Printing handheld terminal device interaction method and system based on Internet of Things

Abstract

The invention discloses a printing handheld terminal device interaction method and system based on the Internet of Things, and the method comprises the steps: building communication connection, collecting operation events and physical property data, and forming an event sequence; the event and physical property data are uploaded and preprocessed, and a reversible interaction power sub-graph is constructed; extracting multiple types of physical property features, coding and fusing the multiple types of physical property features, and generating a multi-modal interaction physical property vector; constructing an improved GraphSAGE network, and executing heterogeneous collaborative convergence to generate a strategy vector; generating a strategy capsule and performing fission and recombination to form an execution sequence to drive printing operation; and uploading printing feedback data, updating the reversible interaction subgraph and improving the GraphSAGE network. According to the method, intelligent generation and dynamic adaptive optimization of an interaction process of the printing handheld terminal are realized by constructing a reversible interaction power sub-graph and fusing an improved GraphSAGE network of physical property characteristics and a strategy capsule fission mechanism.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of Internet of Things intelligent terminal, and particularly relates to a printing handheld terminal device interaction method and system based on Internet of Things. BACKGROUND

[0002] Under the background of rapid development of Internet of Things technology, printing handheld terminal devices are widely used in logistics distribution, warehouse management, on-site bill printing and asset identification, etc. The existing printing handheld terminal devices usually rely on fixed interface processes and preset interaction logic to complete task execution, and their interaction modes are mostly written in firmware, lacking the ability to adjust in real time based on user behavior changes, device state changes and environmental changes. With the increasing complexity of terminal application scenarios, the user operation path presents diversity, the task structure presents dynamics, and the device state presents instability, so the traditional static interaction mode cannot meet the needs of high-frequency on-site operations.

[0003] Although the existing Internet of Things printing terminal has basic data reporting and task issuing capabilities, it can only upload operation logs and device states, and cannot structurally understand the interface switching relationship and event triggering relationship in the interaction process, nor can it model user operation habits, resulting in the terminal being unable to adaptively adjust the interaction process according to different user operation characteristics. The existing technology has a low degree of utilization of the physical property data collected by the terminal, and most of it is only used for display or alarm, without forming an effective mechanism combined with interaction strategies, making it difficult to dynamically select a more reasonable interaction path under complex task conditions.

[0004] Once the current printing handheld terminal interaction process is fixed on the device side, subsequent optimization of interface logic, addition of verification steps or adjustment of exception handling processes must be realized through firmware upgrade, which makes the interaction logic adjustment cycle long, the operation and maintenance cost high, and the device collaboration efficiency low. The existing technology also lacks the ability to reversely deduce the interaction steps, cannot reason about historical interaction paths to optimize future interactions, and also lacks a mechanism to dynamically reorganize the interaction process locally on the terminal. The existing technology has obvious deficiencies in interaction understanding, strategy adaptability and process reconfigurability.

[0005] Therefore, how to provide a printing handheld terminal device interaction method and system based on Internet of Things is a problem that needs to be solved by those skilled in the art. SUMMARY

[0006] One purpose of the present application is to propose an Internet of Things-based printing handheld terminal device interaction method and system, which fully utilizes reversible interaction dynamic subgraph modeling, improved GraphSAGE network and strategy capsule dynamic restructuring technology, and structurally expresses, strategically generates and process-level adaptively controls the interaction process of the printing handheld terminal.

[0007] According to an embodiment of the present application, an Internet of Things-based printing handheld terminal device interaction method comprises:

[0008] The printing handheld terminal device establishes a communication connection with the Internet of Things platform, collects user interface operation events and multi-modal physical property data, forms an interaction event sequence and a physical property data sequence, and uploads the interaction event sequence and the physical property data sequence to the Internet of Things platform.

[0009] The interaction event sequence is time-aligned and type-normalized, and the interaction nodes and edges are constructed to form a reversible interaction dynamic subgraph.

[0010] Based on the physical property data sequence, time physical property features, behavior physical property features, device load physical property features and environmental physical property features are extracted, and the various physical property features are normalized, encoded and fused to generate a multi-modal fusion interaction physical property vector.

[0011] An improved GraphSAGE network is constructed, the local subgraph features of the reversible interaction dynamic subgraph are input into the graph structure channel, the multi-modal fusion interaction physical property vector is input into the physical property channel, and heterogeneous collaborative convergence is performed to generate a strategy vector.

[0012] According to the strategy vector, a plurality of strategy capsules are generated, and the strategy capsules are identified by version and then distributed to the printing handheld terminal device through the Internet of Things platform, and the strategy capsules are fissioned to generate a strategy capsule execution sequence corresponding to the current task to drive the printing module to perform a printing operation.

[0013] The printing handheld terminal device uploads the printing quality label and the printing session feedback data to the Internet of Things platform, updates the dynamics parameters of the reversible interaction dynamic subgraph and updates the improved GraphSAGE network.

[0014] Optionally, the user interface operation event includes an interface button click event, a text input event, an interface jump event, an interface stay event, a cancel operation event, a return operation event, a long press operation event, a sliding operation event, and a print instruction trigger event.

[0015] Optionally, the multi-modal physical property data includes device power data, device voltage data, print head temperature data, consumable remaining amount data, signal strength data, environment temperature data, environment humidity data, environment illumination data, device posture data, and device motion state data.

