the executed end based on the fusion analysis;

[0012] The communication connections for building the knowledge graph include:

[0013] The feasibility analysis unit is updated to perform an update feasibility analysis on the operation and execution logic of knowledge graph construction. The knowledge graph is divided into execution logic and preparation logic, and the logic is updated according to the real-time execution actions.

[0014] The abnormal operation identification unit identifies abnormal operations during the execution process of the knowledge graph; it performs synchronous analysis by combining the execution logic and the preparation logic, and makes logical execution decisions for the knowledge graph based on the analysis results.

[0015] Furthermore, the process of the execution-side analysis unit is as follows:

[0016] The structure is divided to identify the wrist and elbow joints at the execution end, which are uniformly marked as execution joints. Based on the execution joints, the action coordination points at the execution end are determined, and the action trajectory of each joint at the execution end is determined.

[0017] The displacement deviation of the joint movement trajectory within the action space during continuous action execution, as well as the offset distance of the mate point in the position during the continuous execution phase of the action, are obtained and compared with the corresponding thresholds.

[0018] If the displacement deviation of the joint movement trajectory in the action space exceeds the displacement deviation threshold during continuous execution of the action, or if the offset distance of the mate point in the continuous execution phase of the action exceeds the offset distance threshold, it is inferred that the execution end is currently in an unstable continuous operation phase.

[0019] If the displacement deviation of the joint movement trajectory within the action space during continuous execution of the action does not exceed the displacement deviation threshold, and the offset distance of the mating point within the continuous execution phase of the action does not exceed the offset distance threshold, then it is inferred that the execution end is currently in a stable and continuous operation phase.

[0020] Furthermore, the frequency of the angle drift during the unstable continuous operation phase was obtained, and the sensitivity of the internal force feedback during the stable continuous operation phase was also collected.

[0021] If the execution angle drift frequency shows an increasing trend during the unstable continuous operation phase, or the internal force feedback sensitivity shows a fluctuating trend during the stable continuous operation phase, then the current period is set as a risky operation phase, and a risky operation signal is generated; if the execution angle drift frequency does not show an increasing trend during the unstable continuous operation phase, and the internal force feedback sensitivity does not show a fluctuating trend during the stable continuous operation phase, then the current period is set as a safe operation phase, and a safe operation signal is generated.

[0022] Furthermore, the process of the executed end analysis unit is as follows:

[0023] The distance offset between the real-time position of the executed end and the execution position of the executing end is obtained, and the deviation frequency of the real-time position positioning detection of the executed end is collected.

[0024] If the distance offset between the real-time position of the executed end and the execution position of the executing end exceeds the set distance offset threshold, or if the deviation frequency of the real-time position positioning detection of the executed end exceeds the detection deviation frequency threshold, it is inferred that the executed end is in a stage of inefficient cooperation, and a cooperation control signal is generated and sent to the mechanical control platform.

[0025] If the distance offset between the real-time position of the executed end and the execution position of the executing end does not exceed the set distance offset threshold, and the deviation frequency of the real-time position positioning detection of the executed end does not exceed the detection deviation frequency threshold, it is inferred that the executed end is in the efficient cooperation stage, generates a stable cooperation signal and sends it to the mechanical control platform.

[0026] Furthermore, the process of merging analysis units is as follows:

[0027] The overlapping periods of the risky running phase and the inefficient coordination phase are extracted and marked as the fault tolerance analysis period; the percentage of real-time production loss of the executed end when the execution end is offset during the fault tolerance analysis period is obtained, specifically the amount of coffee spilled; the deviation of the clamping angle of the execution end when the position of the executed end is offset during the fault tolerance analysis period is obtained.

[0028] If the percentage of real-time production loss at the executed end exceeds the allowable loss threshold when the executed end performs an offset during the fault tolerance analysis period, or if the deviation of the clamping angle at the executed end exceeds the angle deviation threshold when the executed end's position shifts during the fault tolerance analysis period, a low fault tolerance signal is generated and sent to the mechanical control platform. If the percentage of real-time production loss at the executed end does not exceed the allowable loss threshold when the executed end performs an offset during the fault tolerance analysis period, and the deviation of the clamping angle at the executed end does not exceed the angle deviation threshold when the executed end's position shifts during the fault tolerance analysis period, a high fault tolerance signal is generated and sent to the mechanical control platform.

[0029] Furthermore, the process of updating the feasibility analysis unit is as follows:

[0030] When trajectory deviation occurs during the execution of successive actions in the execution logic, the success rate deviation in the preparation logic after the successive action order is swapped is obtained, and the corresponding successive action order swap is used as the knowledge graph update. The impact data of the execution logic corresponding to the successive actions is collected, and the impact data is set as the trigger condition for the knowledge graph update execution. When the knowledge graph is executed according to the original set logic, the trigger adjustment detection is performed. After the trigger is generated, the knowledge graph is replaced and a high feasibility signal is generated and sent to the knowledge graph building end. The knowledge graph building end transmits the logic within the knowledge graph determined in real time to the mechanical control platform.

