Control method and control system for intelligent front flushing closestool

Abstract

The invention discloses a control method and a control system for an intelligent front flushing closestool, and relates to the technical field of intelligent control of sanitary ware. The method comprises the following steps: based on a target detection mechanism of multi-source information fusion, identifying a flushing target type by collecting a sitting posture state, a forward inclination angle and target area position data, and generating a flushing command parameter set; in combination with the real-time water temperature, water pressure and nozzle state, multiple flushing control vectors are constructed and dynamically converted into an execution sequence, and a nozzle is driven to perform accurate control; flushing feedback information is collected, the recognition model and the control parameters are updated, and a closed-loop flushing control path is constructed. Through multi-source information fusion identification, dynamic flushing control sequence generation and a feedback-based closed-loop regulation and control mechanism, high-precision target identification, efficient flushing execution and adaptive control optimization are realized.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology for sanitary ware, specifically to a control method and control system for an intelligent pre-flush toilet. Background Technology

[0002] Currently, smart toilets are widely used in home hygiene, with core functions including automatic flushing, heating, deodorization, drying, and multiple cleaning modes. Traditional smart toilets mainly rely on timed spraying, preset pressure, and fixed-angle flushing control modes to complete the cleaning process. Some high-end devices have introduced user motion recognition and water pressure adjustment mechanisms, improving the level of intelligence in certain areas. However, existing technologies still have significant limitations in the precise control, dynamic recognition, and adaptive adjustment capabilities of the pre-flush function.

[0003] Current flushing control systems generally lack a fusion sensing mechanism for multi-source status information, making it impossible to accurately identify different user sitting postures, forward tilt angles, and corresponding flushing target areas. This results in difficulty in achieving differentiated and personalized adjustment of nozzle control paths.

[0004] In existing systems, the identification of rinsing targets mostly relies on static sensor feedback or fixed area mapping, lacking the ability to model changes in the forward area structure in real time. This makes it difficult to effectively adapt to the micro-motion adjustments made by users during the rinsing process, affecting rinsing accuracy and cleaning effect.

[0005] Traditional flushing path control strategies typically employ fixed templates and linear control commands, lacking interaction with flushing efficiency feedback information and failing to enable the control model to learn and continuously optimize based on actual flushing results.

[0006] Currently, there is no effective mechanism to achieve joint closed-loop updates of nozzle control behavior and recognition model after rinsing. The rinsing control path is prone to error accumulation, recognition drift and efficiency decay during long-term use, affecting the system's continuous adaptability and user experience.

[0007] Therefore, there is an urgent need for an intelligent control method and system architecture for forward flushing tasks, which can realize flushing target detection based on multi-source fusion, flushing command generation combined with dynamic feature recognition model, and adaptive regulation mechanism based on flushing efficiency feedback, so as to build an intelligent flushing control path with closed-loop capability, thereby improving the overall performance of toilet system in terms of cleaning accuracy, intelligent use and long-term stability. Summary of the Invention

[0008] Based on the shortcomings of the prior art described above, the purpose of this invention is to provide a control method and control system for an intelligent pre-flush toilet to solve the aforementioned technical problems.

[0009] To achieve the above objectives, the present invention provides the following technical solution: a control method for an intelligent pre-flush toilet, comprising:

[0010] A target detection mechanism based on multi-source information fusion is constructed, collecting sitting posture data, forward tilt angle data, and relative position data of the target area, and constructing a set of flushing target partition labels based on structural adjustment state variables;

[0011] The execution part identification and judgment process involves matching the features of the flushing target partition label set with the forward part identification model to filter out the current flushing target type, and generating a flushing command parameter set based on structural state constraints.

[0012] The system acquires real-time water temperature, water pressure, and nozzle spray status parameters, and generates multiple flushing control vectors based on the location labels and flushing intensity indicators in the flushing command parameter set.

[0013] By combining the nozzle rotation angle and the water pressure dynamic adjustment coefficient, a nonlinear transformation is performed on multiple flushing control vectors to form a flushing action execution sequence.

[0014] The flushing action sequence is mapped to the nozzle drive control command to achieve joint control of the spray angle, water pressure and flow rate, and water output rhythm;

[0015] The displacement status feedback and flushing residue feedback after flushing are collected, and the flushing efficiency index vector is reconstructed by combining the initial flushing parameters.

[0016] The forward part identification model and flushing command parameter set are updated based on the flushing efficiency index vector to form a closed-loop flushing control path.

[0017] The present invention is further configured such that the construction of the target detection mechanism based on multi-source information fusion includes:

[0018] Collect sitting posture data to generate information describing the pressure distribution of the user on the toilet surface;

[0019] Extract forward tilt angle data and construct a height sequence structure that reflects the degree of upper body tilt of the user;

[0020] Acquire the relative position data of the target area, and extract the relative positioning parameters of the forward region in the spatial coordinate system based on the point cloud depth information;

[0021] The sitting posture data, forward tilt angle data, and target area relative position data are fused to construct a structural adjustment state quantity. By matching it with predefined area mapping rules, a set of flushing target partition labels is generated.