[0016] Optionally, the forming of the reversible interaction force subgraph with forward deduction and reverse deduction capabilities includes:

[0017] Receiving the interaction event sequence and the physical property data sequence, sorting and deduplicating the interaction events according to timestamps, eliminating incomplete events and duplicate reporting events, and generating a standard interaction event sequence arranged in chronological order;

[0018] According to the interface identifier, event type, task stage, abnormality marker, and print result marker corresponding to each interaction event, type-normalizing the interaction events, and respectively assigning a forward state marker and a backtracking state marker to each interaction event;

[0019] Merging events with the same interface identifier, event type combination, and the same forward state marker and backtracking state marker into the same interaction node, recording the corresponding entry condition, exit condition, associated task stage, and the number of successful events and failed events in each interaction node, and forming an interaction node set with forward state information and backtracking state information;

[0020] According to the time sequence relationship and interface switching relationship between adjacent events in the standard interaction event sequence, establishing a directed interaction edge between the interaction node sets, recording the trigger condition, associated successful result marker, and failure result marker for each interaction edge, assigning a backtracking trigger marker according to the failure event proportion and abnormal event proportion, marking the interaction edges with high failure proportion and high abnormal proportion as reverse priority backtracking edges, and forming an interaction edge set containing forward transition edges and reverse backtracking edges;

[0021] Selecting a node subset and an edge subset related to the print task from the interaction node set and the interaction edge set, respectively establishing a forward adjacent node set and a backtracking adjacent node set for each interaction node based on the forward state marker and the backtracking state marker of each node and the backtracking trigger marker on the interaction edge, and combining the node subset, the edge subset, and the forward adjacency relationship and the backtracking adjacency relationship to form a reversible interaction force subgraph.

[0022] Optionally, the generating of the multi-modal fusion interaction physical property vector includes:

[0023] The physical property data sequence is aligned according to the time stamp, and the multi-modal physical property data corresponding to each interaction event is divided into a time physical property sub-sequence, a behavior physical property sub-sequence, a device load physical property sub-sequence and an environment physical property sub-sequence, and a time index relationship corresponding to the interaction event sequence is established for each sub-sequence;

[0024] In the time physical property sub-sequence, the physical property data in each window is statistically processed according to a preset time window, and time physical property features including stay duration, input duration, event trigger frequency in the window are extracted, and the time physical property features are normalized to obtain a standardized time physical property feature set;

[0025] In the behavior physical property sub-sequence, the number of false touches, the number of repeated operations, the number of cancel operations, the number of return operations and the operation sequence stability are counted according to the user interface operation event and the corresponding physical property data record, and a behavior physical property feature set is formed, and the behavior physical property features are encoded and normalized to obtain a standardized behavior physical property feature set;

[0026] In the device load physical property sub-sequence and the environment physical property sub-sequence, device load physical property features and environment physical property features are extracted based on device power change, print head temperature change, consumable remaining amount change, and environment temperature change, environment humidity change, and environment illumination change, and the device load physical property features and the environment physical property features are subjected to outlier correction and dimensionless processing to form a standardized device load physical property feature set and a standardized environment physical property feature set;

[0027] The standardized time physical property feature set, the standardized behavior physical property feature set, the standardized device load physical property feature set and the standardized environment physical property feature set are combined according to a preset feature splicing order, and the combined feature set is mapped to a fixed-length feature representation through a unified encoding method to form a multi-modal fusion interaction physical property vector.

[0028] Optionally, the generation strategy vector includes:

[0029] An overall structure of an improved GraphSAGE network is constructed, the improved GraphSAGE network including an input layer, a graph structure channel, a physical property channel, a heterogeneous collaborative convergence layer and a strategy vector output layer, the input layer receiving local subgraph features of reversible interaction dynamic subgraphs and multi-modal fusion interaction physical property vectors;

[0030] In the graph structure channel, a subset of interaction nodes and a subset of interaction edges related to the target printing task in the reversible interaction dynamic subgraph are input as a local subgraph, an initial feature vector of each interaction node in the local subgraph is assigned, an adjacency relationship description is established based on the directed connection relationship between the interaction nodes, the node features of the local subgraph are updated layer by layer using a multi-layer graph aggregation structure, and a graph feature representation of the graph structure channel is obtained;

[0031] In the physical property channel, the multi-modal fusion interaction physical property vector is divided into multiple physical property sub-vectors according to time physical property, behavior physical property, device load physical property and environment physical property, a physical property confidence coefficient is generated based on the missing condition, fluctuation amplitude and historical stability of each physical property sub-vector, and a physical property feature representation is obtained by using a gate re-weighting structure driven by the physical property confidence coefficient to screen and combine the physical property sub-vectors;

[0032] In the heterogeneous collaborative convergence layer, the graph feature representation and the physical property feature representation are aligned and combined according to a preset fusion order, and a heterogeneous collaborative convergence operation is performed on the combined heterogeneous features, the heterogeneous collaborative convergence operation including joint weighting, nonlinear transformation and inter-layer transmission of the graph features and the physical property features, to generate a fusion feature representation;

[0033] The fusion feature representation is input into a policy vector output layer to generate a policy vector through linear mapping, the policy vector describing an interaction policy configuration corresponding to the current printing task.

[0034] Optionally, the generating of the policy capsule execution sequence corresponding to the current task and the driving of the printing module to perform the printing operation include:

[0035] The interaction policy configuration parameters are parsed from the policy vector output by the improved GraphSAGE network, the interaction policy configuration parameters including a target interaction node set, an execution priority, a mandatory node label, a skippable node label, an exception handling label and an energy weight parameter;

[0036] A corresponding policy capsule is generated for each interaction node in the target interaction node set, and the interface template information, the trigger condition set, the skippable output relationship set, the execution parameter set and the energy weight parameter are encapsulated in the same policy capsule to form a multi-capsule set to be issued, which is associated with a current task identifier, a device type identifier and a policy version number and then issued to a printing handheld terminal device through an Internet of Things platform;

[0037] After receiving the multi-capsule set, the printing handheld terminal device performs adaptive screening according to the current task type, the real-time state of the device and the historical execution context, and removes the policy capsules that do not match the task constraints or conflict with the device state to obtain an initial candidate policy capsule set.

