Beverage fusion method and system for realizing taste and flavor customization

Abstract

The invention relates to the technical field of automatic coffee machines, and provides a drink fusion method for realizing taste and flavor customization, which comprises the following steps: S1, obtaining fusion control parameters according to user requirements; s2, generating an execution track based on the fusion control parameters; s3, controlling the injection flow rate; s4, executing fusion operation and monitoring a fusion state; and S5, recording and optimizing fusion data. The refined physical control of the automatic coffee machine on the beverage fusion process is realized, the customized taste and flavor of the user are stably reproduced, and the technical limitations of the existing equipment in the aspects of track adaptability, flow velocity accuracy, taste parameter mapping, closed-loop feedback and the like are effectively broken through. High flexibility and expandability are achieved, various series mechanical arms from four shafts to seven shafts are supported, and different cost and performance requirements are met; parallel mechanisms such as De lta are compatible, and high-speed fusion application is achieved; cooperative configuration of a mechanical arm and an auxiliary platform is supported, and motion distribution is optimized; and a collaborative robot can be accessed to realize man-machine collaborative operation.

Description

Technical Field

[0001] The present invention relates to the technical field of automatic coffee machines, and in particular to a beverage blending method and system for achieving customized taste and flavor. Background Art

[0002] As coffee consumption is upgrading towards personalization and refinement, automated coffee machines have gradually evolved from simple "standardized production tools" to "customized taste solutions." Users' demand for coffee is no longer limited to basic strength adjustments, but extends to more refined taste levels—for example, "clear layered transitions between milk foam and coffee," "a dense and smooth fusion texture is felt when the drink enters the mouth," and "a harmonious balance of bitter and sweet flavors in the mouth." These demands pose new challenges to the core capabilities of automated coffee machines: not only must mechanical actions be precisely executed, but the taste and flavor of the beverage must be directly shaped through physical control of the fusion process. However, current automated coffee machines still have many limitations in the key technical aspects of fusion control, making it difficult to meet this core demand.

[0003] The primary bottleneck is the lack of adaptability in trajectory control. Existing automated coffee machines often use pre-set, fixed patterns, such as straight-line injection, circular loops, or simple zigzag motions, and cannot dynamically adjust to actual scenarios. Different cup sizes have significantly different requirements for the injection path. Narrow-mouthed cups have limited inner wall space, requiring a compact injection trajectory to prevent liquid overflow. A wide trajectory can easily cause milk foam to splash outside the cup. Wide-mouthed cups require a wider trajectory coverage to ensure even distribution of milk foam and coffee within the cup. A narrow trajectory can result in excessive mixing in some areas and insufficient fusion in others. Furthermore, the characteristics of milk foam place specific demands on trajectory height. Fresh milk foam has a delicate bubble structure. If the injection height is too low, the impact of the liquid surface can cause the bubbles to break, resulting in a loss of smoothness. Pre-frothed milk foam has more stable bubbles, so a lower injection height can be used to improve fusion efficiency. However, existing fixed trajectories cannot account for these variables, resulting in significant fluctuations in fusion performance across the same machine when changing cup sizes or milk foam types, directly impacting taste consistency.

[0004] Another core shortcoming is the lack of precision in flow rate control. Existing equipment often calculates milk foam flow rate based on traditional fluid mechanics principles, considering only basic factors like the pitcher's tilt angle and liquid gravity. However, it ignores the complex properties of milk foam as a non-Newtonian fluid. The fluidity of milk foam changes dynamically with external conditions. During the pouring process, bubbles in fresh milk foam gradually coalesce and burst, causing viscosity to decrease over time and naturally increasing flow rate. Pre-whipped milk foam, after a long period of rest, develops a more stable internal structure and a slower viscosity decay. However, under high shear forces (such as during rapid pouring), its fluidity can suddenly increase. These characteristics lead to significant errors in traditional flow rate calculation methods, and the actual pouring speed often deviates significantly from the expected one. Excessively high flow rates can impact the oil layer on the coffee surface, disrupting the layered texture. Excessively low flow rates prolong the fusion process, leading to a temperature imbalance between the milk foam and the coffee. A rapid drop in the milk foam temperature can result in a loss of density, while excessive dilution of the coffee temperature can weaken the flavor.

[0005] The disconnect between taste and parameters further restricts customization capabilities. Existing automated coffee machines are unable to convert user demands for abstract tastes such as "distinct layers" and "dense and smooth" into quantifiable and controllable physical parameters. For example, users who pursue a "clear sense of layering" need to control the degree of mixing of milk foam and coffee at a lower level, which corresponds to a specific number of injection trajectory circles (fewer cycles), flow rate change rhythm (slow and even injection) and milk foam injection timing (inject immediately after coffee extraction is completed); while "fully integrated smoothness" requires more trajectory cycles, slightly faster flow rates and longer integration time. However, existing equipment lacks such mapping logic and can only rely on preset programs for rough adjustments. It is impossible to accurately match parameters according to user needs, resulting in "personalized customization" remaining on the surface and making it difficult to truly achieve differentiated shaping of taste.

[0006] The lack of a closed-loop feedback mechanism exacerbates the lack of control accuracy. During the injection process, existing equipment often relies solely on preset time or volume parameters to determine the fusion endpoint, and lacks monitoring and adjustment of the real-time fusion status. In actual operation, the characteristics of the milk foam may vary slightly due to factors such as the raw material batch and ambient temperature. For example, if the fat content of a batch of milk is slightly higher, the viscosity of the milk foam after whipping will be higher than the normal value. When injected according to standard parameters, the actual fusion speed will be slower, and the final drink will taste heavier. When the ambient temperature is low, the fluidity of the milk foam will decrease. If the flow rate is not adjusted, insufficient injection volume may result. Due to the lack of real-time monitoring (such as judging the mixing uniformity through image recognition and sensing the injection speed through weight sensors), the equipment cannot correct these deviations in a timely manner and can only passively accept the differences between the finished product and the expected results, making stability difficult to guarantee.

[0007] In addition, the multivariable coupling problem in the fusion process also makes it difficult for existing control logic to cope with it. Factors such as the range of the injection trajectory, the speed of the flow rate, the characteristics of the milk foam, and the length of the fusion time affect and restrict each other: expanding the trajectory range helps to improve the uniformity of mixing, but may destroy the layered taste; increasing the flow rate can shorten the fusion time, but may increase the impact of the liquid surface; extending the fusion time can make the flavor more coordinated, but may cause the milk foam to defoam. These contradictions make it difficult to achieve the goal of a complex taste by adjusting a single parameter. A control system that can coordinate and control multiple variables is needed, which is exactly the core capability that current automated coffee machines lack.

[0008] In summary, the limitations of current automated coffee machines in trajectory adaptability, flow rate accuracy, taste parameter mapping, closed-loop feedback, and multivariable collaborative control make it difficult to achieve refined physical control of the fusion process, hindering the industrialization of personalized taste customization. To address these issues, a new fusion control technology system is urgently needed. Through dynamic trajectory generation, precise flow rate control, taste parameter mapping, and real-time feedback correction, this system can overcome the customization bottleneck of automated coffee machines and truly achieve the core goal of "one cup, one parameter, customizable taste." Summary of the Invention

[0009] To address the above issues, the present invention aims to provide a method and system for beverage blending that achieves customized taste and flavor. By enabling refined physical control of the beverage blending process by an automated coffee machine, the method can stably reproduce user-customized taste and flavor, effectively overcoming the technical limitations of existing equipment in terms of trajectory adaptability, flow rate accuracy, taste parameter mapping, and closed-loop feedback. The technical solution of the present invention is highly flexible and scalable:

[0010] Versatility of actuators:

[0011] Supports various serial robotic arms from 4 to 7 axes to meet different cost and performance requirements;

[0012] Compatible with parallel mechanisms such as Delta to achieve high-speed fusion applications;

[0013] Supports collaborative configuration of robotic arm + auxiliary platform to optimize motion distribution;

[0014] It can be connected to collaborative robots to achieve human-machine collaborative operation.

[0015] Diversity of sensing technologies:

[0016] Fusion detection is not limited to specific sensors and can be flexibly configured based on accuracy and cost;

[0017] Support multi-sensor data fusion to improve detection reliability;

[0018] Compatible with new sensing technologies such as spectral analysis, acoustic imaging, etc.

[0019] Intelligent trajectory generation:

[0020] Not limited to elliptical trajectories, supporting a variety of trajectory shapes and combinations;

[0021] The trajectory parameters can be automatically optimized according to the container shape;

[0022] Support user-defined tracks to meet creative needs;

[0023] It has the ability to adaptively adjust and optimize the fusion effect in real time.

[0024] Advanced control algorithm:

[0025] Mix multiple control models to balance theoretical accuracy and practical adaptability;

[0026] Support online learning and parameter optimization;

[0027] New AI algorithms can be expanded to continuously improve performance.

[0028] The above-mentioned object of the present invention is achieved through the following technical solutions:

[0029] A method for achieving customized taste and flavor of beverage fusion, comprising the following steps:

[0030] S1: Obtain fusion control parameters according to user requirements;

[0031] S2: generating an execution trajectory based on the fusion control parameters;

[0032] S3: control injection flow rate;

[0033] S4: perform fusion operation and monitor fusion status;

[0034] S5: Record and optimize fusion data.

[0035] Furthermore, in step S2, the execution mechanism for executing the execution trajectory is a motion device having at least four controllable degrees of freedom, including at least one of the following:

[0036] Four-axis to seven-axis serial robotic arms;

[0037] Parallel kinematic mechanisms with three to six degrees of freedom;

[0038] A combination of a robotic arm and one or more auxiliary motion platforms;

[0039] collaborative robotic systems;

[0040] SCARA robot with rotating platform.

[0041] Furthermore, the actuator adopts the following adaptive configuration scheme, specifically:

[0042] Automatic detection of connected actuator type and number of degrees of freedom;

[0043] Load the corresponding kinematic model according to the test results;

[0044] Adaptive adjustment of trajectory generation and control parameters;

[0045] Supports hot swapping and dynamic reconfiguration of actuators.