[0022] The present invention is further configured such that the execution part identification and judgment process includes:

[0023] The target partition label set for rinsing is embedded to form a high-dimensional label structure feature;

[0024] The label structure features are matched with the baseline feature set in the preset forward part recognition model. The optimal index recognition is performed based on the multidimensional feature differences to select the part recognition results corresponding to the current flushing target type.

[0025] Construct a state constraint vector based on the current flushing target type and structural state variables;

[0026] The state constraint vector is fused and mapped with the control template corresponding to the recognition result to generate a set of flushing command parameters for flushing control path scheduling.

[0027] The present invention is further configured such that the generation of multiple flushing control vectors includes:

[0028] Collect water temperature signals, water pressure signals, and nozzle spray status information during the current flushing cycle to construct an operational status observation vector that includes thermal state, hydraulic state, and structural attitude;

[0029] Based on the flushing target type index and intensity index vector in the flushing command parameter set, the corresponding part adjustment structure is matched, and the target type adjustment tensor is constructed.

[0030] The running status observation vector and the target type adjustment tensor are fused and mapped to generate a set of multiple flushing control vectors.

[0031] The present invention is further configured such that the formation of the rinsing action execution sequence includes:

[0032] Collect the multi-degree-of-freedom rotation angle parameters of the nozzle and the corresponding dynamic water pressure adjustment coefficient, and construct a mapping kernel function based on the nonlinear coupling relationship between the two;

[0033] A nonlinear transformation is performed on multiple flushing control vectors based on a mapping kernel function to generate the transformed flushing control signal.

[0034] The transformed control signals are dynamically updated based on the time series to form a multi-dimensional, multi-temporal flushing action execution sequence.

[0035] The present invention is further configured such that the combined control of the injection angle, water pressure flow rate, and water outlet rhythm includes:

[0036] Multidimensional temporal feature extraction is performed on the flushing action execution sequence;

[0037] Construct a drive command generation function, and combine it with preset drive weight parameters to map multidimensional features into nozzle injection angle adjustment commands, water pressure and flow rate adjustment commands, and water output rhythm control commands;

[0038] Perform nonlinear constraint mapping on the generated drive commands to ensure that the control parameters are within a safe range;

[0039] Based on the clock-synchronized output of standardized drive commands, the joint dynamic control of the nozzle is completed.

[0040] The present invention is further configured such that the reconstructed flushing efficiency index vector includes:

[0041] Collect the displacement status feedback and flushing residue feedback after the flushing is completed, and construct a set of feedback information reflecting the nozzle operating accuracy and the degree of flushing residue.

[0042] Extract the set of parameters set during the initial flushing phase to form the initial control reference vector;

[0043] The feedback information set is fused with the initial control reference vector to generate a multi-dimensional structural tensor. The structural tensor is then transformed according to a preset nonlinear mapping rule to generate a flushing efficiency index vector.

[0044] The present invention is further configured such that the updated forward part identification model and flushing command parameter set include:

[0045] Extract the multidimensional feedback quantity from the flushing efficiency index vector, construct the mapping response structure, and perform nonlinear feature adjustment on the baseline feature set of each target part in the forward part recognition model according to the feature matrix corresponding to the target recognition category to generate the updated target recognition model feature set.

[0046] A control response parameter mapping structure is constructed synchronously, and an updated flushing command parameter set is generated based on the coupling relationship between the flushing efficiency index and the existing flushing command parameter set.

[0047] The present invention is further configured such that the closed-loop flushing control path includes:

[0048] The updated target recognition model feature set is used as the input for subsequent part recognition and judgment processes;

[0049] The updated flushing command parameter set is used as the input for the subsequent flushing control vector generation process;

[0050] A response update structure driven by flushing feedback is established. After each flushing cycle, flushing feedback indicators are continuously collected, and the target recognition model feature set and flushing command parameter set are updated sequentially to form an iterative control path oriented towards changes in flushing status.

[0051] The present invention also provides a control system for an intelligent pre-flush toilet, the system comprising:

[0052] Multi-source target detection module: Constructs a target detection mechanism based on multi-source information fusion, collects sitting posture data, forward tilt angle data and target area relative position data, and constructs a set of flushing target partition labels based on structural adjustment state variables;

[0053] Target recognition and command generation module: The execution part recognition and judgment process is carried out by matching the features of the flushing target partition label set with the forward part recognition model to filter out the current flushing target type and generate the flushing command parameter set based on structural state constraints;

[0054] Status acquisition and control vector generation module: acquires real-time water temperature, water pressure and nozzle spray status parameters, and generates multiple flushing control vectors based on the part labels and flushing intensity indicators in the flushing command parameter set;

[0055] Flushing action construction module: Combines nozzle rotation angle and water pressure dynamic adjustment coefficient, performs nonlinear transformation on multiple flushing control vectors to form a flushing action execution sequence;

[0056] Nozzle drive execution module: maps the flushing action execution sequence to nozzle drive control commands, and completes the joint control of spray angle, water pressure and flow rate and water output rhythm;

[0057] Flushing Feedback and Index Reconstruction Module: Collects displacement status feedback and flushing residue feedback after flushing is completed, and reconstructs the flushing efficiency index vector by combining the initial flushing parameters;

[0058] Model and instruction set update module: Updates the forward part identification model and flushing command parameter set according to the flushing efficiency index vector to form a closed-loop flushing control path.