[0038] Perform fission processing on the composite strategy capsule in the initial candidate strategy capsule set:

[0039] Split into the minimum functional unit of interface display, field verification, exception handling, user confirmation and security check, and split different condition branches into independent sub-capsules with the trigger condition as the boundary;

[0040] When the energy weight parameter exceeds the preset threshold or the device state is in a limited state, split the high-cost step and the key check step into independent sub-capsules and move to the priority position of the execution sequence;

[0041] The sub-capsules obtained by splitting are sequentially maintained and conflict is resolved according to the dependency relationship;

[0042] Perform consistency check on the fission result to ensure consistent association of sub-capsules in version number, task identification and trigger condition, forming a candidate sub-capsule set;

[0043] According to the energy weight parameter, execution priority and trigger condition in the candidate sub-capsule set, construct a strategy capsule execution sequence that meets the task execution sequence constraint, map to the interface display step, field verification step, exception prompt step and user confirmation step, complete the dynamic interaction process of the current printing task, and drive the printing module to perform printing operation.

[0044] Optionally, the dynamics parameters of the reversible interactive dynamic subgraph are updated and the GraphSAGE network is improved, including:

[0045] After completing the current printing task, the printing handheld terminal device packs the printing quality label corresponding to this interaction process, the printing result mark, the interaction event summary, the strategy capsule execution sequence, the exception record and the time consumption statistics into the printing session feedback data;

[0046] The Internet of Things platform analyzes and aligns the printing session feedback data, establishes a one-to-one correspondence between the strategy version number, the device type identification, the task identification and the strategy capsule execution sequence, associates and stores the interaction event summary and the printing quality label according to the session, and generates a standardized feedback record;

[0047] The Internet of Things platform updates the reversible interactive dynamic subgraph based on the standardized feedback record, recalculates the state statistics of the interaction node according to the number of entries, the number of successes and the number of failures of the interaction node, recalculates the transfer statistics of the interaction edge according to the number of triggers, the number of successful transfers and the number of failed transfers of the interaction edge, updates the reverse backtracking mark and the backtracking trigger threshold according to the exception record, and generates an updated version of the reversible interactive dynamic subgraph and saves the version number;

[0048] The Internet of Things platform updates parameters of the improved GraphSAGE network based on standardized feedback records, jointly updates parameters of the graph structure channel and the physical property channel by using a training configuration corresponding to a policy version number, generates a new policy version, and delivers the new policy version to the printing handheld terminal device.

[0049] The printing handheld terminal device interaction system based on the Internet of Things according to the embodiment of the application comprises the following modules:

[0050] The data acquisition module is configured to acquire user interface operation events and multi-modal physical property data, and form an interaction event sequence and a physical property data sequence.

[0051] The event processing and subgraph construction module is configured to perform time alignment and type normalization on the interaction event sequence, and generate a reversible interaction dynamic subgraph.

[0052] The physical property feature fusion module is configured to extract time physical property features, behavior physical property features, device load physical property features, and environment physical property features from the physical property data sequence, and generate a multi-modal fusion interaction physical property vector.

[0053] The strategy generation module is configured to construct an improved GraphSAGE network, perform heterogeneous collaborative convergence, and generate a strategy vector.

[0054] The strategy capsule execution module is configured to generate a strategy capsule according to the strategy vector and deliver the strategy capsule to the terminal, perform fission processing on the strategy capsule, and form a strategy capsule execution sequence.

[0055] The feedback update module is configured to receive printing quality labels and printing session feedback data, update dynamics parameters of the reversible interaction dynamic subgraph, and update the improved GraphSAGE network.

[0056] The printing handheld terminal device interaction system based on the Internet of Things according to the embodiment of the application has the following beneficial effects:

[0057] The printing handheld terminal device interaction system based on the Internet of Things according to the embodiment of the application has the following beneficial effects:

[0058] The improved GraphSAGE network is adopted to construct a double-channel strategy generation structure, which respectively processes interactive dynamic subgraph features and multi-modal fusion interactive physical property vectors, and forms a strategy vector highly sensitive to current tasks, device states and user behaviors through a heterogeneous collaborative convergence mechanism. The interactive process can be automatically adjusted according to different use scenarios. The structure overcomes the defects that the existing printing handheld terminal relies on the solidification interface process, cannot adapt to different task complexity and device state fluctuations, and makes the interactive strategy no longer rely on fixed logic, but can be intelligently generated according to real-time state, improves the interactive efficiency and reduces the error rate.

[0059] The application further introduces a strategy capsule and a fission execution mechanism, which disassembles the interactive process into independent strategy units that can be configured and combined, so that the printing handheld terminal can complete the dynamic recombination of the interactive process without firmware upgrade. User interface adjustment, field verification adjustment and exception handling adjustment can be realized through the strategy capsule and fission execution, which greatly reduces the system operation and maintenance cost and improves the adaptive ability and collaborative performance of the terminal. Through the feedback data of the printing session, the reversible interactive dynamic subgraph and the improved GraphSAGE network are continuously updated, and the application realizes the continuous evolution of the interactive strategy, so that the system maintains efficient, stable and intelligent interactive experience in long-term use. BRIEF DESCRIPTION OF DRAWINGS

[0060] The accompanying drawings are included to provide a further understanding of the application, and constitute a part of the specification, which together with the embodiments of the application, are used to explain the application, and do not constitute a limitation on the application. In the drawings:

[0061] Figure 1 A flowchart of a printing handheld terminal device interaction method based on the Internet of Things is provided for the application;

[0062] Figure 2 A structure schematic diagram of a printing handheld terminal device interaction system based on the Internet of Things is provided for the application. DETAILED DESCRIPTION

[0063] The application will now be further described in detail in conjunction with the drawings. These drawings are simplified schematic diagrams, and only illustrate the basic structure of the application in a schematic manner, and therefore only show the components related to the application.