[0046] Furthermore, in step S2, the form of the three-dimensional trajectory as the execution trajectory can be dynamically selected or combined, including:

[0047] Basic geometric trajectories: ellipse, circle, spiral, polygon;

[0048] Complex parameter trajectories: Lissajous figures, fractal trajectories, Bezier curves;

[0049] Adaptive trajectory: A trajectory that is dynamically adjusted based on real-time fusion feedback;

[0050] User-defined trajectory: defined by parameterized equations or teaching.

[0051] Furthermore, in step S4, the fusion state is monitored using a multimodal fusion degree detection technology, including a combination of at least two of the following detection methods:

[0052] Visual detection: at least one of visible light camera, multispectral camera, thermal imaging camera, and structured light 3D camera;

[0053] Electrical detection: at least one of conductivity sensor, dielectric constant sensor, and impedance spectrum analyzer;

[0054] Acoustic detection: at least one of ultrasonic sensor and acoustic impedance sensor;

[0055] Optical detection: at least one of laser scattering instrument, turbidimeter, and fluorescence spectrometer;

[0056] Other detection: at least one of pH sensor, viscosity sensor, and density sensor.

[0057] Furthermore, in step S3, controlling the injection flow rate requires adopting a mixed fluid model, wherein the mixed fluid model includes at least one of the following or a combination thereof:

[0058] Fluid dynamics models based on physical laws;

[0059] Data-driven machine learning models, including but not limited to single-shot learning, deep learning, and reinforcement learning;

[0060] Fuzzy control model;

[0061] Expert system model;

[0062] Hybrid models combine the advantages of two or more of the above models.

[0063] Furthermore, before step S1, a parameter-taste mapping database is established. The input of the parameter-taste mapping database is the user's taste requirements and the real-time characteristics of the milk foam, and the output is the corresponding fusion control trajectory parameters including the ellipse semi-axis, fusion depth, fusion number of circles, and fusion time. Specifically,

[0064] Conducting experimental data collection, systematically varying fusion control trajectory parameters (including trajectory morphology parameters, fusion depth height_bias, number of fusion circles num_circles, and fusion duration duration_s), and milk foam characteristics (including carbonation, whipping degree, and temperature) to prepare multiple groups of differentiated beverages. The trajectory morphology parameters include the ellipse semi-axis, spiral radius, and Lissajous parameters.

[0065] Conduct sensory calibration, with professional baristas evaluating each drink group's bitter-sweet balance, sweetness intensity, and taste levels to generate expert-annotated data.

[0066] Quantify the degree of fusion of each group of drinks using a preset fusion calculation formula, establish a multivariate regression model of fusion control trajectory parameters + milk foam characteristics → fusion degree → sensory evaluation, and store it as the initial parameter-taste mapping database;

[0067] The calculation formula for the fusion degree is:

[0068] Fusion_Rate = (V_mixed / V_total) × K_foam × K_motion, where V_mixed is the volume actually mixed through vortex motion, V_total is the total volume of liquid in the cup, K_foam is the milk foam characteristic coefficient, which is determined by the gas content and the degree of whipping, and K_motion is the motion intensity coefficient, which is determined by the number of fusion circles, fusion depth and fusion speed.

[0069] Furthermore, the parameter-taste mapping database adopts a cloud-based parameter-taste mapping database system, and the database system specifically includes:

[0070] Support multi-device data sharing and synchronization;

[0071] Use federated learning to protect user privacy;

[0072] Achieve cross-regional taste preference analysis;

[0073] Support crowdsourced development of new taste patterns.

[0074] Furthermore, in step S1, obtaining fusion control parameters according to user requirements includes: matching customized requirement input with parameters, receiving user taste requirements and real-time characteristics of milk foam, calling parameter-taste mapping database, and outputting corresponding fusion control trajectory parameters including trajectory morphology parameters, fusion depth, number of fusion circles, and fusion duration, specifically:

[0075] S11: Analyze user taste requirements and quantify the natural language input of users into quantitative indicators including bitter-sweet balance, sweetness perception intensity, and taste levels;

[0076] S12: Detect and infer the milk foam's real-time characteristics. The viscosity sensor obtains the milk foam's viscosity value in real time, and the temperature sensor obtains the milk foam's temperature. The flow rate variation curve during the injection process is analyzed to infer the milk foam's gas content. The degree of whipping is determined based on the viscosity and gas content.

[0077] S13: Calling and matching the parameter mapping database, selecting the historical data subset closest to the current milk foam characteristics in the parameter-taste mapping database, and based on the fusion target corresponding to the user's taste requirements and the selected trajectory shape, in the selected historical data subset:

[0078] Prioritize trajectory shape matching: Select the optimal trajectory type based on container shape and fusion requirements;

[0079] Matching trajectory parameters:

[0080] For elliptical trajectory: high-carbonated milk foam matches the larger semi-axis, low-carbonated milk foam matches the smaller semi-axis;

[0081] For spiral trajectory: adjust the pitch and radius change rate according to the fusion depth requirements;

[0082] For adaptive trajectories: set initial parameters and adjust sensitivity;

[0083] Match fusion depth height_bias: heavy whipping matches shallower depth, light whipping matches deeper depth;

[0084] Matching the number of fusion circles num_circles: Based on the fusion target and the milk foam characteristic coefficient, a low value is used for high carbonation / heavy whipping, and a high value is used for low carbonation / light whipping;

[0085] Matching fusion duration duration_s: positively correlated with the number of fusion cycles and compensated by the milk foam temperature;

[0086] S14: Perform dynamic parameter correction, including correction of fusion time based on temperature, correction of fusion number based on viscosity, correction of safe operating area, and correction of cup shape based on visual system recognition.

[0087] Furthermore, in step S2, generating an execution trajectory based on the fused control parameters includes: performing three-dimensional trajectory generation, calling a trajectory generation function based on the fused control parameters to generate three-dimensional path points including an ellipse, a spiral, a Lissajous figure, or an adaptive trajectory that changes with time, as an input path for the actuator motion planning, and outputting a Cartesian trajectory X, Y, Z sequence, specifically:

[0088] S21: Call the trajectory generation function go_ellipses_3d and enter the following parameters:

[0089] Trajectory type parameter: specifies the shape of the generated trajectory, including ellipse / spiral / lissajous / adaptive trajectory;

[0090] Center pose parameters: the spatial coordinates center_x, center_y, center_z defined by the geometric center of the cup, corresponding to the decomposition value of the center pose target_pose;

[0091] Fusion control parameters: including the x-axis radius_x and the y-axis radius_y of the ellipse semi-axis, which determine the major and minor axis sizes of the ellipse in the horizontal plane;

[0092] Number of fusion circles num_circles: determines the number of times the trajectory is repeated;

[0093] Height bias height_bias: determines the trajectory offset in the Z-axis direction;

[0094] Total trajectory duration_s: the total time to complete all fusion cycles;

[0095] S22: Trajectory generation is driven by discrete time steps dt_s. In each time step, the coordinates of the three-dimensional path points are calculated according to the following rules:

[0096] x(t)=center_x+radius_x*cos(2*pi*t / (duration_s / num_circles));

[0097] y(t)=center_y+radius_y*sin(2*pi*t / (duration_s / num_circles));

[0098] z(t)=center_z+height_bias;

[0099] Where t is the current time step, which ranges from greater than or equal to 0 to less than or equal to duration_s;

[0100] 2*pi*t / (duration_s / num_circles): defines the periodic motion angle of the ellipse, and each circle takes duration_s / num_circles;

[0101] center_x, center_y: X and Y coordinates of the center point of the ellipse trajectory;

[0102] radius_x, radius_y: the length of the semi-axis of the ellipse in the X and Y directions;

[0103] center_z: the initial Z coordinate of the center point of the trajectory;

[0104] height_bias: offset in the Z-axis direction, controlling the track height;

[0105] S23: Dynamically construct trajectory morphology

[0106] Fixed height ellipse: If height_bias is a constant value, an elliptical trajectory is generated in the horizontal plane, and the Z coordinate remains unchanged;

[0107] Spiral descending ellipse: If height_bias decreases linearly with the number of fusion turns, a trajectory that spirally contracts / expands along the negative Z-axis is generated;

[0108] S24: Output the three-dimensional path point sequence of continuous time steps as the X, Y, Z coordinate sequence of the Cartesian space trajectory of the robot arm motion planning, which serves as the input of the subsequent inverse kinematics solution.

[0109] Furthermore, in step S3, controlling the injection flow rate includes: performing flow rate control planning, based on a mixed fluid model including a physical model and a single learning model, combining the current pouring angle and the remaining volume, predicting and controlling the milk foam flow rate in real time, and outputting a sequence of pitch angles of the pouring posture, specifically:

[0110] A hybrid fluid control framework was constructed that integrated a physical model with a single-shot learning model. The physical model, based on the Bernoulli equation and the Coanda effect, combined geometric parameters such as the width and curvature of the spout of the latte art pot to derive the theoretical relationship between milk foam flow rate and pouring angle. The single-shot learning model analyzed historical injection data to establish a mapping relationship between pouring angle, residual volume, and flow rate deviation, dynamically correcting the prediction error of the physical model.

[0111] Execute the real-time flow rate prediction and control process. The current pouring pitch angle is collected through the IMU sensor, and the remaining volume of milk foam is obtained by an electronic scale or liquid level sensor. The theoretical flow rate is first calculated using the physical model, and then the correction value output by the single learning model is called and integrated to obtain the real-time predicted flow rate. Based on the deviation between the predicted flow rate and the target flow rate, the pouring angle is dynamically adjusted through the PID control algorithm to gradually converge the actual flow rate to the target value.

[0112] The output pitch angle sequence of the pouring posture is used as the input command for the robot arm motion planning, and dual constraints are set: first, the pouring angle is limited to a preset range to prevent milk foam overflow or too low a flow rate; second, the flow rate change rate is controlled to ensure a smooth and stable injection process and prevent sudden changes in flow rate from disrupting the fusion of milk foam and coffee.