[0059] This invention provides a control method and control system for an intelligent front-flush toilet. The method constructs a target detection mechanism based on multi-source information fusion, collecting sitting posture data, forward tilt angle data, and relative position data of the target area. It then constructs a set of flushing target zone labels based on structural adjustment state variables. A part identification and judgment process is executed, matching the flushing target zone label set with a forward part identification model to filter the current flushing target type. A flushing command parameter set is generated based on structural state constraints. Real-time water temperature, water pressure, and nozzle spray state parameters are acquired. Multiple flushing control vectors are generated based on the part labels and flushing intensity index in the flushing command parameter set. A nonlinear transformation is performed on the multiple flushing control vectors, combining the nozzle rotation angle and water pressure dynamic adjustment coefficient, to form a flushing action execution sequence. This flushing action execution sequence is mapped to nozzle drive control commands to achieve joint control of spray angle, water pressure flow rate, and water flow rhythm. Displacement state feedback and flushing residue feedback are collected after flushing, and the flushing efficiency index vector is reconstructed based on the initial flushing parameters. The forward part identification model and flushing command parameter set are updated according to the flushing efficiency index vector, forming a closed-loop flushing control path. The beneficial effects include:

[0060] 1. Improve the accuracy and dynamic adaptability of part recognition: By constructing a target detection mechanism based on multi-source information fusion, and introducing sitting posture data, forward tilt angle data and target area relative position information, dynamic perception and zonal recognition of the washing target are realized, which significantly improves the adaptability of the forward part recognition model to different usage postures and structural changes, and enhances the accuracy of the judgment of the washing object.

[0061] 2. Optimize the generation and execution efficiency of flushing control sequence: By combining water temperature, water pressure and nozzle status parameters, multiple flushing control vectors are dynamically generated, and nonlinear transformation is implemented using nozzle rotation angle and water pressure adjustment coefficient to form an efficient flushing action execution sequence, thereby achieving joint control of spray angle, water pressure flow rate and water output rhythm, and improving the responsiveness and smoothness of flushing operation.

[0062] 3. Implement a feedback-based closed-loop flushing control mechanism: Collect the displacement status feedback and residual feedback after flushing is completed, combine them with the initial flushing parameters to construct a flushing efficiency index vector, and use it to update the forward part identification model and flushing command parameter set to construct a closed-loop adaptive flushing control path, effectively improving the self-correction capability and overall control stability of the flushing strategy.

[0063] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0064] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings:

[0065] Figure 1 A flowchart illustrating a control method for an intelligent pre-flush toilet, as shown in an exemplary embodiment of the present invention;

[0066] Figure 2 This is a schematic diagram illustrating the structure of a control system for an intelligent pre-flush toilet, as shown in an exemplary embodiment of the present invention. Detailed Implementation

[0067] The embodiments of the present invention will be described below with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention and not for limiting the scope of protection of the present invention.

[0068] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0069] In the following description, numerous details are explored to provide a more thorough explanation of embodiments of the invention. However, it will be apparent to those skilled in the art that embodiments of the invention may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring embodiments of the invention.

[0070] Example 1:

[0071] A control method for an intelligent pre-flush toilet, such as Figure 1 As shown, it includes:

[0072] A target detection mechanism based on multi-source information fusion is constructed, collecting sitting posture data, forward tilt angle data, and relative position data of the target area, and constructing a set of flushing target partition labels based on structural adjustment state variables;

[0073] The execution part identification and judgment process involves matching the features of the flushing target partition label set with the forward part identification model to filter out the current flushing target type, and generating a flushing command parameter set based on structural state constraints.

[0074] The system acquires real-time water temperature, water pressure, and nozzle spray status parameters, and generates multiple flushing control vectors based on the location labels and flushing intensity indicators in the flushing command parameter set.

[0075] By combining the nozzle rotation angle and the water pressure dynamic adjustment coefficient, a nonlinear transformation is performed on multiple flushing control vectors to form a flushing action execution sequence.

[0076] The flushing action sequence is mapped to the nozzle drive control command to achieve joint control of the spray angle, water pressure and flow rate, and water output rhythm;

[0077] The displacement status feedback and flushing residue feedback after flushing are collected, and the flushing efficiency index vector is reconstructed by combining the initial flushing parameters.

[0078] The forward part identification model and flushing command parameter set are updated based on the flushing efficiency index vector to form a closed-loop flushing control path.