[0064] REFERENCE Figure 1 A printing handheld terminal device interaction method based on the Internet of Things, comprising:

[0065] The printing handheld terminal device and the Internet of Things platform establish a communication connection, collect user interface operation events and multi-modal physical property data, form an interactive event sequence and physical property data sequence;

[0066] upload the interaction event sequence and the physical property data sequence to the Internet of Things platform, perform time alignment and type normalization processing on the interaction event sequence, construct interaction nodes and interaction edges, and form a reversible interaction dynamic subgraph;

[0067] Based on the physical property data sequence, time physical property features, behavior physical property features, device load physical property features, and environment physical property features are extracted, normalized, encoded, and fused to generate a multi-modal fusion interaction physical property vector;

[0068] An improved GraphSAGE network is constructed, the local subgraph features of the reversible interaction dynamic subgraph are input into the graph structure channel, and the multi-modal fusion interaction physical property vector is input into the physical property channel to perform heterogeneous collaborative convergence and generate a strategy vector;

[0069] According to the strategy vector, a plurality of strategy capsules are generated, the strategy capsules are identified by version and then distributed to the printing handheld terminal device through the Internet of Things platform, the strategy capsules are processed by fission to generate a strategy capsule execution sequence corresponding to the current task, and the printing module is driven to perform a printing operation;

[0070] The printing handheld terminal device uploads the printing quality label and the printing session feedback data to the Internet of Things platform, updates the dynamics parameters of the reversible interaction dynamic subgraph, and updates the improved GraphSAGE network.

[0071] In this embodiment, the user interface operation events include interface button click events, text input events, interface jump events, interface stay events, cancel operation events, return operation events, long press operation events, sliding operation events, and printing instruction triggering events.

[0072] In this embodiment, the multi-modal physical property data includes device power data, device voltage data, print head temperature data, consumable remaining amount data, signal strength data, environment temperature data, environment humidity data, environment illumination data, device posture data, and device motion state data.

[0073] In this embodiment, the reversible interaction dynamic subgraph with forward deduction and reverse deduction capabilities is formed, including:

[0074] The interaction event sequence and the physical property data sequence are received, the interaction events are sorted and deduplicated according to timestamps, incomplete events and duplicate reported events are removed, and a standard interaction event sequence arranged in chronological order is generated;

[0075] According to the interface identifier, event type, task stage, abnormality marker, and printing result marker corresponding to each interaction event, the interaction event is type-normalized, and a forward state marker and a backtracking state marker are respectively assigned to each interaction event;

[0076] Events with the same interface identifier, event type combination, and the same forward state marker and backtracking state marker are merged into the same interaction node, and the corresponding entry condition, exit condition, associated task phase, and the number of successful events and failed events in the node are recorded for each interaction node to form an interaction node set with forward state information and backtracking state information;

[0077] According to the time sequence relationship and interface switching relationship between adjacent events in the standard interaction event sequence, a directed interaction edge is established between the interaction node set, and the trigger condition, associated success result marker and failure result marker are recorded for each interaction edge. According to the failure event proportion and the abnormal event proportion, the backtracking trigger marker is allocated, and the interaction edge with high failure proportion and high abnormal proportion is marked as a reverse priority backtracking edge to form an interaction edge set containing forward transition edges and reverse backtracking edges.

[0078] In the interaction node set and the interaction edge set, a node subset and an edge subset related to the print task are selected, and based on the forward state marker and the backtracking state marker of each node and the backtracking trigger marker on the interaction edge, a forward adjacent node set and a backtracking adjacent node set are established for each interaction node. The node subset, the edge subset, and the forward adjacent relationship and the backtracking adjacent relationship are combined to form a reversible interaction dynamic subgraph.

[0079] In the embodiment, the generation of the multi-modal fusion interaction physical property vector comprises:

[0080] The physical property data sequence is aligned according to the timestamp, and the multi-modal physical property data corresponding to each interaction event is divided into a time physical property sub-sequence, a behavior physical property sub-sequence, a device load physical property sub-sequence, and an environment physical property sub-sequence. A time index relationship corresponding to the interaction event sequence is established for each sub-sequence;

[0081] In the time physical property sub-sequence, the physical property data in each window is statistically processed according to a preset time window, and time physical property features including the stay duration, the input duration, and the event trigger frequency in the window are extracted. The time physical property features are normalized to obtain a standardized time physical property feature set;

[0082] In the behavior physical property sub-sequence, the number of false touches, the number of repeated operations, the number of canceled operations, the number of return operations, and the operation sequence stability are counted according to the user interface operation event and the corresponding physical property data record to form a behavior physical property feature set. The behavior physical property features are encoded and normalized to obtain a standardized behavior physical property feature set;

[0083] In the device load physical property sub-sequence and the environment physical property sub-sequence, device load physical property features and environment physical property features are extracted based on device power changes, print head temperature changes, consumable remaining amount changes, and environment temperature changes, environment humidity changes, and environment illumination changes, respectively, and the device load physical property features and the environment physical property features are subjected to outlier correction and dimension unification processing to form a standardized device load physical property feature set and a standardized environment physical property feature set, and the device load physical property features and the environment physical property features are subjected to outlier correction and dimension unification processing, specifically as follows:

[0084] Mutant outliers in the device load physical property features and the environment physical property features are detected, and outlier records exceeding the upper or lower limit of the feature mean value range are replaced with smooth interpolation values of adjacent time step features;

[0085] Physical property features of different dimensions are respectively mapped to a unified numerical interval according to feature categories, and device power, temperature, humidity, and illumination features are made comparable in the same feature space through maximum-minimum interval transformation or interval linear scaling;

[0086] After interval unification, each type of physical property feature is subjected to feature smoothing processing in time sequence order, and features with a continuous fluctuation amplitude exceeding a preset threshold are smoothed using a sliding window mean value to maintain stability of the feature sequence;

[0087] The standardized time physical property feature set, the standardized behavior physical property feature set, the standardized device load physical property feature set, and the standardized environment physical property feature set are combined according to a preset feature splicing order, and the combined feature set is mapped to a fixed-length feature representation through a unified coding method to form a multi-modal fusion interaction physical property vector, and the preset feature splicing order is time physical property features first, behavior physical property features in the middle, device load physical property features second, and environment physical property features last.