[0031] Furthermore, the process of the abnormal operation identification unit is as follows:

[0032] By combining execution logic and preparation logic for synchronous analysis, and comparing the sensor feedback and expected state results of each key node in the coffee making process in real time, the system obtains the identifiers of each abnormal execution link, expected state parameters, actual state parameters, and abnormal type labels during the execution of the knowledge graph logic; and constructs the abnormal characteristics of the execution link. Environmental parameters are obtained in real time by temperature and humidity sensors deployed in the production area, constructing the environmental characteristics of the execution environment, including ambient temperature, ambient humidity, and environmental state standards. Real-time parameters of key raw materials are obtained from the raw material management system, constructing the raw material state characteristics of the execution link, including coffee bean moisture content, milk temperature, and raw material freshness labels.

[0033] Abnormal operation features are extracted based on the historical logic execution process of the knowledge graph. Abnormal operation features are compared with the current execution stage. If there are overlapping abnormal features, the causes of the current overlapping abnormal features are extracted based on the historical logic execution process, and a set of cause data is constructed. The cause extraction direction is environmental and raw material.

[0034] Furthermore, a corresponding analysis is performed on the collected data and the causal data set based on the current environmental characteristics and raw material state characteristics: the fluctuation trend of any type of collected data and the fluctuation trend of the causal data set are obtained, and the trend overlap rate is obtained by comparing the fluctuation trends. That is, the fluctuation trend comparison uses the current fluctuation speed and fluctuation span as comparison parameters, and the highest value as the comparison standard; the rate of decrease of the interval value corresponding to the fluctuation speed or fluctuation span is set as the trend overlap rate.

[0035] Furthermore, if the trend overlap rate continues to increase, the anomaly type label is extracted from the overlapping anomaly features, and the anomaly execution link identifier, expected state parameter, and actual state parameter corresponding to the anomaly type label are obtained from the operation log. The anomaly type label is set as a high-weight anomaly type, and the parameter deviation frequency is statistically analyzed by comparing the expected state parameter and the actual state parameter. When the parameter deviation frequency exceeds the set frequency threshold, the high-weight anomaly type and the corresponding anomaly execution link identifier are sent to the knowledge graph building end. If the knowledge graph building end has no updated execution logic, a mechanical control signal is generated and sent to the mechanical control platform. The mechanical control platform performs corresponding link execution control according to the high-weight anomaly type and the corresponding anomaly execution link identifier to reduce the probability of anomaly generation.

[0036] This invention also proposes a knowledge graph-based control method for a coffee robot's robotic arm. The specific control method steps are as follows:

[0037] Execution-side analysis: Perform execution-side analysis on the coffee making process to identify risky and safe operating segments based on the execution-side analysis.

[0038] The execution end is analyzed to identify the efficient and inefficient cooperation stages.

[0039] The fusion analysis sheet performs fusion analysis on risky runtime segments and inefficient coordination stages, and infers the fault tolerance performance of the execution end and the executed end based on the fusion analysis;

[0040] The feasibility analysis was updated, and the operational execution logic of the knowledge graph construction was updated. The knowledge graph was divided into execution logic and preparation logic, and the logic was updated according to the real-time execution actions.

[0041] Anomaly detection involves identifying abnormal operations during the execution of the knowledge graph; combining the execution logic and preparation logic for synchronous analysis; and using the analysis results to make logical execution decisions for the knowledge graph.

[0042] Compared with the prior art, the beneficial effects of the present invention are:

[0043] 1. This invention can effectively identify the stability changes of the robotic arm during continuous movements, promptly distinguish between stable operation and abnormal shaking, and provide reliable data support for subsequent risk period determination by quantifying the correlation between joint motion deviation and end-point position offset, thereby reducing the waste of raw materials caused by the loss of control of the robotic arm during coffee making.

[0044] This invention also enables early warning of robotic arm performance degradation by superimposing angle drift frequency trend analysis and force feedback sensitivity monitoring, thus avoiding control lag caused by misjudgment of a single parameter. Simultaneously, it can dynamically identify potential risk periods in the robotic arm control system, triggering timely warnings when the angle drift frequency increases or the force feedback sensitivity becomes abnormal, thereby reducing the probability of material spillage or operational errors due to robotic arm malfunction during coffee making. By distinguishing safe operating periods, unnecessary control logic switching can be avoided, maintaining efficient system operation in a stable state.

[0045] This invention can effectively identify abnormal states in the cooperation between the robotic arm and the target object, and promptly trigger adjustments to the control strategy to avoid problems such as coffee spillage and container collisions caused by positional deviation or misalignment. At the same time, it maintains a stable operating rhythm during the high-efficiency phase, improving the coordination and reliability of the coffee preparation process.