[0079] The present invention is further configured such that the construction of the target detection mechanism based on multi-source information fusion includes:

[0080] Data on the user's sitting posture is collected to form information describing the pressure distribution on the toilet's contact surface; specifically, the pressure distribution of the contact array is assumed to be a two-dimensional matrix P = [p ij ] m×n Define the sitting posture state mapping tensor as Where, p ij ∈[0,500] represents the pressure value in the i-th row and j-th column of the seat cushion, in N / m², φ i ∈R + The row weight modulation factor is used to represent the difference in longitudinal force transmission, ∈ i ∈R + To apply nonlinear power-law adjustment to regions of different heights, A∈R n To represent the combined intensity of each column in the pressure tensor for lateral symmetry inference, this tensor is used to represent the non-uniform distribution of the center of gravity of the sitting posture in a two-dimensional plane. The sensitivity to the recognition of asymmetrical sitting posture can be improved by using a nested power superposition structure.

[0081] Extract the forward tilt angle data to construct a height sequence structure reflecting the degree of upper body tilt of the user; specifically, define the height map obtained from the infrared thermal camera contour extraction as H = [h i ] 1≤i≤s , where h iThis represents the maximum upper edge height of the human body on the i-th scan line of the height map, in mm. The approximate index for the forward tilt angle is defined as follows: Where, μ i ∈R + For scanning weighting factors, used to enhance key areas, ρ∈R + γ is the exponential enhancement coefficient used to amplify forward tilt variations. i ∈R + As a offset suppression constant, it adjusts the response to small changes, δ∈R + To suppress redundancy and ensure smooth handling of excessive fluctuations, ζ is used to describe the current user's upper body forward lean. A strongly nonlinear kernel function based on the height map is designed to improve the sensitivity to dynamic sitting posture shifts.

[0082] Acquire the relative position data of the target area, and extract the relative positioning parameters of the forward region in the spatial coordinate system based on the point cloud depth information; specifically, let the spatial depth sensing point cloud matrix be D={d k ∈R 3 |k=1,…,N}, by selecting the set of points mapped to the hip and pelvic regions Define the coordinates of the center of the forward region as Where, d k Let ω be the position vector of the k-th point. k ∈R + U represents the local density weighting coefficient. k ∈R 3 This is a local normal vector used for directionality correction. To represent tensor product operations and enhance spatial coupling, this formula constructs a tensor-quantized position kernel based on the anterior pelvic region extracted from structured light point clouds, generating a spatial center point C. f This describes the precise spatial location of the target area relative to the toilet coordinate system.

[0083] The structural adjustment state quantity is constructed by fusing sitting posture data, forward tilt angle data, and target area relative position data. This quantity is then matched with predefined region mapping rules to generate a set of flushing target zone labels. Specifically, the structural adjustment state quantity is defined as a triplet. The structural adjustment state variables are mapped to the label space using a partitioning projection function. Where, λ i Let Ω(·) represent the label number of the i-th rinsing area. Let Ω(·) be the multidimensional projection mapping function. Based on the discrete index constraint and threshold condition decision generation results, a three-dimensional state set is constructed as input by fusing the sitting posture lateral tensor, forward tilt angle index and spatial region coordinates. This set is then mapped to the predefined rinsing label space to complete the rinsing area target recognition under the current user state.

[0084] The present invention is further configured such that the execution part identification and judgment process includes:

[0085] The target partition label set for rinsing is embedded to form a high-dimensional label structure feature; specifically, the target partition label set for rinsing is defined as... Define higher-order embedding functions Where, λ i To represent the index of the i-th partition in the current label space, κ j ∈R + V represents the partition importance factor, indicating label priority. j ∈R s Let T be the label embedding vector, the static representation corresponding to the j-th partition. j ∈R s×s The label structure weight tensor is used to couple the spatial correlation of intervals, E∈R s×s For the joint feature representation of the partition label set, this step combines the static spatial representation vectors corresponding to each label by constructing a high-order tensor product structure, and introduces a weight tensor T. j The contextual coupling between partitions is captured to form a unified embedding structure E, which is used to align with the feature space of the recognition model.

[0086] The label structure features are matched with the baseline feature set in the pre-defined forward part recognition model. Optimal index recognition is performed based on multi-dimensional feature differences to select the part recognition results corresponding to the current flushing target type. Specifically, the baseline feature set of the trained forward part recognition model is defined as follows: Define the location identification matching distance function as follows: Select the index corresponding to the minimum distance Among them, Υ(E,M) k ) is the structure matching deviation function, used to measure the difference between the measured structure matrix E and the k-th reference structure matrix M. k The degree of matching deviation between them, M k α is the identification template for the k-th target part. uvk ∈R + The location-part-feature modulation factor, E, represents the local weight. uv Let represent the element in the u-th row and v-th column of the structure-state matrix E. M represents the k-th reference feature. k The element in the u-th row and v-th column, β k ∈R + δ is the template stability factor used to limit anomalous alignment. k Let k be the distance evaluation value between the k-th template and the label feature tensor. *The optimal matching site label index is determined by this matching function, which explicitly considers the heteroscedasticity and modal coupling of each location in the label space using a higher-order distance metric. This avoids the use of traditional Euclidean or linear similarity functions, ensuring accurate matching of complex anatomical structures. Finally, the type index k of the current target is determined through a minimum distance strategy. * ;