[0113] Furthermore, in step S4, performing a fusion operation and monitoring the fusion status includes: synthesizing the Cartesian trajectory X, Y, Z sequence and the Pitch angle sequence through a trajectory synthesizer to generate a final synchronous trajectory, converting the final synchronous trajectory into an actuator control instruction to drive the injection action, calculating the real-time fusion degree through image and conductivity detection, and correcting the trajectory parameters and flow rate based on the deviation between the fusion degree and the target value to ensure that the target fusion degree is achieved, specifically:

[0114] The trajectory synthesizer receives the Cartesian trajectory X, Y, and Z sequence and the pitch angle sequence, and achieves synchronous synthesis through time axis alignment: based on the total trajectory duration, the spatial path points with X, Y, and Z coordinates are bound to the corresponding pitch angle at the time, generating a final synchronized trajectory containing position and posture information, ensuring that the robot arm synchronizes with the preset tilt angle changes while performing spatial motion;

[0115] The final synchronized trajectory is input into the robotic arm motion controller, which converts it into angle commands for each joint through inverse kinematics. This drives the robotic arm to execute the injection motion according to the planned path. During the injection process, a high-speed camera captures real-time images of the interface between the milk foam and coffee. Combined with a conductivity probe, the uniformity of the mixture is tested, and the real-time degree of integration, including layered integration or uniform integration, is comprehensively calculated.

[0116] Based on the deviation between the real-time detection value of the fusion degree and the target value, the control parameters are dynamically corrected: if the layered fusion degree is too high, the number of trajectory circles is reduced and the rate of change of the pouring angle is lowered; if the uniform fusion degree is insufficient, the trajectory radius is increased and the flow rate is increased. The corrected parameters are updated to the trajectory synthesizer in real time to ensure that the final fusion degree converges to the target range and guarantees the stable reproduction of the preset taste.

[0117] Furthermore, in step S5, recording and optimizing the fusion data includes: performing taste verification and database updating, verifying the matching degree between the taste of the beverage and the user's needs, entering the fusion control trajectory parameters + milk foam characteristics → fusion degree → sensory evaluation new data into the parameter-taste mapping database, and optimizing the subsequent matching accuracy, specifically:

[0118] After the drink is prepared, the taste compatibility is evaluated through a multi-dimensional verification mechanism: image analysis technology is used to identify the layered structure or uniformity of the milk foam and coffee, and the actual degree of integration is compared with the target value. An electronic tongue sensor collects sweetness and bitterness flavor data, and combines it with preset sensory evaluation standards to generate a quantitative taste score. The system integrates visual characteristics and flavor data to form a final judgment on whether the drink tastes meet user requirements.

[0119] Recording the complete preparation process data into the parameter-taste mapping database: specifically, fusion control trajectory parameters including ellipse semi-axis, fusion depth, number of fusion turns, and fusion time, milk foam characteristics including carbonation, whipping degree, and temperature, real-time fusion degree data, and sensory evaluation results. Each record maintains the mapping relationship between parameters, forming a complete data chain of trajectory parameters + milk foam characteristics → fusion degree → sensory evaluation.

[0120] Iteratively optimize the mapping model based on newly added data: The system regularly analyzes the deviation between historical data and newly added data, adjusts the mapping relationship between parameters and taste scores through machine learning algorithms, pays attention to the taste deviation scenarios that appear frequently in user feedback, and specifically strengthens the matching accuracy of relevant parameter combinations, gradually improving the database's prediction accuracy for new milk foam types and personalized taste requirements.

[0121] Furthermore, the method for achieving beverage blending with customized taste and flavor also includes using a modular beverage blending control device to blend beverages, specifically including:

[0122] Core control module, compatible with different types of actuators;

[0123] Replaceable actuator interface, supporting 4-7 axis robotic arms and other motion devices;

[0124] Scalable sensor interface to support multimodal sensing devices;

[0125] Unified software architecture, adaptive to different hardware configurations.

[0126] Furthermore, the method for achieving customized taste and flavor beverage fusion also includes optimizing the taste based on user feedback, specifically including:

[0127] Collect real-time user feedback through mobile apps or touch screens;

[0128] Reinforcement learning algorithm is used to optimize fusion parameters;

[0129] Support personalized taste profiles;

[0130] Achieve taste reproduction across devices.

[0131] A beverage fusion system for achieving customized taste and flavor for executing the beverage fusion method for achieving customized taste and flavor as described above, comprising:

[0132] Fusion parameter acquisition module, used to obtain fusion control parameters according to user needs;

[0133] An execution trajectory generation module, configured to generate an execution trajectory based on the fusion control parameters;

[0134] An injection flow rate control module, used for controlling the injection flow rate;

[0135] A fusion execution monitoring module is used to execute fusion operations and monitor fusion status;

[0136] The fusion data optimization module is used to record and optimize the fusion data.

[0137] Compared with the prior art, the present invention has at least one of the following beneficial effects:

[0138] (1) Breaking through the bottleneck of trajectory adaptability: Through the dynamic generation technology of three-dimensional elliptical trajectories, parameters such as the elliptical semi-axis, fusion depth, and number of trajectory turns can be adjusted in real time according to the characteristics of the milk foam (carbon content, degree of whipping) and the difference in cup shape. For example, highly carbonated milk foam matches a larger semi-axis with a higher fusion depth, while heavily whipped milk foam uses a shallower depth and a lower number of turns. This effectively solves the problem of fluctuations in the fusion effect of a fixed trajectory after changing the cup shape or milk foam type, ensuring consistency of taste in different scenarios.

[0139] (2) Improving flow rate control accuracy: A hybrid fluid control framework based on a physical model and a single-shot learning model not only considers the theoretical relationship between the geometric parameters of the latte art cylinder and the Bernoulli equation, but also uses historical data to correct flow rate deviations caused by non-Newtonian fluid characteristics. Combined with PID closed-loop control, the flow rate prediction error is ≤±0.2ml / s, avoiding problems such as excessive flow rate impacting the grease layer and excessive flow rate causing temperature imbalance caused by ignoring the dynamic characteristics of milk foam in traditional equipment.

[0140] (3) Establishing a precise taste parameter mapping mechanism: Through the parameter-taste mapping database, users’ abstract requirements such as “clear layers” and “dense and smooth” are converted into quantifiable fusion targets (e.g., a layered fusion degree ≤ 30% corresponds to a layered taste), and further decomposed into physical parameters such as elliptical trajectory and flow rate. This solves the “demand-parameter” disconnect in existing equipment and enables direct conversion from user language to control instructions.

[0141] (4) Strengthening closed-loop feedback and dynamic correction capabilities: During the injection process, high-speed cameras and conductivity probes are used to detect the degree of fusion in real time. Based on the deviation, trajectory parameters are dynamically adjusted (such as reducing the number of turns to improve the sense of stratification and increasing the radius to enhance uniformity) to ensure that the degree of fusion converges to the target range (error ≤ ±5%). This overcomes the shortcomings of traditional equipment that relies on preset parameters and cannot cope with differences in raw material batches and environmental fluctuations, significantly improving the stability of the finished product.

[0142] (5) Database iteration and continuous optimization: By collecting full data on "control parameters + milk foam characteristics + fusion degree + sensory evaluation" and combining it with machine learning algorithms to continuously optimize the mapping model, the parameter matching accuracy under new data is improved by ≥10%. This not only adapts to conventional milk foam types, but also quickly responds to new ingredients such as plant-based milk foam, supporting long-term personalized taste upgrades.

[0143] (6) Stable reproduction of customized taste and flavor: Combining the above technological innovations, the system can accurately control the mixing degree, flow field morphology and temperature balance of milk foam and coffee, and achieve industrial reproduction of customized tastes such as "clear layers" and "bitter-sweet balance". Compared with existing equipment, it can not only ensure the consistency of taste of a single cup of beverage, but also meet the differentiated needs of different users, promoting the upgrade of automated coffee machines from "standardized production" to "personalized solutions".

[0144] (7) Wide range of hardware compatibility: Through the universal design of the actuator, the present invention can be adapted to more than 90% of industrial robots and collaborative robots on the market, without the need for dedicated hardware, significantly lowering the threshold for technology application. Enterprises can choose the appropriate hardware configuration based on their budget and needs, from economical 4-axis SCARA to high-end 7-axis collaborative robots, all of which can achieve taste customization.

[0145] (8) Continuity of technological evolution: The modular and open system architecture ensures the sustainable development of technology. With the emergence of new sensors, AI algorithms, and actuators, the system can quickly integrate new technologies through software upgrades or module replacements, avoiding premature equipment obsolescence. The system lifecycle is expected to be extended to more than 10 years.

[0146] (9) Openness of the ecosystem: Through standardized interfaces and cloud databases, the present invention supports the construction of an open technology ecosystem. Third-party developers can contribute new trajectory algorithms and taste patterns, and users can share and download taste configurations, forming an innovative model similar to the "App Store" and promoting technological progress across the industry.

[0147] (10) Convenience of global application: The cloud database supports cross-regional taste preference analysis and configuration sharing. Chain coffee brands can ensure taste consistency across stores worldwide while making local adjustments based on regional characteristics. The system supports multilingual interfaces and localized taste descriptions, facilitating global promotion. BRIEF DESCRIPTION OF THE DRAWINGS

[0148] Figure 1 The overall flow chart of the beverage fusion method for achieving customized taste and flavor of the present invention;

[0149] Figure 2 This is a diagram showing the principle of controlling taste during the fusion process of the present invention;

[0150] Figure 3 This is a structural diagram of the parameter-taste mapping database of the present invention;

[0151] Figure 4 Schematic diagram of the trend change of eddy current intensity and trajectory radius of the present invention;

[0152] Figure 5 This is a schematic diagram of the change of fusion depth over time in the present invention;

[0153] Figure 6 This is a schematic diagram of the boundary of the safe operating area of ​​the present invention;

[0154] Figure 7 A diagram of the three-dimensional elliptical trajectory generation and synchronous execution system of the present invention;

[0155] Figure 8 This is a schematic diagram of the three-dimensional trajectory generation of the go_elipses_3d function of the present invention;

[0156] Figure 9 Schematic diagram of the effect of flow rate on fusion quality in the present invention;

[0157] Figure 10 The overall structure diagram of the beverage fusion system for achieving taste and flavor customization of the present invention. DETAILED DESCRIPTION

[0158] To make the purpose, technical solutions, and advantages of the embodiments of this application more clear, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0159] Those skilled in the art will appreciate that, unless otherwise stated, the singular forms "a," "an," "said," and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present invention refers to the presence of the stated features, integers, steps, operations, elements, and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0160] First embodiment

[0161] In traditional beverage production, once the raw materials are determined, the taste and flavor are basically fixed. The core requirement of this invention is to break this limitation and introduce a precisely controlled physical fusion stage to achieve the goal of using exactly the same raw materials to create drinks with different tastes and flavors. Figure 1 As shown, this embodiment provides a beverage fusion method for achieving customized taste and flavor, comprising the following steps:

[0162] S1: Obtain fusion control parameters according to user requirements.