[0046] This invention also integrates time-period analysis and multi-dimensional parameter monitoring to more accurately identify system-level fault-tolerant bottlenecks, whereas traditional methods may misjudge such complex faults as being caused by a single factor. This effectively solves the fault-tolerant judgment problem caused by the disconnect between execution accuracy and the state of the controlled object in the robotic arm control system. In the coffee-making scenario, it can accurately identify the risk of coffee spillage caused by the combined effects of robotic arm vibration and material container offset, and adjust operating parameters in a timely manner to avoid material waste. At the same time, through a dual monitoring mechanism of gripping angle, it prevents tipping accidents caused by cup positioning deviation, improving the system robustness under complex working conditions. Attached Figure Description

[0047] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0048] Figure 1 This is a system logic architecture diagram of the present invention;

[0049] Figure 2 This is a system principle block diagram of the present invention. Detailed Implementation

[0050] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0052] In existing technologies, the control of coffee robot robotic arms mostly relies on preset programs to execute fixed action processes, lacking the ability to analyze the dynamic state of the execution end in real time. When there is a deviation in the coordination between the robotic arm and the equipment being executed, such as the material table and coffee machine, the system has difficulty adjusting the operating parameters in time, resulting in interruption of the production process or waste of raw materials. For example, if the joint displacement of the robotic arm is not identified in time during continuous production, it may cause problems such as coffee spillage or failure to pick up materials, affecting the overall control efficiency.

[0053] To address the aforementioned issues, the inventors discovered that conventional systems cannot effectively integrate the analysis results of the execution end's action state and the execution end's coordination state. By studying the correlation between the robotic arm's motion trajectory and the material platform's position, they proposed dividing the control system into two core modules: a mechanical control platform and a knowledge graph construction end. The former identifies risky operating periods and inefficient coordination stages through multi-dimensional data analysis, while the latter optimizes the execution logic based on a dynamically updated knowledge graph, forming a closed-loop control mechanism.

[0054] Please see Figures 1-2 As shown, a knowledge graph-based control system for a coffee robot robotic arm includes a coffee machine operation system, which is equipped with a mechanical control platform and a knowledge graph building terminal.

[0055] The mechanical control platform has communication connections with an execution end analysis unit, an executed end analysis unit, and a fusion analysis unit; the mechanical control platform is used to perform hardware operation control analysis on the robotic arm.

[0056] The mechanical control platform generates an execution-end analysis signal and sends it to the execution-end analysis unit.

[0057] After receiving the data, the execution analysis unit performs execution analysis on the coffee making process, where the execution end is represented by the robotic arm of the coffee robot.

[0058] The execution end is structurally divided according to the facility structure, that is, the wrist joint and elbow joint of the execution end are identified and uniformly marked as execution joints. The action coordination points of the execution end are obtained based on the execution joints, and the action trajectory of each joint of the execution end is determined by combining the action coordination points.

[0059] If the operating cycle of the execution end continues to increase, then the motion trajectory of the execution joints and the synchronization of motion coordination points are analyzed based on the motion trajectory:

[0060] That is, to obtain the displacement deviation of the joint movement trajectory within the action space during continuous execution of the action, as well as the offset distance of the mate point in the position during the continuous execution phase of the action;

[0061] The displacement deviation of the joint movement trajectory within the action space during continuous action execution and the offset distance of the mating point during the continuous action execution phase are compared with the displacement deviation threshold and the offset distance threshold, respectively. The displacement deviation threshold and the offset distance threshold are parameters that are manually set according to the robot's equipment model and facility structure operating characteristics.

[0062] If the displacement deviation of the joint movement trajectory in the action space exceeds the displacement deviation threshold during continuous execution of the action, or if the offset distance of the mate point in the continuous execution phase of the action exceeds the offset distance threshold, it is inferred that the execution end is currently in an unstable continuous operation phase.

[0063] If the displacement deviation of the joint movement trajectory within the action space during continuous execution of the action does not exceed the displacement deviation threshold, and the offset distance of the mating point within the continuous execution phase of the action does not exceed the offset distance threshold, then it is inferred that the execution end is currently in a stable and continuous operation phase.

[0064] The frequency of the execution angle drift during the unstable continuous operation phase is obtained, and the force feedback sensitivity during the stable continuous operation phase is collected. The force feedback sensitivity is the sensitivity of adjusting the motion amplitude according to the real-time grasping weight, specifically the parameter for setting the amplitude based on the weight.

[0065] If the frequency of angle drift increases during the unstable continuous operation phase, or the force feedback sensitivity fluctuates during the stable continuous operation phase, it is inferred that a risk has arisen in the current operation. The current time period is set as the risk operation period, and a risk operation signal is generated and sent to the mechanical control platform. After receiving the signal, the mechanical control platform adjusts the connection of the facility structure at the execution end and adjusts the execution action.