[0087] A state constraint vector is constructed based on the current flushing target type and structural state variables; specifically, let the sitting posture state tensor be A, the forward tilt angle index be ζ, and the relative position vector be C. f Define the state constraint vector as Where, θ i ∈R + For state adjustment weights, γ,η∈R + χ² is a non-linear adjustment index. i ∈R + As the rinsing stability adjustment factor, Q∈R s For state constraint vectors, this state factor is used to adjust command configuration preferences under different posture conditions, construct an adaptive parameter basis for target recognition results, and provide an input structure for subsequent command parameter generation;

[0088] The state constraint vector is fused and mapped with the control template corresponding to the recognition result to generate a flushing command parameter set for flushing control path scheduling. Specifically, it is indexed k according to the target type. * State constraint vector Define the command parameter function as follows in, For the kth * The basic instruction vector corresponding to each part type. This indicates a position-synchronized convolution mapping operation. As the flushing command parameter set, which serves as the input for subsequent path control, this step fuses the identified location with the current structural state information through a synchronous mapping mechanism. The output command parameter set includes multi-dimensional flushing control parameters such as direction, intensity, and coverage, and has dynamic adaptability and correlation with physiological locations.

[0089] The present invention is further configured such that the generation of multiple flushing control vectors includes:

[0090] The system collects water temperature, water pressure, and nozzle spray status information during the current flushing cycle to construct an operational status observation vector that includes thermal state, hydraulic state, and structural attitude. Specifically, the real-time acquired status is defined as a triplet. Where, τ t ∈R + Given the current water temperature, π t ∈R + Given the current water pressure, st ∈R d The nozzle's current state vector, with dimensions including injection opening, rotation angle, and oscillation rate, is used to construct the operational state observation vector. Where σ1,σ2∈R + For temperature and pressure adjustment factors, ρ∈R + The nonlinear adjustment index is W∈R 1×d Z is the nozzle state projection matrix. t ∈R 3 This is the fused state observation vector;

[0091] Based on the flushing target type index and intensity index vector in the flushing command parameter set, the corresponding part adjustment structure is matched, and the target type adjustment tensor is constructed; specifically, the flushing intensity index vector is defined as I=[ι1,ι2,…,ι m ]∈R m , where each ι j Different control targets (such as flow intensity, angle adjustment rate, coverage area, etc.) are represented, where m is the number of control targets; a target type adjustment tensor is constructed, and its calculation formula is: in, For the part type k * Related control modulation factor, μ ij ,ν ij ∈R + ξ is a nonlinear transformation exponent. ij ∈R + As a stability constraint factor;

[0092] The operational status observation vector is fused and mapped with the target type adjustment tensor to generate a set of multiple flushing control vectors, and the set of multiple control output vectors is as follows: Where q is the number of control channels, u i ∈R represents the output value of the i-th control channel. This is the complete set of control output vectors.

[0093] The present invention is further configured such that the formation of the rinsing action execution sequence includes:

[0094] The multi-degree-of-freedom rotation angle parameters of the nozzle and the corresponding dynamic water pressure adjustment coefficient are collected, and a mapping kernel function is constructed based on the nonlinear coupling relationship between the two. Specifically, the nozzle rotation angle state vector is collected as Θ. t =[θ1,θ2,…,θ p ] □ ∈R p The dynamic adjustment coefficient vector of water pressure is P. t =[π1,π2,…,π p ] □ ∈Rp , where θ i π represents the rotation angle of the i-th nozzle's degree of freedom. i This represents the dynamic adjustment coefficient of water pressure corresponding to the i-th degree of freedom, and the mapping kernel function is constructed as follows: Where, α i ,β i ,γ∈R + δ is the nonlinear transformation exponent. i ∈R + The regularization coefficients prevent the denominator from being zero and map the kernel function. The nonlinear coupling relationship between the quantification rotation angle and the water pressure adjustment coefficient;

[0095] A nonlinear transformation is performed on multiple flushing control vectors based on a mapping kernel function to generate the transformed flushing control signal. Specifically, the calculation formula for the nonlinear transformation of multiple flushing control vectors based on a mapping kernel function is as follows: Where, η i ,ζ i ∈R + To adjust the exponent and control the magnitude and sensitivity of the change, v i ∈R represents the shuffling control signal after transformation;

[0096] The transformed control signal is dynamically updated based on the time series, forming a multi-dimensional, multi-temporal flushing action execution sequence. Specifically, the flushing action execution sequence is as follows: Wherein, the time distance from the walk length T is the execution cycle length. It updates dynamically over time to ensure consistent actions and timely responses.