[0163] The taste of a latte depends largely on the degree of fusion between milk foam and espresso. This invention accurately quantifies the degree of fusion and establishes a scientific relationship between fusion and taste.

[0164] Quantitative definition of fusion and taste characteristics:

[0165] Low fusion (fusion <30%): The milk foam mostly sinks beneath the crema, forming a distinct layered structure. Because the crema layer is mostly bitter, the first thing you experience is the bitterness and richness of the coffee, followed by the smoothness of the milk foam. This low fusion preserves the unique flavor characteristics of both coffee and milk, creating distinct layers. Medium fusion (fusion 30%-70%): The milk foam partially blends with the coffee, forming a gradual transition layer. The taste is balanced and smooth, and the sweetness of the milk begins to neutralize the bitterness of the coffee, creating a harmonious blend of milk and coffee.

[0166] Extremely high integration (integration >70%): The milk foam and coffee are thoroughly blended, creating a seamless experience and fully integrated flavors. This high level of integration maximizes the sweetness of the milk, as thorough mixing allows the delicate texture and temperature characteristics of the milk foam to create an optimal sensory experience in the mouth, enhancing the overall sweetness. For 100% integration, the system allows you to skip the injection step and proceed directly to the integration process, achieving complete homogenization through a larger trajectory radius, deeper integration depth, and a higher number of integration cycles.

[0167] The impact of milk foam characteristics on taste:

[0168] The air content and degree of whipping of the milk foam are other key factors influencing the final taste. The more air that enters the milk, the more creamy the texture and the stronger the perceived sweetness. This is because: highly aerated milk foam has a finer microbubble structure, increasing the contact area with the taste buds; the addition of air changes the rheological properties of the milk foam, causing it to release more slowly in the mouth, prolonging the perception of sweetness; and moderately whipped milk foam has optimal stability and blends better with the coffee. Over-whipping causes the milk foam to separate, while under-whipping prevents the formation of a fine microbubble structure.

[0169] like Figure 2 As shown in the schematic diagram of the fusion process, this invention precisely controls the robotic arm's motion trajectory (e.g., number of revolutions, depth, and speed) while injecting milk foam, actively guiding the flow and mixing of liquids in the cup. This quantifies and controls the fusion process, providing customers with customized taste options. The system automatically adjusts fusion parameters based on the actual gas content and degree of whipping of the milk foam to ensure the target taste is achieved.

[0170] The liquid in the cup (coffee + milk foam) is a complex fluid system. The vortex intensity, shape and stability generated by the movement of the robotic arm are highly nonlinearly related to multiple factors such as the radius of the trajectory, speed, injection depth, injection flow rate, cup shape, liquid viscosity, etc. How to establish a quantitative relationship between these parameters and the final fusion degree (taste) is the biggest technical challenge. The core trick of the present invention is that it does not attempt to fully simulate complex fluid dynamics, but decouples and parameterizes this complex problem. We found that the final fusion degree is mainly determined by several key physical actions, and these actions can be effectively controlled by several core parameters of the go_elipses_3d function (radius, depth, number of turns). This allows us to bypass complex fluid simulations and control the final taste by directly controlling these parameters.

[0171] Through extensive physical experiments, we systematically varied parameters like radius, height_bias, and duration_s. Professional baristas then conducted sensory calibration on the finished product, creating a valuable database mapping trajectory parameters, integration level, and flavor. This database specifically records: the bitter-sweet balance at varying degrees of integration; the effect of foam carbonation on sweetness perception; and the correlation between optimal integration parameters, cup shape, and foam characteristics.

[0172] Specifically, in this embodiment, in order to quantify the degree of fusion and establish a scientific relationship between the degree of fusion and taste, it is necessary to establish a parameter-taste mapping database before step S1, such as Figure 3As shown in the structure diagram of the parameter-taste mapping database, the input of the parameter-taste mapping database is the user's taste requirements and the real-time characteristics of the milk foam, and the output is the corresponding fusion control trajectory parameters including the ellipse semi-axis, fusion depth, fusion circle number, and fusion time. Specifically,

[0173] Conducting experimental data collection, systematically varying fusion control trajectory parameters (including trajectory morphology parameters, fusion depth height_bias, number of fusion circles num_circles, and fusion duration duration_s), and milk foam characteristics (including carbonation, whipping degree, and temperature) to prepare multiple groups of differentiated beverages. The trajectory morphology parameters include the ellipse semi-axis, spiral radius, and Lissajous parameters.

[0174] Conduct sensory calibration, with professional baristas evaluating each drink group's bitter-sweet balance, sweetness intensity, and taste levels to generate expert-annotated data.

[0175] Quantify the degree of fusion of each group of drinks using a preset fusion calculation formula, establish a multivariate regression model of fusion control trajectory parameters + milk foam characteristics → fusion degree → sensory evaluation, and store it as the initial parameter-taste mapping database;

[0176] The calculation formula for the fusion degree is:

[0177] Fusion_Rate = (V_mixed / V_total) × K_foam × K_motion, where V_mixed is the volume actually mixed through vortex motion, V_total is the total volume of liquid in the cup, K_foam is the milk foam characteristic coefficient, which is determined by the gas content and the degree of whipping, and K_motion is the motion intensity coefficient, which is determined by the number of fusion circles, fusion depth and fusion speed.

[0178] In this embodiment, the milk foam characteristic coefficient is between 0.8 and 1.2. The milk foam characteristic is perceived by the system as follows: by analyzing the flow rate curve during pouring, the system can infer the gas content of the milk foam in real time. The flow rate curve for milk foam with high gas content is flatter, while the flow rate curve for milk foam with low gas content is steeper. Based on this characteristic, the system automatically adjusts the K_foam coefficient to ensure the target degree of blending is achieved under different milk foam conditions.

[0179] Furthermore, the parameter-taste mapping database adopts a cloud-based parameter-taste mapping database system, and the database system specifically includes:

[0180] Support multi-device data sharing and synchronization;

[0181] Use federated learning to protect user privacy;

[0182] Achieve cross-regional taste preference analysis;

[0183] Support crowdsourced development of new taste patterns.

[0184] In addition, in step S1, obtaining fusion control parameters according to user requirements includes: matching customized requirement input with parameters, receiving user taste requirements and real-time characteristics of milk foam, calling parameter-taste mapping database, and outputting corresponding fusion control trajectory parameters including trajectory morphology parameters, fusion depth, number of fusion circles, and fusion duration, specifically:

[0185] S11: Analyze user taste requirements and quantify the natural language input of users into quantitative indicators including bitter-sweet balance, sweetness perception intensity, and taste levels;

[0186] S12: Detect and infer the milk foam's real-time characteristics. The viscosity sensor obtains the milk foam's viscosity value in real time, and the temperature sensor obtains the milk foam's temperature. The flow rate variation curve during the injection process is analyzed to infer the milk foam's gas content. The degree of whipping is determined based on the viscosity and gas content.

[0187] S13: Calling and matching the parameter mapping database, selecting the historical data subset closest to the current milk foam characteristics in the parameter-taste mapping database, and based on the fusion target corresponding to the user's taste requirements and the selected trajectory shape, in the selected historical data subset:

[0188] Prioritize trajectory shape matching: Select the optimal trajectory type based on container shape and fusion requirements;

[0189] Matching trajectory parameters:

[0190] For elliptical trajectory: high-carbonated milk foam matches the larger semi-axis, low-carbonated milk foam matches the smaller semi-axis;

[0191] For spiral trajectory: adjust the pitch and radius change rate according to the fusion depth requirements;

[0192] For adaptive trajectories: set initial parameters and adjust sensitivity;

[0193] Match fusion depth height_bias: heavy whipping matches shallower depth, light whipping matches deeper depth;

[0194] Matching the number of fusion circles num_circles: Based on the fusion target and the milk foam characteristic coefficient, a low value is used for high carbonation / heavy whipping, and a high value is used for low carbonation / light whipping;

[0195] Matching fusion duration duration_s: positively correlated with the number of fusion cycles and compensated by the milk foam temperature;

[0196] S14: Perform dynamic parameter correction, including correction of fusion time based on temperature, correction of fusion number based on viscosity, correction of safe operating area, and correction of cup shape based on visual system recognition.

[0197] (1) Safe operating area modification: spill prevention and grease protection

[0198] Inappropriate parameter combinations can cause liquid to splash out of the cup or damage the coffee's precious crema layer due to excessive vortexes. A "Safe Operation Zone" must be defined to ensure that all parameter combinations fall within this area. To address the need to "avoid liquid splashing and protect the crema layer," the system pre-defines safe parameter boundaries:

[0199] Pitch angle constraint: Limit the pitch angle range (e.g., 15°≤θ≤60°) to prevent excessive pitch angles from causing liquid splashing, or excessive pitch angles from causing strong eddy currents that damage the grease layer (Crema).

[0200] Trajectory radius constraint: The ellipse semi-axis (radius_x / radius_y) does not exceed 80% of the cup mouth radius to prevent the milk foam from flowing beyond the cup body. Figure 6 Shown is a schematic diagram of the safe operating area boundary.

[0201] Eddy current intensity threshold: The upper limit of eddy current intensity (such as velocity gradient) is calibrated by simulation or experiment.

[0202] ≤50s -1 ), when the predicted eddy current intensity approaches the threshold, the number of fusion turns or the trajectory radius is automatically reduced to ensure the integrity of the grease layer.

[0203] (2) Cup shape adaptation correction: cross-cup shape fusion consistency

[0204] The same set of trajectory parameters will produce completely different fluid effects in wide-mouth cups and narrow-mouth cups. The algorithm needs to be able to automatically adjust the trajectory parameters based on the cup shape identified by the vision system to achieve a consistent fusion effect. Based on the cup shape recognition of the vision system (measuring the cup diameter and cup wall slope), the parameters are dynamically adapted:

[0205] Trajectory size scaling: The ellipse semi-axis (radius_x / radius_y) is scaled proportionally according to the cup mouth diameter. For example, if the diameter of a wide-mouth cup increases by 20%, the semi-axis will be scaled up by 20% to maintain the "relative coverage ratio" of the milk foam injection.