[0066] If the frequency of the angle drift does not increase during the unstable continuous operation phase and the force feedback sensitivity does not fluctuate during the stable continuous operation phase, it is inferred that no operational risk has occurred. The current period is then set as a safe operating period, and a safe operating signal is generated and sent to the mechanical control platform.

[0067] In response to the fact that traditional robotic arm control methods typically only monitor the displacement data of a single joint, this invention uses a dual monitoring mechanism that integrates the motion trajectory of multiple joints and the positional offset of the end effector to more comprehensively evaluate the operating status of the robotic arm. Furthermore, in view of the fact that the fixed threshold judgment method used in the existing technology cannot adapt to different action scenarios, this invention achieves accurate status recognition of complex operation scenarios by dynamically matching displacement parameters within the action space.

[0068] The above technical solution can effectively identify the stability changes of the robotic arm during continuous movements, and promptly distinguish between stable operation and abnormal shaking. When making cappuccino, which requires continuous milk foam injection and latte art, the system can detect the offset of the wrist joint in the arc trajectory in real time, avoiding pattern distortion caused by mechanical vibration. This solution provides reliable data support for subsequent risk period judgment by quantifying the correlation between joint movement deviation and end point offset, thereby reducing the waste of raw materials caused by the loss of control of the robotic arm during coffee making.

[0069] In addition, by superimposing angle drift frequency trend analysis and force feedback sensitivity monitoring, early warnings are issued in the early stages of robotic arm performance degradation, avoiding control lag problems caused by misjudgment of a single parameter. Potential risk periods of the robotic arm control system are dynamically identified, and early warnings are triggered in time when the angle drift frequency increases or the force feedback sensitivity is abnormal. This reduces the probability of raw material spillage or operational errors caused by the robotic arm going out of control during coffee making. At the same time, by distinguishing safe operating periods, unnecessary control logic switching can be avoided, maintaining the efficient operation of the system in a stable state.

[0070] The mechanical control platform generates an analysis signal for the executed end and sends it to the analysis unit for the executed end.

[0071] After receiving the data, the execution end analysis unit analyzes the execution end, which is represented as a coffee container used by the coffee robot for processing, such as a coffee cup.

[0072] The distance offset between the real-time position of the executed end and the execution position of the executing end is obtained, and the deviation frequency of the real-time position positioning detection of the executed end is collected.

[0073] If the distance offset between the real-time position of the executed end and the execution position of the executing end exceeds the set distance offset threshold, or if the deviation frequency of the real-time position detection of the executed end exceeds the detection deviation frequency threshold, it is inferred that the executed end is in a stage of inefficient cooperation. A cooperation control signal is generated and sent to the mechanical control platform. After receiving the signal, the mechanical control platform adjusts the positioning detection accuracy of the executed end and accurately controls its position.

[0074] If the distance offset between the real-time position of the executed end and the execution position of the executing end does not exceed the set distance offset threshold, and the deviation frequency of the real-time position detection of the executed end does not exceed the detection deviation frequency threshold, it is inferred that the executed end is in a highly efficient cooperation stage, generates a stable cooperation signal and sends it to the mechanical control platform; it should be explained that the distance offset threshold and the detection deviation frequency threshold are parameters artificially set by those skilled in the art in combination with the uncontrollable influences on the operation of the coffee robot, such as the position offset caused by equipment vibration, etc.

[0075] Traditional robotic arm control relies heavily on preset paths to execute actions, lacking the ability to respond in real time to changes in the position of target objects in dynamic environments. For example, when a coffee cup is accidentally moved, a conventional system may continue to operate along the original trajectory, leading to a collision. This invention uses a dual-parameter monitoring mechanism to proactively identify abnormalities in the collaborative state between the execution end and the target object, providing a quantitative basis for dynamically adjusting the control strategy.

[0076] Through the above technical solution, the present invention can effectively identify abnormal states of cooperation between the robotic arm and the target object, trigger timely adjustments to the control strategy, avoid problems such as coffee spillage and container collisions caused by positional deviation or misalignment, maintain a stable operating rhythm during the high-efficiency phase, and improve the coordination and reliability of the coffee preparation process.

[0077] After obtaining the risky runtime phase and the inefficient coordination phase, they are sent together to the fusion analysis unit;

[0078] After receiving the data, the fusion analysis unit performs fusion analysis on the risky runtime phase and the inefficient coordination phase.

[0079] The overlapping periods of the risky running phase and the inefficient coordination phase are extracted and marked as the fault tolerance analysis period; the percentage of real-time production volume loss of the executed end when the execution end deviates during the fault tolerance analysis period is obtained, specifically the amount of coffee spilled; the deviation of the clamping angle of the execution end when the position of the executed end deviates during the fault tolerance analysis period is obtained. It should be noted that the coffee spillage caused by the clamping angle deviation is also included in the percentage of production volume loss.