[0097] The present invention is further configured such that the combined control of the injection angle, water pressure flow rate, and water outlet rhythm includes:

[0098] Multidimensional temporal feature extraction is performed on the flushing action execution sequence; specifically, the flushing action execution sequence is decomposed into multidimensional temporal features, and a feature tensor F = [f ijk ]∈R q×T×r Where r represents the number of feature extraction dimensions, f ijk =Ψ k (v i (t j ),t j ) represents the k-th characteristic function with respect to time t. j The mapping of the i-th control signal, Ψ k It is a non-linear feature extraction function, defined as follows: Where, α k ,γ k ,λk ∈R + Using parameters, the flushing action execution sequence is mapped to a high-dimensional feature space through nonlinear mapping and timing adjustment, capturing dynamic evolution laws and control characteristics, and improving the response accuracy of drive command generation;

[0099] A drive command generation function is constructed, which, combined with preset drive weight parameters, maps multidimensional features into nozzle injection angle adjustment commands, water pressure and flow rate regulation commands, and water outlet rhythm control commands; specifically, a set of nozzle drive control commands is defined. Control command d(t) at each moment j )=Φ(F :,j,: W d ), where F :,j,: ∈R q×r Represents time t j The feature matrix of all channels, W d ∈R m×q×r To drive the weight tensor, Φ denotes the tensor product mapping function, specifically: Where m = 1, 2, 3 correspond to the injection angle adjustment command, water pressure and flow rate adjustment command, and injection rhythm control command, respectively. The driving weight tensor is combined with the extracted features to achieve multi-dimensional information fusion and generate joint driving commands that meet the needs of different control dimensions.

[0100] The generated drive commands undergo nonlinear constraint mapping to ensure that the control parameters remain within a safe range. Specifically, the drive commands are mapped to ensure that the control parameters conform to physical and safety constraints, and a mapping function d is defined. * (t j )=Γ(d(t j ),C), where C=[c1,c2,c3]∈R 3 To limit the maximum amplitude of the control parameter, d * (t j ) represents the standardized driving instructions after constraint mapping, and the mapping is defined as follows: To ensure that the output control signal is bounded and continuous, a nonlinear mapping function is introduced to compress the over-limit command within the safe range, ensuring that there is no risk of overload during the drive execution process;

[0101] Based on the clock-synchronized output of standardized drive commands, the joint dynamic control of the nozzle is completed.

[0102] The present invention is further configured such that the reconstructed flushing efficiency index vector includes:

[0103] The displacement status feedback and flushing residue feedback are collected after flushing to construct a set of feedback information reflecting the nozzle operating accuracy and the degree of flushing residue. Specifically, the displacement status feedback vector D is collected through a humidity sensor and infrared scanning. f=[d1,d2,…,d p ] □ ∈R p Collect the residual feedback vector R from the rinsing process. f =[r1,r2,…,r q ] □ ∈R q , where d i r represents the displacement error or deviation feedback of the nozzle or cleaning part in the i-th dimension. j This represents the feedback signal of residual flushing material in the j-th dimension;

[0104] Extract the parameter set set during the initial flushing phase to form the initial control reference vector; specifically, extract the initial flushing parameter vector P0 = [p1, p2, ..., p m ] □ ∈R m , where p k This indicates the control parameters set when starting the flushing process, such as spray intensity, nozzle angle, and water pressure.

[0105] The feedback information set is fused with the initial control reference vector to generate a multi-dimensional structure tensor. This structure tensor is then transformed according to a preset nonlinear mapping rule to generate a flushing efficiency index vector. Specifically, the multi-dimensional structure tensor is generated, and the calculation formula is: G e =[g ijk ]∈R p×q×m The fusion function is defined as follows: Where, α ijk ,β ijk ,γ ijk ∈R + For nonlinear adjustment exponent, ∈ ijk ∈R + To avoid zero denominators, the regularization coefficient is used to generate the flushing efficiency index vector. The calculation formula is: E = [e1, e2, ..., e n ] □ ∈R n Obtained through tensor dimensionality reduction mapping Where, η ijkl ∈R represents the fusion weight coefficient, δ ijkl ∈R + is a non-linear exponent, and n represents the dimension of the efficiency index.

[0106] The present invention is further configured such that the updated forward part identification model and flushing command parameter set include:

[0107] Extract the multidimensional feedback quantity from the flushing efficiency index vector, construct a mapping response structure, and perform nonlinear feature adjustment on the baseline feature set of each target part in the forward part recognition model according to the feature matrix corresponding to the target recognition category, generating an updated target recognition model feature set; specifically, define the nonlinear transformation coefficient matrix: Ξ=[ξ ij ]∈R n×d , Among them, e i Let α be the flushing efficiency index vector. ij ,β ij ∈R + ε is the transformation exponent. ij ∈R + This is a regularization term to suppress weight imbalance, where d is the total number of part categories in the forward model, corresponding to... The feature set of the trained forward recognition model is: Constructing a response update tensor: The update rules are as follows: in, In response to the enhancement factor, F i ∈R s×1 Let G be the feedback feature vector of the i-th indicator. k ∈R s×1 Let γ be the direction vector for class k, where γ∈R + In response to the amplification factor;

[0108] A control response parameter mapping structure is constructed synchronously. Based on the coupling relationship between the flushing efficiency index and the existing flushing command parameter set, an updated flushing command parameter set is generated. Specifically, let the flushing command parameter set used in the previous round of control be: The control update matrix is: K p ∈R n×z , The update rule is as follows Where, μ ij ,ν ij ,η ij ∈R + To control the response conditioning exponent, λ∈R + Here, E represents the nonlinear gain control parameter, and E is the flushing efficiency index vector.