[0088] In this embodiment, the generation strategy vector includes:

[0089] The overall structure of the improved GraphSAGE network is constructed, which includes an input layer, a graph structure channel, a physical property channel, a heterogeneous collaborative convergence layer, and a strategy vector output layer. The input layer receives local subgraph features of reversible interactive dynamic subgraphs and multi-modal fusion interaction physical property vectors.

[0090] In the graph structure channel, an interactive node subset and an interactive edge subset related to the target print task in the reversible interactive dynamic subgraph are taken as local subgraph inputs, an initial feature vector is assigned to each interactive node in the local subgraph, an adjacency relationship description is established based on the directed connection relationship between the interactive nodes, a multi-layer graph aggregation structure is used to update the node features of the local subgraph layer by layer to obtain a graph feature representation of the graph structure channel.

[0091] In the physical property channel, the multi-modal fusion interaction physical property vector is divided into multiple physical property sub-vectors according to time physical property, behavior physical property, device load physical property and environment physical property, a physical property confidence coefficient is generated based on the missing condition, fluctuation amplitude and historical stability of each physical property sub-vector, a gated re-weighting structure driven by the physical property confidence coefficient is used to screen and combine the physical property sub-vectors, and a physical property feature representation is obtained, wherein:

[0092] The physical property confidence coefficient is generated based on the missing condition, fluctuation amplitude and historical stability of each physical property sub-vector, and specifically:

[0093] The basic confidence value is determined according to the missing proportion of the physical property sub-vector in the recent time window, and the basic confidence value of the sub-vector is reduced when the missing proportion increases;

[0094] On the basis of the basic confidence value, the fluctuation amplitude of the physical property sub-vector in the current time window is combined, and the corresponding confidence value is adjusted downward when the fluctuation amplitude exceeds a preset threshold;

[0095] The confidence value is compensated according to the long-term historical stability of the physical property sub-vector, the compensation weight is increased when the feature is stable in the history of multiple tasks, and finally the basic confidence value, the fluctuation correction value and the stability compensation value are synthesized to obtain the physical property confidence coefficient;

[0096] The physical property sub-vectors are screened and combined by using the gated re-weighting structure driven by the physical property confidence coefficient, and specifically:

[0097] The physical property confidence coefficient is input as a gating weight to a gating unit, and the activation degree of each physical property sub-vector is adjusted, so that the high-confidence sub-vector obtains a higher activation weight;

[0098] The physical property sub-vectors after gating are subjected to re-weighting processing, and the confidence coefficient is used as a feature weighting factor to reduce the influence degree of the low-confidence sub-vector and increase the contribution degree of the high-confidence sub-vector;

[0099] The physical property sub-vectors after re-weighting are combined according to a preset splicing order to form a physical property feature representation with differentiated feature contribution degrees;

[0100] In the heterogeneous collaborative fusion layer, the graph feature representation and the physical property feature representation are aligned and combined according to a preset fusion order, a heterogeneous collaborative fusion operation is performed on the combined heterogeneous features, the heterogeneous collaborative fusion operation includes joint weighting, nonlinear transformation and inter-layer transmission of the graph features and the physical property features, and a fused feature representation is generated, wherein the preset fusion order is to give priority to aligning the graph feature representation output by the graph structure channel, to introduce the physical property feature representation output by the physical property channel, to place the graph features as a dominant feature sequence in the front part, and to place the physical property features as an auxiliary feature sequence in the rear part for fusion processing, and the heterogeneous collaborative fusion operation performed on the combined heterogeneous features is specifically as follows:

[0101] According to the feature source labels and the interaction importance labels of the graph features and the physical property features, the two types of features are input into a joint weighting unit, the dominant weighting coefficients are assigned to the graph features, and the auxiliary weighting coefficients are assigned to the physical property features, so that the dominant features occupy a higher weight in the fusion process;

[0102] The jointly weighted heterogeneous features are input into a nonlinear transformation unit, a nonlinear mapping of the feature space is performed on the heterogeneous features by using an activation function, so that the graph features and the physical property features form a superimposable fused representation in a unified feature domain;

[0103] The fused features after the nonlinear transformation are input into an inter-layer transmission unit, the features are transmitted and updated layer by layer in a multi-layer fusion structure, the high-dimensional graph structure information and the multi-modal physical property information are continuously strengthened in the inter-layer interaction, and finally a fused feature representation used for strategy generation is formed;

[0104] The fused feature representation is input into a strategy vector output layer, a strategy vector is generated by linear mapping, the strategy vector describes the interaction strategy configuration corresponding to the current printing task, and the strategy vector is generated by linear mapping, specifically as follows:

[0105] The fused feature representation is subjected to feature weighting processing according to a preset weight distribution rule, the dominant features from the graph structure channel and the auxiliary features from the physical property channel are combined in a weighted manner to obtain a to-be-mapped feature vector, and the preset weight distribution rule is as follows:

[0106] The features from the graph structure channel are assigned with dominant weights, and the features reflecting the importance of the interaction nodes, the transfer relationship of the interaction edges and the stability of the sub-graphs are taken as the main contribution features;

[0107] The features from the physical property channel are assigned with auxiliary weights, and the weight proportions are dynamically set according to the influence degrees of the time physical property, the behavior physical property, the device load physical property and the environmental physical property on the current printing task;

[0108] The dominant weights and the auxiliary weights are subjected to normalization processing, so that the two types of weights satisfy the constraint that the weight sum is one in the same scale space;

[0109] Input the to-be-mapped feature vector into a linear mapping unit, linearly combine each feature component in the vector according to the corresponding position of a fixed mapping matrix, and generate a linear combination result for representing an interaction policy element;

[0110] Perform scale compression processing on the linear combination result, normalize the output result in a preset policy vector range, and take the normalized vector as a final policy vector.