[0080] If the percentage of real-time production loss at the executed end exceeds the allowable loss threshold when the executed end shifts during the fault tolerance analysis period, or if the deviation of the clamping angle at the executed end exceeds the angle deviation threshold when the executed end shifts position during the fault tolerance analysis period, it is inferred that the fault tolerance performance analysis is abnormal when the executed end and the executed end are operating together. A low fault tolerance signal is generated and sent to the mechanical control platform. After receiving the low fault tolerance signal, the mechanical control platform adjusts the executed end and the executed end and controls the fault tolerance rate of the coordinated operation. That is, when the executed end and the executed end are set to operate, the set parameters are reserved, such as the clamping angle reservation setting.

[0081] If, during the fault tolerance analysis period, the percentage of real-time production loss at the executed end does not exceed the allowable loss threshold when the executed end shifts position, and the deviation of the clamping angle at the executed end does not exceed the angle deviation threshold when the executed end shifts position, then it can be inferred that the fault tolerance performance analysis is normal when the executed end and the executed end work together. A high fault tolerance signal is generated and sent to the mechanical control platform. After receiving the signal, the mechanical control platform can immediately put the executed end and the executed end into operation after adjustments are made. It should be noted that the allowable loss threshold and the angle deviation threshold are threshold parameters that are manually set to cope with the unavoidable impact of mechanical operation during coffee making.

[0082] This invention, by integrating time-period analysis and multi-dimensional parameter monitoring, can more accurately identify system-level fault-tolerant bottlenecks, such as simultaneously capturing coffee spillage caused by milk pitcher positioning deviation and robotic arm vibration, whereas traditional methods may misjudge such complex faults as being caused by a single factor.

[0083] The above technical solution effectively solves the error tolerance problem caused by the disconnect between the execution accuracy and the state of the controlled object in the robotic arm control system. In the coffee making scenario, it can accurately identify the risk of coffee spillage caused by the combined effects of robotic arm vibration and material container offset, and adjust the operating parameters in time to avoid material waste. At the same time, through the dual monitoring mechanism of gripping angle, it prevents tipping accidents caused by cup positioning deviation and improves the system robustness under complex working conditions.

[0084] The knowledge graph construction terminal communication connection includes an update feasibility analysis unit and an abnormal operation identification unit;

[0085] After the mechanical control platform completes monitoring and corresponding mechanical control, the knowledge graph building end, which serves as the operation and execution logic of the coffee robot, needs to be continuously updated and anomaly monitoring performed.

[0086] The knowledge graph construction end generates an update feasibility analysis signal and sends it to the update feasibility analysis unit.

[0087] After receiving the updated feasibility analysis unit, an updated feasibility analysis is performed on the operational execution logic of the knowledge graph construction.

[0088] The knowledge graph is divided into execution logic and preparation logic. When trajectory deviation occurs during the execution of successive actions in the execution logic, the success rate deviation in the preparation logic after the successive action order is swapped is obtained, and the corresponding successive action order swap is used as the knowledge graph update. The impact data of the execution logic corresponding to the successive actions is collected, and the impact data is set as the trigger condition for knowledge graph update execution. When the knowledge graph executes according to the original set logic, trigger adjustment detection is performed. After the trigger is generated, the knowledge graph is replaced and a high feasibility signal is generated and sent to the knowledge graph building end. The knowledge graph building end outputs the data to the mechanical control platform according to the real-time determined logic within the knowledge graph. The impact data specifically includes impact parameters such as vibration frequency during joint movement.

[0089] Traditional coffee robot control systems typically use fixed programs to control the robotic arm's movements, making it impossible to dynamically adjust the execution logic based on real-time working conditions. Conventional solutions can only issue error alarms or stop the machine to wait for manual intervention when encountering trajectory deviations, lacking autonomous optimization capabilities. In contrast, this invention establishes a correlation model between the sequence of actions and the success rate of preparation, enabling adaptive updates of the execution logic and effectively solving the problem of process interruptions caused by abnormal robotic arm movements.

[0090] Through the above technical solutions, this invention can quickly generate alternative solutions when the robotic arm's movements deviate, avoiding the failure of the entire production process due to a single link failure. By quantitatively analyzing the impact of logical adjustments on the final preparation result, it ensures that each update effectively improves operational reliability. At the same time, the dynamic setting mechanism of trigger conditions reduces unnecessary logical replacement operations, maintains system operational stability, and ultimately achieves a dual improvement in coffee making efficiency and success rate.

[0091] After the knowledge graph update is completed, an abnormal operation identification signal is generated and sent to the abnormal operation identification unit.