[0206] Height compensation adjustment: For narrow-mouth cups (steep cup walls), appropriately increase height_bias (e.g., +2mm) to prevent milk foam from bouncing back against the cup walls; for wide-mouth cups, reduce height_bias (e.g., -2mm) to prevent flow dispersion;

[0207] Effect normalization: By adjusting trajectory parameters, the relative vortex intensity generated by milk foam injection in different cup types is ensured to be consistent (for example, the vortex kinetic energy deviation is ≤10%), ultimately achieving cross-cup adaptation of "same taste requirement → same fusion effect".

[0208] (3) Synergistic logic of temperature and viscosity correction (supplementary association)

[0209] Temperature correction: For every ±5°C deviation of the milk foam temperature from the reference value (60°C), the froth fusion time will be adjusted by ±5%. At the same time, the froth fusion time will be adjusted according to the cup shape: narrow-mouth cups are temperature-sensitive, so the froth fusion time will be adjusted by an additional 2%.

[0210] Viscosity correction: For every ±10cP deviation of the milk foam viscosity from the baseline value (40cP), the number of fusion circles will be synchronized by ±1 circle, and the safety area will be dynamically contracted (for example, when the viscosity is high, the upper limit of the trajectory radius is reduced by 5%) to prevent high-viscosity milk foam from overflowing due to the increase in the number of circles.

[0211] To achieve precise control of the fusion process, two core technical issues must be addressed:

[0212] Precise control of robotic arm motion: The robotic arm needs to perform smooth and stable trajectories (such as continuous elliptical motion) in three-dimensional space to generate the desired liquid surface vortex.

[0213] Precisely matching liquid flow rates: The flow rate of milk foam injection must be precisely controlled while the robotic arm moves. Synchronizing the flow rate with the arm speed is key to forming a stable vortex, avoiding liquid splashing, and ultimately achieving repeatable blending results.

[0214] Therefore, the pursuit of "multiple tastes" directly gave rise to the need for two underlying technologies: "precise robotic arm control" and "precise liquid flow rate matching." The flow rate control technology of the present invention uses a hybrid model architecture, combining physical modeling (based on the Bernoulli equation and the Coanda effect) with machine learning (single-shot learning adaptation) to achieve precise flow rate prediction and control for different fluids. This flow rate control technology, combined with fusion trajectory generation technology, constitutes the core of the present invention.

[0215] like Figure 7 The three-dimensional elliptical trajectory generation and synchronous execution system diagram is shown in the following figure, which illustrates the precise control of the fusion process:

[0216] (1) Parameter Mapping: The "fusion level" selected by the user or set by the system is mapped to a set of specific physical parameters, such as radius_x, radius_y (fusion range), height_bias (fusion depth), and duration_s (fusion time / number of turns). For 100% fusion mode, the system automatically selects the maximum parameter value, such as a larger radius (up to 80% of the cup diameter), a deeper fusion depth (15-20mm), and a greater number of fusion turns (6-8 turns).

[0217] (2) Trajectory generation: Call the core go_elipses_3d function to generate an accurate three-dimensional elliptical (or circular) trajectory based on the above parameters. This trajectory defines all spatial positions of the end of the robotic arm during the fusion process.

[0218] (3) Synchronous Planning: The system performs motion planning and flow rate planning in parallel. On the one hand, it generates a 3D trajectory (X, Y, Z) in Cartesian space, and on the other hand, it calculates the required pouring posture (especially the pitch angle to control the flow rate) based on the fusion level. These two parts are converted into coordinated motion of each degree of freedom of the actuator through an inverse kinematics solver. In 100% fusion mode, injection and fusion actions are performed simultaneously, without waiting for injection to complete.

[0219] (4) Synthesis and Execution: All instructions for multiple controllable degrees of freedom are synthesized into a unified, time-synchronized trajectory and handed over to the controller for execution. The robotic arm will drive the latte art cylinder in a smooth circular motion while injecting milk foam, generating a stable vortex in the cup, thereby achieving efficient and controllable fusion.

[0220] The above process is achieved by the following steps S2-S4 of the present invention.

[0221] S2: Generate an execution trajectory based on the fusion control parameters.

[0222] In step S2, the execution mechanism for executing the execution trajectory is a motion device with at least four controllable degrees of freedom, including at least one of the following:

[0223] Four-axis to seven-axis serial robotic arms;

[0224] Parallel kinematic mechanisms with three to six degrees of freedom;

[0225] A combination of a robotic arm and one or more auxiliary motion platforms;

[0226] collaborative robotic systems;

[0227] SCARA robot with rotating platform.

[0228] The actuator adopts the following adaptive configuration scheme, specifically:

[0229] Automatic detection of connected actuator type and number of degrees of freedom;

[0230] Load the corresponding kinematic model according to the test results;

[0231] Adaptive adjustment of trajectory generation and control parameters;

[0232] Supports hot swapping and dynamic reconfiguration of actuators.

[0233] In step S2, the form of the three-dimensional trajectory as the execution trajectory can be dynamically selected or combined, including:

[0234] Basic geometric trajectories: ellipse, circle, spiral, polygon;

[0235] Complex parameter trajectories: Lissajous figures, fractal trajectories, Bezier curves;

[0236] Adaptive trajectory: A trajectory that is dynamically adjusted based on real-time fusion feedback;

[0237] User-defined trajectory: defined by parameterized equations or teaching.

[0238] In step S2, generating an execution trajectory based on the fusion control parameters includes: performing three-dimensional trajectory generation, calling a trajectory generation function based on the fusion control parameters to generate three-dimensional path points including ellipse, spiral, Lissajous figure or adaptive trajectory that changes with time, as the input path of the actuator motion planning, and outputting a Cartesian trajectory X, Y, Z sequence, specifically:

[0239] S21: Call the trajectory generation function go_ellipses_3d and enter the following parameters:

[0240] Trajectory type parameter: specifies the shape of the generated trajectory, including ellipse / spiral / lissajous / adaptive trajectory;

[0241] Center pose parameters: the spatial coordinates center_x, center_y, center_z defined by the geometric center of the cup, corresponding to the decomposition value of the center pose target_pose;

[0242] Fusion control parameters: including the x-axis radius_x and the y-axis radius_y of the ellipse semi-axis, which determine the major and minor axis sizes of the ellipse in the horizontal plane;

[0243] Number of fusion circles num_circles: determines the number of times the trajectory is repeated;

[0244] Height bias height_bias: determines the trajectory offset in the Z-axis direction;

[0245] Total trajectory duration_s: the total time to complete all fusion cycles;

[0246] S22: Trajectory generation is driven by discrete time steps dt_s. In each time step, the coordinates of the three-dimensional path points are calculated according to the following rules:

[0247] x(t)=center_x+radius_x*cos(2*pi*t / (duration_s / num_circles));

[0248] y(t)=center_y+radius_y*sin(2*pi*t / (duration_s / num_circles));

[0249] z(t)=center_z+height_bias;

[0250] Where t is the current time step, which ranges from greater than or equal to 0 to less than or equal to duration_s;

[0251] 2*pi*t / (duration_s / num_circles): defines the periodic motion angle of the ellipse, and each circle takes duration_s / num_circles;

[0252] center_x, center_y: X and Y coordinates of the center point of the ellipse trajectory;

[0253] radius_x, radius_y: the length of the semi-axis of the ellipse in the X and Y directions;

[0254] center_z: the initial Z coordinate of the center point of the trajectory;

[0255] height_bias: offset in the Z-axis direction, controlling the track height;

[0256] like Figure 8 As shown in the schematic diagram of the three-dimensional trajectory generated by the go_elipses_3d function, by combining these points, a complete, smooth spiral descent or fixed-height elliptical trajectory is formed. This trajectory is then fed into the robot's inverse kinematics solver and converted into specific rotation instructions for each joint. Simultaneously, a hybrid fluid control model (combining a physical model and a single-shot learning model) is invoked to predict the flow rate in real time based on the current pouring angle and remaining volume. Closed-loop control is then used to maintain a constant or slowly varying flow rate, ensuring the stability of the fusion process.

[0257] S23: Dynamically construct trajectory morphology

[0258] Fixed height ellipse: If height_bias is a constant value, an elliptical trajectory is generated in the horizontal plane, and the Z coordinate remains unchanged;

[0259] Spiral descending ellipse: If height_bias decreases linearly with the number of fusion turns, a trajectory that spirally contracts / expands along the negative Z-axis is generated;

[0260] S24: Output the three-dimensional path point sequence of continuous time steps as the X, Y, Z coordinate sequence of the Cartesian space trajectory of the robot arm motion planning, which serves as the input of the subsequent inverse kinematics solution.

[0261] S3: Control the injection flow rate. Figure 9 Shown is a schematic diagram of the effect of flow rate on fusion quality.

[0262] In step S3, a mixed fluid model is required to control the injection flow rate. The mixed fluid model includes at least one of the following or a combination thereof:

[0263] Fluid dynamics models based on physical laws;

[0264] Data-driven machine learning models, including but not limited to single-shot learning, deep learning, and reinforcement learning;

[0265] Fuzzy control model;

[0266] Expert system model;

[0267] Hybrid models combine the advantages of two or more of the above models.

[0268] In step S3, controlling the injection flow rate includes: performing flow rate control planning, based on a mixed fluid model including a physical model and a single learning model, combining the current pouring angle and the remaining volume, predicting and controlling the milk froth flow rate in real time, and outputting a sequence of pitch angles of the pouring posture:

[0269] A hybrid fluid control framework was constructed that integrated a physical model with a single-shot learning model. The physical model, based on the Bernoulli equation and the Coanda effect, combined geometric parameters such as the width and curvature of the spout of the latte art pot to derive the theoretical relationship between milk foam flow rate and pouring angle. The single-shot learning model analyzed historical injection data to establish a mapping relationship between pouring angle, residual volume, and flow rate deviation, dynamically correcting the prediction error of the physical model.

[0270] Execute the real-time flow rate prediction and control process. The current pouring pitch angle is collected through the IMU sensor, and the remaining volume of milk foam is obtained by an electronic scale or liquid level sensor. The theoretical flow rate is first calculated using the physical model, and then the correction value output by the single learning model is called and integrated to obtain the real-time predicted flow rate. Based on the deviation between the predicted flow rate and the target flow rate, the pouring angle is dynamically adjusted through the PID control algorithm to gradually converge the actual flow rate to the target value.