[0111] In the embodiment, the generation of the strategy capsule execution sequence corresponding to the current task drives the printing module to perform a printing operation, including:

[0112] According to the interaction strategy configuration parameters parsed from the policy vector output by the improved GraphSAGE network, the interaction strategy configuration parameters include a target interaction node set, an execution priority, a mandatory node label, a skippable node label, an exception handling label, and an energy weight parameter;

[0113] For each interaction node in the target interaction node set, a corresponding strategy capsule is generated, and the interface template information, the trigger condition set, the skippable output relationship set, the execution parameter set, and the energy weight parameter are encapsulated in the same strategy capsule to form a multi-capsule set to be issued, which is associated with a current task identifier, a device type identifier, and a strategy version number, and then issued to a printing handheld terminal device through an Internet of Things platform. For each interaction node in the target interaction node set, a corresponding strategy capsule is generated, specifically:

[0114] According to the interface requirements, field input structure, and operation constraints of the target interaction node, a corresponding interface template is selected from a template library, and the interface display content, fillable fields, and triggerable operations are encapsulated into the template field of the strategy capsule;

[0115] According to the task logic condition, device state condition, and physical property feature condition of the target interaction node, a trigger condition set and a skippable output relationship set are generated, and the entry condition, exit condition, and next node jump relationship are written into the strategy capsule;

[0116] According to the complexity, key degree, and failure risk of the target interaction node, an execution parameter set and an energy weight parameter are generated, and the field verification rule, exception handling requirement, confirmation step, and energy weight are attached to the strategy capsule to form a complete strategy capsule structure;

[0117] After the printing handheld terminal device receives the multi-capsule set, it is adapted and filtered according to the current task type, real-time device state, and historical execution context, and the strategy capsules that do not match the task constraints or conflict with the device state are removed to obtain an initial candidate strategy capsule set.

[0118] Perform fission processing on the composite strategy capsule in the initial candidate strategy capsule set:

[0119] Split into the minimum functional unit of interface display, field verification, exception handling, user confirmation and security check, and split different condition branches into independent sub-capsules with the trigger condition as the boundary;

[0120] When the energy weight parameter exceeds the preset threshold or the device state is in a limited state, split the high-cost step and the key check step into independent sub-capsules and move to the priority position of the execution sequence;

[0121] The sub-capsules obtained by splitting are sequentially maintained and conflict is resolved according to the dependency relationship;

[0122] Perform consistency check on the fission result to ensure consistent association of sub-capsules in version number, task identification and trigger condition, forming a candidate sub-capsule set;

[0123] According to the energy weight parameter, execution priority and trigger condition in the candidate sub-capsule set, construct a strategy capsule execution sequence that meets the task execution sequence constraint, map to the interface display step, field verification step, exception prompt step and user confirmation step, complete the dynamic interaction process of the current printing task, and drive the printing module to perform printing operation.

[0124] In the embodiment, the dynamics parameters of the reversible interactive dynamic subgraph are updated and the GraphSAGE network is improved, including:

[0125] After completing the current printing task, the printing handheld terminal device packages the printing quality label corresponding to this interaction process, the printing result mark, the interaction event summary, the strategy capsule execution sequence, the exception record and the time consumption statistics into printing session feedback data;

[0126] The Internet of Things platform analyzes and aligns the printing session feedback data, establishes a one-to-one correspondence between the strategy version number, the device type identification, the task identification and the strategy capsule execution sequence, and stores the interaction event summary and the printing quality label in association according to the session, and generates a standardized feedback record;

[0127] The Internet of Things platform updates the reversible interactive dynamic subgraph based on the standardized feedback record, recalculates the state statistics of the interaction node according to the number of entries, the number of successes and the number of failures of the interaction node, recalculates the transfer statistics of the interaction edge according to the number of triggers, the number of successful transfers and the number of failed transfers of the interaction edge, updates the reverse backtracking marker and the backtracking trigger threshold according to the exception record, generates an updated version of the reversible interactive dynamic subgraph and saves the version number;

[0128] The Internet of Things platform updates parameters of the improved GraphSAGE network based on standardized feedback records, jointly updates parameters of the graph structure channel and the physical property channel by using a training configuration corresponding to a policy version number, generates a new policy version, and delivers the new policy version to the printing handheld terminal device.

[0129] Reference Figure 2 An Internet of Things-based printing handheld terminal device interaction system includes the following modules:

[0130] A data acquisition module is configured to acquire user interface operation events and multi-modal physical property data, and form an interaction event sequence and a physical property data sequence.

[0131] An event processing and subgraph construction module is configured to perform time alignment and type normalization on the interaction event sequence, and generate a reversible interactive dynamic subgraph.

[0132] A physical property feature fusion module is configured to extract time physical property features, behavior physical property features, device load physical property features, and environmental physical property features from the physical property data sequence, and generate a multi-modal fusion interaction physical property vector.

[0133] A strategy generation module is configured to construct an improved GraphSAGE network, perform heterogeneous collaborative convergence, and generate a strategy vector.

[0134] A strategy capsule execution module is configured to generate a strategy capsule based on the strategy vector and deliver the strategy capsule to the terminal device, perform fission processing on the strategy capsule, and form a strategy capsule execution sequence.

[0135] A feedback update module is configured to receive printing quality labels and printing session feedback data, update dynamic parameters of the reversible interactive dynamic subgraph, and update the improved GraphSAGE network.