[0109] The present invention is further configured such that the closed-loop flushing control path includes:

[0110] The updated target recognition model feature set is used as the input for subsequent part recognition and judgment processes;

[0111] The updated flushing command parameter set is used as the input for the subsequent flushing control vector generation process;

[0112] A response update structure driven by flushing feedback is established. After each flushing cycle, flushing feedback indicators are continuously collected, and the target recognition model feature set and flushing command parameter set are updated sequentially to form an iterative control path oriented towards changes in flushing status. Specifically, the updated target recognition feature set... Input to the next round of target discrimination process, flush command parameter set Inputting into step two enables feedback-driven iterative closed-loop structural adjustment, forming a high-frequency adaptive path: →Next round of identification and control.

[0113] Example 2:

[0114] Please see Figure 2 The exemplary control system for a smart pre-flush toilet includes:

[0115] Multi-source target detection module: Constructs a target detection mechanism based on multi-source information fusion, collects sitting posture data, forward tilt angle data and target area relative position data, and constructs a set of flushing target partition labels based on structural adjustment state variables;

[0116] Target recognition and command generation module: The execution part recognition and judgment process is carried out by matching the features of the flushing target partition label set with the forward part recognition model to filter out the current flushing target type and generate the flushing command parameter set based on structural state constraints;

[0117] Status acquisition and control vector generation module: acquires real-time water temperature, water pressure and nozzle spray status parameters, and generates multiple flushing control vectors based on the part labels and flushing intensity indicators in the flushing command parameter set;

[0118] Flushing action construction module: Combines nozzle rotation angle and water pressure dynamic adjustment coefficient, performs nonlinear transformation on multiple flushing control vectors to form a flushing action execution sequence;

[0119] Nozzle drive execution module: maps the flushing action execution sequence to nozzle drive control commands, and completes the joint control of spray angle, water pressure and flow rate and water output rhythm;

[0120] Flushing Feedback and Index Reconstruction Module: Collects displacement status feedback and flushing residue feedback after flushing is completed, and reconstructs the flushing efficiency index vector by combining the initial flushing parameters;

[0121] Model and instruction set update module: Updates the forward part identification model and flushing command parameter set according to the flushing efficiency index vector to form a closed-loop flushing control path.

[0122] It should be noted that the control system for an intelligent front-flush toilet provided in the above embodiments and the control method for an intelligent front-flush toilet provided in the above embodiments belong to the same concept. The specific ways in which each module and unit performs its operation have been described in detail in the method embodiments, and will not be repeated here. In practical applications, the control system for an intelligent front-flush toilet provided in the above embodiments can be assigned to different functional modules as needed, that is, the internal structure of the system can be divided into different functional modules to complete all or part of the functions described above, and this is not a limitation here.

[0123] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A control method for an intelligent pre-flush toilet, characterized in that, include: A target detection mechanism based on multi-source information fusion is constructed, collecting sitting posture data, forward tilt angle data, and relative position data of the target area, and constructing a set of flushing target partition labels based on structural adjustment state variables; The execution part identification and judgment process involves matching the features of the flushing target partition label set with the forward part identification model to filter out the current flushing target type, and generating a flushing command parameter set based on structural state constraints. The system acquires real-time water temperature, water pressure, and nozzle spray status parameters, and generates multiple flushing control vectors based on the location labels and flushing intensity indicators in the flushing command parameter set. By combining the nozzle rotation angle and the water pressure dynamic adjustment coefficient, a nonlinear transformation is performed on multiple flushing control vectors to form a flushing action execution sequence. The flushing action sequence is mapped to the nozzle drive control command to achieve joint control of the spray angle, water pressure and flow rate, and water output rhythm; The displacement status feedback and flushing residue feedback after flushing are collected, and the flushing efficiency index vector is reconstructed by combining the initial flushing parameters. The forward part identification model and flushing command parameter set are updated based on the flushing efficiency index vector to form a closed-loop flushing control path.

2. The control method for an intelligent pre-flush toilet according to claim 1, characterized in that, Constructing a target detection mechanism based on multi-source information fusion includes: Collect sitting posture data to generate information describing the pressure distribution of the user on the toilet surface; Extract forward tilt angle data and construct a height sequence structure that reflects the degree of upper body tilt of the user; Acquire the relative position data of the target area, and extract the relative positioning parameters of the forward region in the spatial coordinate system based on the point cloud depth information; The sitting posture data, forward tilt angle data, and target area relative position data are fused to construct a structural adjustment state quantity. By matching it with predefined area mapping rules, a set of flushing target partition labels is generated.

3. The control method for an intelligent pre-flush toilet according to claim 1, characterized in that, The execution site identification and judgment process includes: The target partition label set for rinsing is embedded to form a high-dimensional label structure feature; The label structure features are matched with the baseline feature set in the preset forward part recognition model. The optimal index recognition is performed based on the multidimensional feature differences to select the part recognition results corresponding to the current flushing target type. Construct a state constraint vector based on the current flushing target type and structural state variables; The state constraint vector is fused and mapped with the control template corresponding to the recognition result to generate a set of flushing command parameters for flushing control path scheduling.