[0092] After receiving the data, the abnormal operation identification unit identifies abnormal operations during the execution of the knowledge graph.

[0093] By combining execution logic and preparation logic for synchronous analysis, and comparing the sensor feedback and expected state at each key node in the coffee making process in real time, the system obtains the identifiers of abnormal execution links, expected state parameters, actual state parameters, and abnormal type labels for each abnormal execution link in the knowledge graph logic execution process. It also constructs abnormal features for each execution link, specifically operations such as insufficient powder, milk container not connected, and insufficient pressure. It should be noted that the features constructed based on sensor data comparison are all features with a probability of occurrence. The abnormal execution link identifier represents the abnormal production stage, such as the extraction stage or the milk frothing stage.

[0094] By acquiring environmental parameters in real time through temperature and humidity sensors deployed in the production area, the environmental characteristics of the execution environment are constructed, including ambient temperature, ambient humidity, and environmental status standards, specifically high temperature and high humidity conditions.

[0095] Real-time parameters of key raw materials are obtained from the raw material management system to construct the raw material status characteristics of the execution process, including coffee bean moisture content, milk temperature, and raw material freshness label, which can be stored as the number of days as the freshness label.

[0096] Based on the historical logic execution process of the knowledge graph, abnormal operation features are extracted. The abnormal operation features are compared with the current execution stage. If there are overlapping abnormal features, the causes of the current overlapping abnormal features are extracted based on the historical logic execution process, and a set of causes is constructed. The causes are extracted in the direction of environment and raw materials.

[0097] Based on the collected data on current environmental characteristics and raw material conditions, and the corresponding analysis of the inducing factors data set:

[0098] The system acquires the floating trends of arbitrary types of numerical data and the floating trends of the inducing data set, and obtains the trend overlap rate based on the comparison of floating trends. That is, the floating trend comparison uses the floating speed and floating span of the current floating trend as comparison parameters, and the highest value as the comparison standard; the trend overlap rate is set by the decreasing rate of the interval value corresponding to the floating speed or floating span.

[0099] If the trend overlap rate continues to increase, the anomaly type label is extracted from the overlapping anomaly features, and the anomaly execution link identifier, expected state parameter, and actual state parameter corresponding to the anomaly type label are obtained from the operation log. The anomaly type label is set as a high-weight anomaly type, and the parameter deviation frequency is statistically calculated by comparing the expected state parameter and the actual state parameter. When the parameter deviation frequency exceeds the set frequency threshold, the high-weight anomaly type and the corresponding anomaly execution link identifier are sent to the knowledge graph building end. If the knowledge graph building end has no updated execution logic, a mechanical control signal is generated and sent to the mechanical control platform. The mechanical control platform performs corresponding link execution control according to the high-weight anomaly type and the corresponding anomaly execution link identifier to reduce the probability of anomaly occurrence.

[0100] Compared with existing technologies, traditional coffee robot anomaly detection usually relies on a single sensor data threshold for judgment, which cannot distinguish the relationship between operational anomalies and environmental or raw material factors. However, this invention can effectively distinguish between robotic arm control anomalies and external interference factors by integrating multi-dimensional feature analysis of execution logic, environmental parameters and raw material status, combined with a comparison mechanism of historical abnormal operation pattern library. For example, when the robotic arm's gripping action deviates, existing technologies may directly determine it as a mechanical failure, while this invention can further analyze whether it is caused by material expansion due to changes in ambient temperature or frictional changes caused by abnormal moisture content of coffee beans.

[0101] Through the above technical solution, the present invention can improve the accuracy of identifying abnormal operations of coffee robots and quickly locate the environmental or raw material causes of the abnormality. For example, when an abnormal temperature is detected in the milk frothing process, the system can simultaneously analyze the environmental parameters of the refrigeration equipment and the freshness data of the milk raw materials to accurately determine whether the root cause of the abnormality is a refrigeration system failure or an expired raw material problem.

[0102] This processing mechanism enables the control system to adopt differentiated response strategies for different types of triggers, such as adjusting environmental temperature control equipment or triggering raw material replacement prompts, thereby reducing the false judgment rate and improving the efficiency of anomaly handling.

[0103] This invention also proposes a knowledge graph-based control method for a coffee robot's robotic arm. The specific control method steps are as follows:

[0104] Execution-side analysis: Perform execution-side analysis on the coffee making process to identify risky and safe operating segments based on the execution-side analysis.

[0105] The execution end is analyzed to identify the efficient and inefficient cooperation stages.

[0106] The fusion analysis sheet performs fusion analysis on risky runtime segments and inefficient coordination stages, and infers the fault tolerance performance of the execution end and the executed end based on the fusion analysis;

[0107] The feasibility analysis was updated, and the operational execution logic of the knowledge graph construction was updated. The knowledge graph was divided into execution logic and preparation logic, and the logic was updated according to the real-time execution actions.