[0271] The output pitch angle sequence of the pouring posture is used as the input command for the robot arm motion planning, and dual constraints are set: first, the pouring angle is limited to a preset range to prevent milk foam overflow or too low a flow rate; second, the flow rate change rate is controlled to ensure a smooth and stable injection process and prevent sudden changes in flow rate from disrupting the fusion of milk foam and coffee.

[0272] The core of the flow rate control planning in step S3 lies in building a hybrid control logic of "theoretical derivation + empirical correction" and achieving precise and stable control of the milk froth flow rate through closed-loop regulation and constraint mechanisms:

[0273] (1) Construction logic of mixed fluid model

[0274] The physical model is based on the Bernoulli equation (which describes the relationship between flow velocity and potential energy) and the Coanda effect (the characteristic of milk foam flowing along the curved surface of the spout of a latte art pot). Combined with geometric parameters such as the spout width and curvature, the correlation formula of "pour angle → theoretical flow velocity" is derived. However, milk foam is a non-Newtonian fluid (viscosity changes with shear rate) and is affected by the gas content and the degree of whipping. Pure theoretical models are prone to deviations. Therefore, a single learning model is introduced. By analyzing the law of "pour angle-residual volume-actual flow velocity deviation" in historical injection data, a dynamic correction mapping is established to compensate for the prediction error of the physical model under complex fluid characteristics.

[0275] (2) Execution path of real-time flow rate prediction and control

[0276] The system uses the IMU sensor to capture the tilting angle (Pitch angle) of the robotic arm in real time, and uses the electronic scale / liquid level sensor to monitor the remaining volume of milk foam to provide input for the model. The physical model first calculates the theoretical flow rate corresponding to the "current angle + remaining volume", and then calls the single learning model to output the flow rate correction for the current working conditions based on historical data. The two are combined to obtain the real-time predicted flow rate. If there is a deviation between the predicted flow rate and the target flow rate, the tilting angle is dynamically adjusted through the PID control algorithm (for example, if the flow rate is too high, the angle is increased to slow down the injection, and if the flow rate is too low, the angle is reduced to speed up the injection) so that the actual flow rate gradually converges to the target value, forming a closed loop of "perception-prediction-control".

[0277] (3) The core significance of output constraints

[0278] Set dual constraints of "dumping angle range" and "flow rate change rate limit":

[0279] The pouring angle is limited to a preset range (e.g. 15° to 60°) to avoid the milk foam splashing out of the cup due to a large angle, or the flow impacting the coffee oil layer (Crema) due to a small angle, thus destroying the flavor base; the flow rate change rate is controlled (e.g. ≤0.5ml / s 2 ) to ensure a smooth and stable injection process - if the flow rate changes suddenly, the mixed flow field of milk foam and coffee will oscillate violently, which may easily break the milk foam bubbles or disrupt the layered structure, resulting in uncontrolled fusion effect.

[0280] Through the above design, step S3 not only ensures the accuracy of flow rate control (in line with theory and experience), but also takes into account safety (anti-overflow, oil protection) and stability (slowly changing flow rate to maintain flow field stability), laying a key foundation for the reproduction of customized taste.

[0281] S4: Execute the fusion operation and monitor the fusion status.

[0282] In step S4, the fusion status is monitored using a multimodal fusion degree detection technology, including a combination of at least two of the following detection methods:

[0283] Visual detection: at least one of visible light camera, multispectral camera, thermal imaging camera, and structured light 3D camera;

[0284] Electrical detection: at least one of conductivity sensor, dielectric constant sensor, and impedance spectrum analyzer;

[0285] Acoustic detection: at least one of ultrasonic sensor and acoustic impedance sensor;

[0286] Optical detection: at least one of laser scattering instrument, turbidimeter, and fluorescence spectrometer;

[0287] Other detection: at least one of pH sensor, viscosity sensor, and density sensor.

[0288] In step S4, performing the fusion operation and monitoring the fusion status includes: synthesizing the Cartesian trajectory X, Y, Z sequence and the pitch angle sequence through a trajectory synthesizer to generate a final synchronous trajectory, converting the final synchronous trajectory into an actuator control instruction to drive the injection action, calculating the real-time fusion degree through image and conductivity detection, and correcting the trajectory parameters and flow rate based on the deviation between the fusion degree and the target value to ensure that the target fusion degree is achieved. Specifically,

[0289] The trajectory synthesizer receives the Cartesian trajectory X, Y, and Z sequence and the pitch angle sequence, and achieves synchronous synthesis through time axis alignment: based on the total trajectory duration, the spatial path points with X, Y, and Z coordinates are bound to the corresponding pitch angle at the time, generating a final synchronized trajectory containing position and posture information, ensuring that the robot arm synchronizes with the preset tilt angle changes while performing spatial motion;

[0290] The final synchronized trajectory is input into the robotic arm motion controller, which converts it into angle commands for each joint through inverse kinematics. This drives the robotic arm to execute the injection motion according to the planned path. During the injection process, a high-speed camera captures real-time images of the interface between the milk foam and coffee. Combined with a conductivity probe, the uniformity of the mixture is tested, and the real-time degree of integration, including layered integration or uniform integration, is comprehensively calculated.

[0291] Based on the deviation between the real-time detection value of the fusion degree and the target value, the control parameters are dynamically corrected: if the layered fusion degree is too high, the number of trajectory circles is reduced and the rate of change of the pouring angle is lowered; if the uniform fusion degree is insufficient, the trajectory radius is increased and the flow rate is increased. The corrected parameters are updated to the trajectory synthesizer in real time to ensure that the final fusion degree converges to the target range and guarantees the stable reproduction of the preset taste.

[0292] The core of step S4 is to build a control hub for "trajectory-posture synchronous execution + fusion degree closed-loop correction", from spatiotemporal coordination, multi-modal detection to dynamic control, to ensure customized fusion effects layer by layer:

[0293] (1) Trajectory synthesis: precise coupling of spatiotemporal coordination

[0294] The key to the trajectory synthesizer is "time axis alignment" - binding the three-dimensional space trajectory (X, Y, Z coordinate sequence) to the pouring posture (Pitch angle sequence) by timestamp. For example, when the robotic arm moves to a certain spatial point on the elliptical trajectory, it must synchronously match the preset Pitch angle to ensure that the angle, speed and flow field planning of the milk foam injection are completely consistent with the flow field planning. This spatiotemporal coordination of "spatial position + posture angle" directly determines the initial collision form of the milk foam and coffee during the injection process: if the time is misaligned, the preset vortex structure will be destroyed at best, and the milk foam will splash at worst. Therefore, "synchronous binding based on the total duration of the trajectory" is the underlying prerequisite for achieving precise fusion.

[0295] (2) Execution and Detection: Fusion Perception of Multimodal Data

[0296] The robot arm relies on inverse kinematics calculations to convert Cartesian trajectory instructions into joint angle instructions, driving the robot arm to move along the planned path. Fusion detection uses a dual-dimensional solution of "vision + conductivity":

[0297] A high-speed camera captures the interface morphology between milk foam and coffee (e.g., layer thickness and boundary definition), visually reflecting the degree of layer integration (the thicker the interface, the more pronounced the layers). A conductivity probe detects conductivity fluctuations in the mixture (the smaller the standard deviation, the more evenly mixed), quantifying the degree of integration. The two complement each other: vision qualitatively determines whether layers are present, while conductivity quantitatively calculates the degree of integration. Together, they construct a comprehensive picture of the integration state, overcoming the limitations of a single sensor (e.g., vision is easily affected by bubbles, while conductivity cannot identify interface morphology).

[0298] (3) Closed-loop correction: Dynamic regulation of differentiation strategy

[0299] Based on the correction logic of integration deviation, differentiated strategies are designed for the two requirements of "stratification" and "uniformity":

[0300] If the degree of stratification and fusion is too high (e.g., the goal is "clear layers" but the actual mixing is excessive): suppress excessive fusion by reducing the number of trajectory turns (reducing the number of mixing times and retaining the stratification boundaries) and reducing the rate of change of the pitch angle (slowing down the adjustment of the pouring angle to avoid sudden changes in flow rate that exacerbate mixing); if the degree of uniform fusion is insufficient (e.g., the goal is "dense and smooth" but the mixing is uneven): improve fusion efficiency by increasing the trajectory radius (expanding the injection coverage range and strengthening vortex stirring) and increasing the flow rate (increasing the mixing volume per unit time and accelerating homogenization). The corrected parameters are fed back to the trajectory synthesizer in real time, forming a closed loop of "execution → detection → correction → re-execution" to ensure that the fusion degree converges to the target range (error ≤±5%) and ultimately stably reproduce the preset taste.

[0301] The technical value of step S4 lies in establishing a closed loop from "planning → execution → feedback": spatiotemporal synchronization ensures precision, seamlessly connecting "trajectory design" and "actual injection flow"; multi-mode detection enables quantitative perception of fusion effects, addressing the industry's dilemma of relying solely on experience for taste quality; and dynamic correction provides precise control for different fusion requirements, effectively achieving customized goals of "layering" and "uniformity." This step serves as both the execution endpoint for trajectory and flow rate control and the quality gate for subsequent taste verification, supporting the critical transition from "parameter output" to "taste reproduction" for automated coffee machines.

[0302] S5: Record and optimize fusion data.

[0303] In step S5, recording and optimizing the fusion data includes: performing taste verification and database update to verify the matching degree between the taste of the beverage and the user's needs, entering the new data of the fusion control trajectory parameters + milk foam characteristics → fusion degree → sensory evaluation into the parameter-taste mapping database, and optimizing the subsequent matching accuracy, specifically:

[0304] After the drink is prepared, the taste compatibility is evaluated through a multi-dimensional verification mechanism: image analysis technology is used to identify the layered structure or uniformity of the milk foam and coffee, and the actual degree of integration is compared with the target value. An electronic tongue sensor collects sweetness and bitterness flavor data, and combines it with preset sensory evaluation standards to generate a quantitative taste score. The system integrates visual characteristics and flavor data to form a final judgment on whether the drink tastes meet user requirements.