[0136] Embodiment 1

[0137] To verify the feasibility of the present application in implementation, the present application is applied to express sorting of an intelligent warehouse center. Field operators need to use a printing handheld terminal device to generate a large number of package labels, abnormal item labels, and temporary supplementary recording labels every day. The daily package throughput of the warehouse center is about 60,000-70,000 pieces, the sorting line rhythm is compact, and the operators need to maintain continuous operation under different temperature and humidity, different shifts, and different workloads. However, the traditional printing handheld terminal device adopts a fixed interaction process and does not have the ability to adaptively adjust according to task changes, environmental changes, or device state changes, resulting in frequent system errors, printing failures, and process stalls in night shifts, low temperatures, and personnel fatigue.

[0138] In this embodiment, the printing handheld terminal device interaction method based on Internet of Things is applied to deploy 20 printing handheld terminals supporting Internet of Things connection in the warehouse center. The application first collects user interface operation events and device temperature, power, illumination, humidity and other multi-modal physical property data in real time through the data acquisition module and uploads them to the Internet of Things platform. The system collects about 120,000 interaction events in one day, with an average of about 850-1000 physical property data records uploaded per hour per terminal.

[0139] The Internet of Things platform time-aligns and type-normalizes the collected interaction events, maps events such as "confirm printing", "go back one step", "field input error" and the like to standardized event types. Then, according to the logic of the events, interaction nodes are built, and according to the interface jump relationship between the nodes, interaction edges are established. The system cumulatively builds 142 interaction nodes and 268 interaction edges on the 3rd day, and automatically identifies 5 high-frequency process paths and 3 high-risk paths, thereby forming a reversible interaction dynamic subgraph with forward and reverse deduction capabilities.

[0140] The system extracts time physical property features, behavior physical property features, device load physical property features and environmental physical property features from the physical property data. For example, in the case of early shift temperature of about 20℃ and night shift temperature of about 12℃, the decrease in device temperature will affect the printing image quality. The application generates a multi-modal fusion interaction physical property vector that reflects user behavior, device state and environmental state by fusing different physical property features.

[0141] The application constructs an improved GraphSAGE network, inputs the local subgraph features of the reversible interaction dynamic subgraph into the graph structure channel, inputs the multi-modal fusion interaction physical property vector into the physical property channel, and performs a heterogeneous collaborative convergence generation strategy vector. The strategy vector is used to describe the interaction strategy adapted to the current task, which can be flexibly changed according to the device temperature, power decline trend or user mis-touch frequency. For example, when detecting an increase in night shift mis-touch rate, the strategy vector will tend to add more field confirmation steps, thereby reducing false printing.

[0142] The system generates a strategy capsule according to the strategy vector and issues it to the terminal side. The terminal splits the strategy capsule according to the real-time state, for example, splits the confirm printing capsule into a print head verification and a print confirmation sub-capsule, and accordingly reconstructs the interaction process of the current task. In actual testing, when the device power is lower than 20% and the downward trend is obvious, the system will automatically insert a power prompt capsule and a lightweight mode capsule to avoid interruption of the printing process.

[0143] After completing each printing task, the terminal uploads the printing results, printing quality labels, interactive path summaries, and other feedback to the platform. The system dynamically updates the reversible interactive force subgraph and improves the GraphSAGE network based on these feedback, so that the strategy is more suitable for the actual operating environment. During the 10-day operation period, the system accumulates a total of 92,000 printing session feedback data, the interactive path gradually converges, and the strategy generation accuracy improves from 81% to 92%.

[0144] Table 1 Warehouse center field test key performance data statistics table

[0145] Statistical date Daily average print task volume Average time per single task (seconds) Print failure rate User mis-touch frequency (times / day) Low-temperature condition failure rate Strategy generation accuracy rate Day 1 65,200 9.8 6.1% 1,320 4.3% 81% Day 3 66,010 9.2 5.0% 1,210 3.8% 85% Day 5 66,500 8.7 4.2% 1,050 3.1% 88% Day 7 67,100 8.1 3.6% 980 2.7% 90% Day 10 67,420 7.9 3.1% 920 2.2% 92%

[0146] As can be seen from the table data, under the condition of basically stable daily printing task volume, the printing efficiency is significantly improved by the method. The average time consumption of a single task decreases from 9.8 seconds on the first day to 7.9 seconds on the tenth day, showing a continuous downward trend. This indicates that as the reversible interactive force subgraph is continuously improved and the strategy vector generation capability is enhanced, the system can automatically select a more concise and efficient interactive path, reducing unnecessary interface jumps and repeated operations, thereby significantly improving the overall work rhythm.

[0147] The printing failure rate and the low-temperature condition failure rate in the table further demonstrate the advantages of the application in device state and environmental adaptability. The printing failure rate decreases from 6.1% to 3.1%, and the low-temperature condition failure rate decreases from 4.3% to 2.2%, both showing a stable downward trend. This is due to the system's ability to capture key physical characteristics such as temperature drop and power change in real time through physical property feature fusion, and automatically insert necessary steps such as print head pre-check and power prompt during the strategy capsule fission process, thereby avoiding abnormal printing problems caused by environmental fluctuations in advance.

[0148] The changes in user mis-touch frequency and strategy generation accuracy further reflect the synergistic effect of the application in interactive optimization and continuous learning. The mis-touch frequency decreases from 1,320 times to 920 times in 10 days, while the strategy generation accuracy improves from 81% to 92%, indicating that the system continuously optimizes the strategy generation model as feedback data accumulates, more accurately identifies user operation patterns and field work characteristics, and automatically adjusts interactive steps and paths, making the interactive process more stable, smooth, and in line with the actual needs of operating personnel.

[0149] The above describes only the preferred specific embodiments of the application, but the protection scope of the application is not limited thereto. Any skilled person in the art can make equivalent substitutions or changes to the technical solutions and inventive concepts of the application within the scope of the disclosed technology, which should be covered within the protection scope of the application.