[0108] Anomaly detection involves identifying abnormal operations during the execution of the knowledge graph; combining the execution logic and preparation logic for synchronous analysis; and using the analysis results to make logical execution decisions for the knowledge graph.

[0109] Thresholds, preset values, preset ranges, etc. are set for result comparison and analysis to determine whether they are good or bad. The value of these thresholds is determined by a combination of large-scale model analysis of sample data and human experience. They can also be adjusted appropriately based on seasonal or common-sense influences.

[0110] Furthermore, the settings for weighting ratios, influence factors, etc., are based on the magnitude of each parameter's influence on the results. The specific values ​​are allocated to ultimately reflect the impact on the results. The settings for input and storage are also determined by a combination of large-scale model analysis of sample data and human experience. Appropriate adjustments can also be made based on seasonal or rational influence conditions.

[0111] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A knowledge graph-based control system for a coffee robot's robotic arm, characterized in that, It is equipped with a coffee machine operation system, and the coffee machine operation system is equipped with a mechanical control platform and a knowledge graph construction terminal; The mechanical control platform has the following communication connections: The execution-side analysis unit performs execution-side analysis on the coffee making process, identifying risky and safe runtime segments based on this analysis. The process of the execution-side analysis unit is as follows: The structure is divided to identify the wrist and elbow joints at the execution end, which are uniformly marked as execution joints. Based on the execution joints, the action coordination points at the execution end are determined, and the action trajectory of each joint at the execution end is determined. The displacement deviation of the joint movement trajectory within the action space during continuous action execution, as well as the offset distance of the mate point in the position during the continuous action execution phase, are obtained and compared with the corresponding thresholds. If the displacement deviation of the joint movement trajectory in the action space exceeds the displacement deviation threshold during continuous execution of the action, or if the offset distance of the mate point in the continuous execution phase of the action exceeds the offset distance threshold, it is inferred that the execution end is currently in an unstable continuous operation phase. If the displacement deviation of the joint movement trajectory within the action space during continuous execution of the action does not exceed the displacement deviation threshold, and the offset distance of the mating point within the continuous execution phase of the action does not exceed the offset distance threshold, then it is inferred that the execution end is currently in a stable and continuous operation phase. The frequency of the angle drift during the unstable continuous operation phase was obtained, and the sensitivity of the internal force feedback during the stable continuous operation phase was collected. If the frequency of the execution angle drift increases during the unstable continuous operation phase, or if the force feedback sensitivity fluctuates during the stable continuous operation phase, then the current time period is set as the risk operation phase, and a risk operation signal is generated. If the frequency of the execution angle drift does not increase during the unstable continuous operation phase, and the force feedback sensitivity does not fluctuate during the stable continuous operation phase, then the current period is set as the safe operation period, and a safe operation signal is generated. The execution-side analysis unit analyzes the execution-side to determine the efficient and inefficient cooperation stages. The process of the execution-side analysis unit is as follows: The distance offset between the real-time position of the executed end and the execution position of the executing end is obtained, and the deviation frequency of the real-time position positioning detection of the executed end is collected. If the distance offset between the real-time position of the executed end and the execution position of the executing end exceeds the set distance offset threshold, or if the deviation frequency of the real-time position positioning detection of the executed end exceeds the detection deviation frequency threshold, it is inferred that the executed end is in a stage of inefficient cooperation, and a cooperation control signal is generated and sent to the mechanical control platform. If the distance offset between the real-time position of the executed end and the execution position of the executing end does not exceed the set distance offset threshold, and the deviation frequency of the real-time position positioning detection of the executed end does not exceed the detection deviation frequency threshold, it is inferred that the executed end is in the efficient cooperation stage, generates a stable cooperation signal and sends it to the mechanical control platform; The fusion analysis unit performs fusion analysis on risky runtime segments and inefficient coordination phases, and infers the fault tolerance performance of the execution and executed ends based on the fusion analysis; the process of the fusion analysis unit is as follows: The overlapping periods of the risky runtime phase and the inefficient coordination phase are extracted and marked as the fault tolerance analysis period; the percentage of real-time production volume lost by the executed end when the execution end performs an offset during the fault tolerance analysis period is obtained, specifically the amount of coffee spilled. Obtain the deviation of the clamping angle of the execution end when the position of the execution end shifts during the fault tolerance analysis period; If the percentage of real-time production loss at the executed end exceeds the allowable loss threshold when the executed end is offset during the fault tolerance analysis period, or if the deviation of the clamping angle at the executed end exceeds the angle deviation threshold when the executed end is offset during the fault tolerance analysis period, a low fault tolerance signal will be generated and sent to the mechanical control platform. If the percentage of real-time production loss at the executed end does not exceed the allowable loss threshold when the executed end is offset during the fault tolerance analysis period, and the deviation of the clamping angle of the executed end does not exceed the angle deviation threshold when the executed end is offset during the fault tolerance analysis period, then a high fault tolerance signal is generated and sent to the mechanical control platform. The communication connections for building the knowledge graph include: The feasibility analysis unit is updated to perform a feasibility analysis on the operational logic of knowledge graph construction. The knowledge graph is divided into execution logic and preparation logic, and the logic is updated based on real-time execution actions. The process of updating the feasibility analysis unit is as follows: When trajectory deviation occurs during the execution of successive actions in the execution logic, the success rate deviation in the preparation logic after the successive action order is swapped is obtained, and the corresponding successive action order swap is used as knowledge graph update. The impact data of the execution logic corresponding to the successive actions is collected, and the impact data is set as the trigger condition for knowledge graph update execution. When the knowledge graph is executed according to the original set logic, the trigger adjustment detection is performed. After the trigger is generated, the knowledge graph is replaced and a high feasibility signal is generated and sent to the knowledge graph building end. The knowledge graph building end transmits the logic within the knowledge graph determined in real time to the mechanical control platform. The anomaly detection unit identifies anomalies during the execution of the knowledge graph; it performs simultaneous analysis combining the execution logic and the preparation logic, and makes knowledge graph logic execution decisions based on the analysis results; the process of the anomaly detection unit is as follows: By combining execution logic and preparation logic for synchronous analysis, and comparing the sensor feedback and expected state of each key node in the coffee making process in real time, the system obtains the identifiers of each abnormal execution link, expected state parameters, actual state parameters, and abnormal type labels in the knowledge graph logic execution process; and constructs the abnormal features of the execution link. Environmental parameters are acquired in real time by temperature and humidity sensors deployed in the production area to construct the environmental characteristics of the execution environment; This includes ambient temperature, ambient humidity, and environmental condition standards; real-time parameters of key raw materials are obtained from the raw material management system to construct the raw material condition characteristics of the execution process, including coffee bean moisture content, milk temperature, and raw material freshness labels; Abnormal operation features are extracted based on the historical logic execution process of the knowledge graph. Abnormal operation features are compared with the current execution stage. If there are overlapping abnormal features, the causes of the current overlapping abnormal features are extracted based on the historical logic execution process, and a set of cause data is constructed. The cause extraction direction is environmental and raw material.