[0305] Recording the complete preparation process data into the parameter-taste mapping database: specifically, fusion control trajectory parameters including ellipse semi-axis, fusion depth, number of fusion turns, and fusion time, milk foam characteristics including carbonation, whipping degree, and temperature, real-time fusion degree data, and sensory evaluation results. Each record maintains the mapping relationship between parameters, forming a complete data chain of trajectory parameters + milk foam characteristics → fusion degree → sensory evaluation.

[0306] Iteratively optimize the mapping model based on newly added data: The system regularly analyzes the deviation between historical data and newly added data, adjusts the mapping relationship between parameters and taste scores through machine learning algorithms, pays attention to the taste deviation scenarios that appear frequently in user feedback, and specifically strengthens the matching accuracy of relevant parameter combinations, gradually improving the database's prediction accuracy for new milk foam types and personalized taste requirements.

[0307] Furthermore, the method for achieving beverage blending with customized taste and flavor also includes using a modular beverage blending control device to blend beverages, specifically including:

[0308] Core control module, compatible with different types of actuators;

[0309] Replaceable actuator interface, supporting 4-7 axis robotic arms and other motion devices;

[0310] Scalable sensor interface to support multimodal sensing devices;

[0311] Unified software architecture, adaptive to different hardware configurations.

[0312] Furthermore, the method for achieving customized taste and flavor beverage fusion also includes optimizing the taste based on user feedback, specifically including:

[0313] Collect real-time user feedback through mobile apps or touch screens;

[0314] Reinforcement learning algorithm is used to optimize fusion parameters;

[0315] Support personalized taste profiles;

[0316] Achieve taste reproduction across devices.

[0317] Second embodiment

[0318] like Figure 10 As shown, this embodiment provides a beverage fusion system for achieving customized taste and flavor for executing the beverage fusion method for achieving customized taste and flavor as in the first embodiment, comprising:

[0319] Fusion parameter acquisition module 1, used to obtain fusion control parameters according to user needs;

[0320] An execution trajectory generating module 2, configured to generate an execution trajectory based on the fusion control parameters;

[0321] Injection flow rate control module 3, used to control the injection flow rate;

[0322] A fusion execution monitoring module 4 is used to execute fusion operations and monitor fusion status;

[0323] The fusion data optimization module 5 is used to record and optimize the fusion data.

[0324] Third embodiment

[0325] This example provides an economical configuration using a 4-axis SCARA robot. This example demonstrates how to use a 4-axis SCARA robot to achieve beverage blending control. The SCARA robot has three rotational axes and one vertical axis. Although it has fewer degrees of freedom, it can still achieve effective blending control through clever trajectory design:

[0326] Generate fusion trajectory in the horizontal plane using the first three rotation axes;

[0327] The 4th axis controls vertical height to achieve layered fusion;

[0328] Compensate for the shortcomings of attitude control by optimizing trajectory parameters;

[0329] The cost is reduced by more than 50%, which is suitable for small and medium-sized coffee shops.

[0330] Fourth embodiment

[0331] This embodiment provides a high-speed configuration using a Delta parallel robot. This embodiment uses a 3-DOF Delta parallel robot in conjunction with a 2-axis rotation platform to achieve high-speed and precise fusion control:

[0332] Delta robots provide high-speed XYZ motion (up to 10m / s);

[0333] The rotating platform provides tilt angle control;

[0334] Suitable for commercial applications that require rapid integration;

[0335] Fusion time is shortened by 40%, improving production efficiency.

[0336] Fifth embodiment

[0337] This embodiment provides an intelligent configuration for multi-sensor fusion. This embodiment demonstrates the use of a low-cost sensor combination to achieve high-precision fusion detection:

[0338] Use ordinary camera + image processing algorithm instead of high-speed camera

[0339] Using a multi-point conductivity sensor array to improve detection resolution

[0340] Add ultrasonic sensors to detect liquid level and density distribution

[0341] Fusion of multi-sensor data through Kalman filtering

[0342] Detection costs are reduced by 70%, and accuracy is maintained within ±3%.

[0343] Sixth embodiment:

[0344] This embodiment provides an adaptive trajectory generation configuration and implements adaptive trajectory optimization based on reinforcement learning:

[0345] Initially use the basic elliptical trajectory;

[0346] Monitor the fusion effect in real time and calculate the reward function;

[0347] Dynamically adjust trajectory parameters through reinforcement learning algorithms;

[0348] Support online learning and continuously optimize integration strategies;

[0349] The ability to adapt to new beverages and special containers has increased by 80%.

[0350] A computer-readable storage medium stores computer code. When the computer code is executed, the above-described method is performed. A person skilled in the art will appreciate that all or part of the steps in the various methods of the above-described embodiments can be performed by a program instructing related hardware. The program can be stored in a computer-readable storage medium. The storage medium may include a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0351] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiment. All technical solutions based on the concept of the present invention are within the scope of protection of the present invention. It should be noted that for those skilled in the art, various improvements and modifications that do not depart from the principles of the present invention should also be considered within the scope of protection of the present invention.

[0352] The technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above-mentioned embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0353] It should be noted that the above embodiments can be freely combined as needed. The above description is only a preferred embodiment of the present invention. It should be pointed out that those skilled in the art can make several improvements and modifications without departing from the principles of the present invention, and such improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A beverage fusion method for achieving customized taste and flavor, characterized in that: The following steps are involved: S1: Obtain fusion control parameters according to user requirements; S2: generating an execution trajectory based on the fusion control parameters; S3: control injection flow rate; S4: perform fusion operation and monitor fusion status; S5: Record and optimize fusion data.

2. The beverage fusion method for achieving customized taste and flavor according to claim 1, characterized in that: In step S2, the execution mechanism for executing the execution trajectory is a motion device with at least four controllable degrees of freedom, including at least one of the following: Four-axis to seven-axis serial robotic arms; Parallel kinematic mechanisms with three to six degrees of freedom; A combination of a robotic arm and one or more auxiliary motion platforms; collaborative robotic systems; SCARA robot with rotating platform.

3. The beverage fusion method for achieving customized taste and flavor according to claim 2, characterized in that: The actuator adopts the following adaptive configuration scheme, specifically: Automatic detection of connected actuator type and number of degrees of freedom; Load the corresponding kinematic model according to the test results; Adaptive adjustment of trajectory generation and control parameters; Supports hot swapping and dynamic reconfiguration of actuators.

4. The beverage fusion method for achieving customized taste and flavor according to claim 1, characterized in that: In step S2, the form of the three-dimensional trajectory as the execution trajectory can be dynamically selected or combined, including: Basic geometric trajectories: ellipse, circle, spiral, polygon; Complex parameter trajectories: Lissajous figures, fractal trajectories, Bezier curves; Adaptive trajectory: A trajectory that is dynamically adjusted based on real-time fusion feedback; User-defined trajectory: defined by parameterized equations or teaching.

5. The beverage fusion method for achieving customized taste and flavor according to claim 1, characterized in that: In step S4, the fusion status is monitored using a multimodal fusion degree detection technology, including a combination of at least two of the following detection methods: Visual detection: at least one of visible light camera, multispectral camera, thermal imaging camera, and structured light 3D camera; Electrical detection: at least one of conductivity sensor, dielectric constant sensor, and impedance spectrum analyzer; Acoustic detection: at least one of ultrasonic sensor and acoustic impedance sensor; Optical detection: at least one of laser scattering instrument, turbidimeter, and fluorescence spectrometer; Other detection: at least one of pH sensor, viscosity sensor, and density sensor.

6. The beverage fusion method for achieving customized taste and flavor according to claim 1, characterized in that: In step S3, a mixed fluid model is required to control the injection flow rate. The mixed fluid model includes at least one of the following or a combination thereof: Fluid dynamics models based on physical laws; Data-driven machine learning models, including but not limited to single-shot learning, deep learning, and reinforcement learning; Fuzzy control model; Expert system model; Hybrid models combine the advantages of two or more of the above models.

7. The beverage fusion method for achieving customized taste and flavor according to claim 1, characterized in that: Before step S1, a parameter-taste mapping database is established. The input of the parameter-taste mapping database is the user's taste requirements and the real-time characteristics of the milk foam, and the output is the corresponding fusion control trajectory parameters including the ellipse semi-axis, fusion depth, fusion number of circles, and fusion time. Specifically, Conducting experimental data collection, systematically varying fusion control trajectory parameters (including trajectory morphology parameters, fusion depth height_bias, number of fusion circles num_circles, and fusion duration duration_s), and milk foam characteristics (including carbonation, whipping degree, and temperature) to prepare multiple groups of differentiated beverages. The trajectory morphology parameters include the ellipse semi-axis, spiral radius, and Lissajous parameters. Conduct sensory calibration, with professional baristas evaluating each drink group's bitter-sweet balance, sweetness intensity, and taste levels to generate expert-annotated data. Quantify the degree of fusion of each group of drinks using a preset fusion calculation formula, establish a multivariate regression model of fusion control trajectory parameters + milk foam characteristics → fusion degree → sensory evaluation, and store it as the initial parameter-taste mapping database; The calculation formula for the fusion degree is: Fusion_Rate = (V_mixed / V_total) × K_foam × K_motion, where V_mixed is the volume actually mixed through vortex motion, V_total is the total volume of liquid in the cup, K_foam is the milk foam characteristic coefficient, which is determined by the gas content and the degree of whipping, and K_motion is the motion intensity coefficient, which is determined by the number of fusion circles, fusion depth and fusion speed.

8. The beverage fusion method for achieving customized taste and flavor according to claim 7, characterized in that: The parameter-taste mapping database adopts a cloud-based parameter-taste mapping database system, and the database system specifically includes: Support multi-device data sharing and synchronization; Use federated learning to protect user privacy; Achieve cross-regional taste preference analysis; Support crowdsourced development of new taste patterns.