4. The control method for an intelligent pre-flush toilet according to claim 3, characterized in that, The generation of multiple flush control vectors includes: Collect water temperature signals, water pressure signals, and nozzle spray status information during the current flushing cycle to construct an operational status observation vector that includes thermal state, hydraulic state, and structural attitude; Based on the flushing target type index and intensity index vector in the flushing command parameter set, the corresponding part adjustment structure is matched, and the target type adjustment tensor is constructed. The running status observation vector and the target type adjustment tensor are fused and mapped to generate a set of multiple flushing control vectors.

5. The control method for an intelligent pre-flush toilet according to claim 1, characterized in that, The sequence of flushing actions includes: Collect the multi-degree-of-freedom rotation angle parameters of the nozzle and the corresponding dynamic water pressure adjustment coefficient, and construct a mapping kernel function based on the nonlinear coupling relationship between the two; A nonlinear transformation is performed on multiple flushing control vectors based on a mapping kernel function to generate the transformed flushing control signal. The transformed control signals are dynamically updated based on the time series to form a multi-dimensional, multi-temporal flushing action execution sequence.

6. The control method for an intelligent pre-flush toilet according to claim 5, characterized in that, The combined control of jet angle, water pressure, flow rate, and water output rhythm includes: Multidimensional temporal feature extraction is performed on the flushing action execution sequence; Construct a drive command generation function, and combine it with preset drive weight parameters to map multidimensional features into nozzle injection angle adjustment commands, water pressure and flow rate adjustment commands, and water output rhythm control commands; Perform nonlinear constraint mapping on the generated drive commands to ensure that the control parameters are within a safe range; Based on the clock-synchronized output of standardized drive commands, the joint dynamic control of the nozzle is completed.

7. The control method for an intelligent pre-flush toilet according to claim 1, characterized in that, The reconstructed flushing efficiency index vector includes: Collect the displacement status feedback and flushing residue feedback after the flushing is completed, and construct a set of feedback information reflecting the nozzle operating accuracy and the degree of flushing residue. Extract the set of parameters set during the initial flushing phase to form the initial control reference vector; The feedback information set is fused with the initial control reference vector to generate a multi-dimensional structural tensor. The structural tensor is then transformed according to a preset nonlinear mapping rule to generate a flushing efficiency index vector.

8. The control method for an intelligent pre-flush toilet according to claim 7, characterized in that, The updated forward part recognition model and flushing command parameter set include: Extract the multidimensional feedback quantity from the flushing efficiency index vector, construct the mapping response structure, and perform nonlinear feature adjustment on the baseline feature set of each target part in the forward part recognition model according to the feature matrix corresponding to the target recognition category to generate the updated target recognition model feature set. A control response parameter mapping structure is constructed synchronously, and an updated flushing command parameter set is generated based on the coupling relationship between the flushing efficiency index and the existing flushing command parameter set.

9. The control method for an intelligent pre-flush toilet according to claim 8, characterized in that, The closed-loop flushing control path includes: The updated target recognition model feature set is used as the input for subsequent part recognition and judgment processes; The updated flushing command parameter set is used as the input for the subsequent flushing control vector generation process; A response update structure driven by flushing feedback is established. After each flushing cycle, flushing feedback indicators are continuously collected, and the target recognition model feature set and flushing command parameter set are updated sequentially to form an iterative control path oriented towards changes in flushing status.

10. A control system for an intelligent pre-flush toilet, used to implement the control method for an intelligent pre-flush toilet as described in any one of claims 1-9, characterized in that, include: Multi-source target detection module: Constructs a target detection mechanism based on multi-source information fusion, collects sitting posture data, forward tilt angle data and target area relative position data, and constructs a set of flushing target partition labels based on structural adjustment state variables; Target recognition and command generation module: The execution part recognition and judgment process is carried out by matching the features of the flushing target partition label set with the forward part recognition model to filter out the current flushing target type and generate the flushing command parameter set based on structural state constraints; Status acquisition and control vector generation module: acquires real-time water temperature, water pressure and nozzle spray status parameters, and generates multiple flushing control vectors based on the part labels and flushing intensity indicators in the flushing command parameter set; Flushing action construction module: Combines nozzle rotation angle and water pressure dynamic adjustment coefficient, performs nonlinear transformation on multiple flushing control vectors to form a flushing action execution sequence; Nozzle drive execution module: maps the flushing action execution sequence to nozzle drive control commands, and completes the joint control of spray angle, water pressure and flow rate and water output rhythm; Flushing Feedback and Index Reconstruction Module: Collects displacement status feedback and flushing residue feedback after flushing is completed, and reconstructs the flushing efficiency index vector by combining the initial flushing parameters; Model and instruction set update module: Updates the forward part identification model and flushing command parameter set according to the flushing efficiency index vector to form a closed-loop flushing control path.

Images