2. The knowledge graph-based coffee robot robotic arm control system according to claim 1, characterized in that, Based on the collected data and the data set of causes, corresponding analysis is performed on the current environmental characteristics and raw material status characteristics: the fluctuation trend of any type of collected data and the fluctuation trend of the data set of causes are obtained, and the trend overlap rate is obtained by comparing the fluctuation trends. That is, the fluctuation trend comparison uses the current fluctuation speed and fluctuation span as comparison parameters, and the highest value as the comparison standard; the trend overlap rate is set by the decreasing rate of the interval value corresponding to the fluctuation speed or fluctuation span.

3. The knowledge graph-based coffee robot robotic arm control system according to claim 2, characterized in that, If the trend overlap rate continues to increase, the anomaly type label is extracted from the overlapping anomaly features, and the anomaly execution link identifier, expected state parameter, and actual state parameter corresponding to the anomaly type label are obtained from the operation log. The anomaly type label is set as a high-weight anomaly type, and the parameter deviation frequency is statistically calculated by comparing the expected state parameter and the actual state parameter. When the parameter deviation frequency exceeds the set frequency threshold, the high-weight anomaly type and the corresponding anomaly execution link identifier are sent to the knowledge graph building end. If the knowledge graph building end has no updated execution logic, a mechanical control signal is generated and sent to the mechanical control platform. The mechanical control platform performs corresponding link execution control according to the high-weight anomaly type and the corresponding anomaly execution link identifier to reduce the probability of anomaly occurrence.

4. A knowledge graph-based control method for a coffee robot's robotic arm, characterized in that, The specific control method steps for the knowledge graph-based coffee robot robotic arm control system described in any one of claims 1-3 are as follows: Execution-side analysis: Perform execution-side analysis on the coffee making process to identify risky and safe operating segments based on the execution-side analysis. The execution end is analyzed to identify the efficient and inefficient cooperation stages. The fusion analysis sheet performs fusion analysis on risky runtime segments and inefficient coordination stages, and infers the fault tolerance performance of the execution end and the executed end based on the fusion analysis; The feasibility analysis was updated, and the operational execution logic of the knowledge graph construction was updated. The knowledge graph was divided into execution logic and preparation logic, and the logic was updated according to the real-time execution actions. Anomaly detection involves identifying abnormal operations during the execution of the knowledge graph; combining the execution logic and preparation logic for synchronous analysis; and using the analysis results to make logical execution decisions for the knowledge graph.

Images