9. The beverage fusion method for achieving customized taste and flavor according to claim 7, characterized in that: In step S1, obtaining fusion control parameters according to user requirements includes: matching customized requirements input with parameters, receiving user taste requirements and real-time characteristics of milk foam, calling parameter-taste mapping database, and outputting corresponding fusion control trajectory parameters including trajectory morphology parameters, fusion depth, number of fusion circles, and fusion duration. Specifically, S11: Analyze user taste requirements and quantify the natural language input of users into quantitative indicators including bitter-sweet balance, sweetness perception intensity, and taste levels; S12: Detect and infer the milk foam's real-time characteristics. The viscosity sensor obtains the milk foam's viscosity value in real time, and the temperature sensor obtains the milk foam's temperature. The flow rate variation curve during the injection process is analyzed to infer the milk foam's gas content. The degree of whipping is determined based on the viscosity and gas content. S13: Calling and matching the parameter mapping database, selecting the historical data subset closest to the current milk foam characteristics in the parameter-taste mapping database, and based on the fusion target corresponding to the user's taste requirements and the selected trajectory shape, in the selected historical data subset: Prioritize trajectory shape matching: Select the optimal trajectory type based on container shape and fusion requirements; Matching trajectory parameters: For elliptical trajectory: high-carbonated milk foam matches the larger semi-axis, low-carbonated milk foam matches the smaller semi-axis; For spiral trajectory: adjust the pitch and radius change rate according to the fusion depth requirements; For adaptive trajectories: set initial parameters and adjust sensitivity; Match fusion depth height_bias: heavy whipping matches shallower depth, light whipping matches deeper depth; Matching the number of fusion circles num_circles: Based on the fusion target and the milk foam characteristic coefficient, a low value is used for high carbonation / heavy whipping, and a high value is used for low carbonation / light whipping; Matching fusion duration duration_s: positively correlated with the number of fusion cycles and compensated by the milk foam temperature; S14: Perform dynamic parameter correction, including correction of fusion time based on temperature, correction of fusion number based on viscosity, correction of safe operating area, and correction of cup shape based on visual system recognition.

10. The beverage fusion method for achieving customized taste and flavor according to claim 1, characterized in that: In step S2, generating an execution trajectory based on the fusion control parameters includes: performing three-dimensional trajectory generation, calling a trajectory generation function based on the fusion control parameters to generate three-dimensional path points including ellipse, spiral, Lissajous figure or adaptive trajectory that changes with time, as the input path of the actuator motion planning, and outputting a Cartesian trajectory X, Y, Z sequence, specifically: S21: Call the trajectory generation function go_ellipses_3d and enter the following parameters: Trajectory type parameter: specifies the shape of the generated trajectory, including ellipse / spiral / lissajous / adaptive trajectory; Center pose parameters: the spatial coordinates center_x, center_y, center_z defined by the geometric center of the cup, corresponding to the decomposition value of the center pose target_pose; Fusion control parameters: including the x-axis radius_x and the y-axis radius_y of the ellipse semi-axis, which determine the major and minor axis sizes of the ellipse in the horizontal plane; Number of fusion circles num_circles: determines the number of times the trajectory is repeated; Height bias height_bias: determines the trajectory offset in the Z-axis direction; Total trajectory duration_s: the total time to complete all fusion cycles; S22: Trajectory generation is driven by discrete time steps dt_s. In each time step, the coordinates of the three-dimensional path points are calculated according to the following rules: x(t)=center_x+radius_x*cos(2*pi*t / (duration_s / num_circles)); y(t)=center_y+radius_y*sin(2*pi*t / (duration_s / num_circles)); z(t)=center_z+height_bias; Where t is the current time step, which ranges from greater than or equal to 0 to less than or equal to duration_s; 2*pi*t / (duration_s / num_circles): defines the periodic motion angle of the ellipse, and each circle takes duration_s / num_circles; center_x, center_y: X and Y coordinates of the center point of the ellipse trajectory; radius_x, radius_y: the length of the semi-axis of the ellipse in the X and Y directions; center_z: the initial Z coordinate of the center point of the trajectory; height_bias: offset in the Z-axis direction, controlling the track height; S23: Dynamically construct trajectory morphology Fixed height ellipse: If height_bias is a constant value, an elliptical trajectory is generated in the horizontal plane, and the Z coordinate remains unchanged; Spiral descending ellipse: If height_bias decreases linearly with the number of fusion turns, a trajectory that spirally contracts / expands along the negative Z-axis is generated; S24: Output the three-dimensional path point sequence of continuous time steps as the X, Y, Z coordinate sequence of the Cartesian space trajectory of the robot arm motion planning, which serves as the input of the subsequent inverse kinematics solution.

11. The beverage fusion method for achieving customized taste and flavor according to claim 1, characterized in that: In step S3, controlling the injection flow rate includes: performing flow rate control planning, based on a mixed fluid model including a physical model and a single learning model, combining the current pouring angle and the remaining volume, predicting and controlling the milk foam flow rate in real time, and outputting a sequence of pitch angles of the pouring posture, specifically: A hybrid fluid control framework was constructed that integrated a physical model with a single-shot learning model. The physical model, based on the Bernoulli equation and the Coanda effect, combined geometric parameters such as the width and curvature of the spout of the latte art pot to derive the theoretical relationship between milk foam flow rate and pouring angle. The single-shot learning model analyzed historical injection data to establish a mapping relationship between pouring angle, residual volume, and flow rate deviation, dynamically correcting the prediction error of the physical model. Execute the real-time flow rate prediction and control process. The current pouring pitch angle is collected through the IMU sensor, and the remaining volume of milk foam is obtained by an electronic scale or liquid level sensor. The theoretical flow rate is first calculated using the physical model, and then the correction value output by the single learning model is called and integrated to obtain the real-time predicted flow rate. Based on the deviation between the predicted flow rate and the target flow rate, the pouring angle is dynamically adjusted through the PID control algorithm to gradually converge the actual flow rate to the target value. The output pitch angle sequence of the pouring posture is used as the input command for the robot arm motion planning, and dual constraints are set: first, the pouring angle is limited to a preset range to prevent milk foam overflow or too low a flow rate; second, the flow rate change rate is controlled to ensure a smooth and stable injection process and prevent sudden changes in flow rate from disrupting the fusion of milk foam and coffee.

12. The beverage fusion method for achieving customized taste and flavor according to claim 1, characterized in that: In step S4, performing the fusion operation and monitoring the fusion status includes: synthesizing the Cartesian trajectory X, Y, Z sequence and the pitch angle sequence through a trajectory synthesizer to generate a final synchronous trajectory, converting the final synchronous trajectory into an actuator control instruction to drive the injection action, calculating the real-time fusion degree through image and conductivity detection, and correcting the trajectory parameters and flow rate based on the deviation between the fusion degree and the target value to ensure that the target fusion degree is achieved. Specifically, The trajectory synthesizer receives the Cartesian trajectory X, Y, and Z sequence and the pitch angle sequence, and achieves synchronous synthesis through time axis alignment: based on the total trajectory duration, the spatial path points with X, Y, and Z coordinates are bound to the corresponding pitch angle at the time, generating a final synchronized trajectory containing position and posture information, ensuring that the robot arm synchronizes with the preset tilt angle changes while performing spatial motion; The final synchronized trajectory is input into the robotic arm motion controller, which converts it into angle commands for each joint through inverse kinematics. This drives the robotic arm to execute the injection motion according to the planned path. During the injection process, a high-speed camera captures real-time images of the interface between the milk foam and coffee. Combined with a conductivity probe, the uniformity of the mixture is tested, and the real-time degree of integration, including layered integration or uniform integration, is comprehensively calculated. Based on the deviation between the real-time detection value of the fusion degree and the target value, the control parameters are dynamically corrected: if the layered fusion degree is too high, the number of trajectory circles is reduced and the rate of change of the pouring angle is lowered; if the uniform fusion degree is insufficient, the trajectory radius is increased and the flow rate is increased. The corrected parameters are updated to the trajectory synthesizer in real time to ensure that the final fusion degree converges to the target range and guarantees the stable reproduction of the preset taste.

13. The beverage fusion method for achieving customized taste and flavor according to claim 1, characterized in that: In step S5, recording and optimizing the fusion data includes: performing taste verification and database update to verify the matching degree between the taste of the beverage and the user's needs, entering the new data of the fusion control trajectory parameters + milk foam characteristics → fusion degree → sensory evaluation into the parameter-taste mapping database, and optimizing the subsequent matching accuracy, specifically: After the drink is prepared, the taste compatibility is evaluated through a multi-dimensional verification mechanism: image analysis technology is used to identify the layered structure or uniformity of the milk foam and coffee, and the actual degree of integration is compared with the target value. An electronic tongue sensor collects sweetness and bitterness flavor data, and combines it with preset sensory evaluation standards to generate a quantitative taste score. The system integrates visual characteristics and flavor data to form a final judgment on whether the drink tastes meet user requirements. Recording the complete preparation process data into the parameter-taste mapping database: specifically, fusion control trajectory parameters including ellipse semi-axis, fusion depth, number of fusion turns, and fusion time, milk foam characteristics including carbonation, whipping degree, and temperature, real-time fusion degree data, and sensory evaluation results. Each record maintains the mapping relationship between parameters, forming a complete data chain of trajectory parameters + milk foam characteristics → fusion degree → sensory evaluation. Iteratively optimize the mapping model based on newly added data: The system regularly analyzes the deviation between historical data and newly added data, adjusts the mapping relationship between parameters and taste scores through machine learning algorithms, pays attention to the taste deviation scenarios that appear frequently in user feedback, and specifically strengthens the matching accuracy of relevant parameter combinations, gradually improving the database's prediction accuracy for new milk foam types and personalized taste requirements.

14. The beverage fusion method for achieving customized taste and flavor according to claim 1, characterized in that: It also includes beverage blending using a modular beverage blending control device, specifically including: Core control module, compatible with different types of actuators; Replaceable actuator interface, supporting 4-7 axis robotic arms and other motion devices; Scalable sensor interface to support multimodal sensing devices; Unified software architecture, adaptive to different hardware configurations.

15. The beverage fusion method for achieving customized taste and flavor according to claim 1, characterized in that: It also includes taste optimization based on user feedback, including: Collect real-time user feedback through mobile apps or touch screens; Reinforcement learning algorithm is used to optimize fusion parameters; Support personalized taste profiles; Achieve taste reproduction across devices.

16. A beverage fusion system for achieving customized taste and flavor, for executing the beverage fusion method for achieving customized taste and flavor according to any one of claims 1 to 15, characterized in that: include: Fusion parameter acquisition module, used to obtain fusion control parameters according to user needs; An execution trajectory generation module, configured to generate an execution trajectory based on the fusion control parameters; An injection flow rate control module, used for controlling the injection flow rate; A fusion execution monitoring module is used to execute fusion operations and monitor fusion status; The fusion data optimization module is used to record and optimize the fusion data.

Images