Claims

1. A printing handheld terminal device interaction method based on the Internet of Things, characterized in that, The method comprises the following steps: A print handheld terminal device establishes a communication connection with an Internet of Things platform, collects user interface operation events and multi-modal physical property data, and forms an interaction event sequence and a physical property data sequence; The interaction event sequence and the physical property data sequence are uploaded to the Internet of Things platform, the interaction event sequence is subjected to time alignment and type normalization processing, interaction nodes and interaction edges are constructed, and a reversible interaction dynamic subgraph is formed; Based on the physical property data sequence, time physical property features, behavior physical property features, device load physical property features and environmental physical property features are extracted, the various physical property features are subjected to normalization, coding and fusion, and a multi-modal fusion interaction physical property vector is generated; An improved GraphSAGE network is constructed, local subgraph features of the reversible interaction dynamic subgraph are input into a graph structure channel, the multi-modal fusion interaction physical property vector is input into a physical property channel, heterogeneous collaborative convergence is performed, and a strategy vector is generated; According to the strategy vector, a plurality of strategy capsules are generated, the strategy capsules are identified by version and then distributed to the print handheld terminal device through the Internet of Things platform, the strategy capsules are subjected to fission processing, a strategy capsule execution sequence corresponding to a current task is generated, and a print module is driven to perform a print operation; The print handheld terminal device uploads print quality labels and print session feedback data to the Internet of Things platform, updates dynamics parameters of the reversible interaction dynamic subgraph, and updates the improved GraphSAGE network.

2. The printing handheld terminal device interaction method based on the Internet of Things according to claim 1, characterized in that, The user interface operation events include interface button click events, text input events, interface jump events, interface stay events, cancel operation events, return operation events, long press operation events, sliding operation events and print instruction trigger events.

3. The printing handheld terminal device interaction method based on the Internet of Things according to claim 1, characterized in that, The multi-modal physical property data includes device power data, device voltage data, print head temperature data, consumable remaining amount data, signal strength data, environmental temperature data, environmental humidity data, environmental illumination data, device posture data and device motion state data.

4. The printing handheld terminal device interaction method based on the Internet of Things according to claim 1, characterized in that, The method for forming the reversible interaction dynamic subgraph with forward deduction and reverse deduction capabilities comprises the following steps: The interaction event sequence and the physical property data sequence are received, the interaction events are sorted and de-duplicated according to timestamps, incomplete events and repeated reporting events are removed, and a standard interaction event sequence arranged in chronological order is generated; According to the interface identifier, event type, task stage, abnormality marker and print result marker corresponding to each interaction event, the interaction events are subjected to type normalization processing, and the forward state marker and the backtracking state marker are respectively assigned to each interaction event; Events with the same interface identifier, event type combination, forward state marker and backtracking state marker are merged into the same interaction node, the corresponding entry condition, exit condition, associated task stage and the number of successful events and failed events in each interaction node are recorded, and an interaction node set with forward state information and backtracking state information is formed; According to the time sequence and interface switching relationship between adjacent events in the standard interaction event sequence, a directed interaction edge is established between the interaction node set, and a trigger condition, an associated success result mark and a failure result mark are recorded for each interaction edge; a backtracking trigger mark is distributed according to the failure event proportion and the abnormal event proportion; an interaction edge with a high failure proportion and a high abnormal proportion is marked as a reverse priority backtracking edge to form an interaction edge set containing forward transfer edges and reverse backtracking edges; A node subset and an edge subset related to the printing task are selected from the interaction node set and the interaction edge set; a forward adjacent node set and a backtracking adjacent node set are established for each interaction node based on the forward state mark and the backtracking state mark of each node and the backtracking trigger mark on the interaction edge; and the node subset, the edge subset, the forward adjacent relationship and the backtracking adjacent relationship are combined to form a reversible interaction dynamic subgraph.

5. The printing handheld terminal device interaction method based on the Internet of Things according to claim 1, characterized in that, The generating the multi-modal fusion interaction property vector comprises: aligning the property data sequence according to the time stamp, dividing the multi-modal property data corresponding to each interaction event into a time property sub-sequence, a behavior property sub-sequence, a device load property sub-sequence and an environment property sub-sequence, and establishing a time index relationship between each sub-sequence and the interaction event sequence; In the time property sub-sequence, the property data in each window is statistically processed according to a preset time window, and time property features including stay duration, input duration, event trigger frequency in the window are extracted; the time property features are normalized to obtain a standardized time property feature set; In the behavior property sub-sequence, the number of false touches, the number of repeated operations, the number of canceled operations, the number of return operations and the operation sequence stability are counted according to the user interface operation event and the corresponding property data record to form a behavior property feature set; the behavior property features are encoded and normalized to obtain a standardized behavior property feature set; In the device load property sub-sequence and the environment property sub-sequence, device load property features and environment property features are extracted based on device power change, print head temperature change, consumable remaining amount change, and environment temperature change, environment humidity change and environment illumination change, respectively; the device load property features and the environment property features are subjected to outlier correction and dimensionless processing to form a standardized device load property feature set and a standardized environment property feature set; The standardized time property feature set, the standardized behavior property feature set, the standardized device load property feature set and the standardized environment property feature set are combined according to a preset feature splicing order; the combined feature set is mapped to a fixed-length feature representation through a unified encoding method to form a multi-modal fusion interaction property vector.

6. The printing handheld terminal device interaction method based on the Internet of Things according to claim 1, characterized in that, The generating the strategy vector comprises: An improved GraphSAGE network is constructed, which includes an input layer, a graph structure channel, a property channel, a heterogeneous collaborative convergence layer and a strategy vector output layer; the input layer receives the local subgraph features of the reversible interaction dynamic subgraph and the multi-modal fusion interaction property vector; In the graph structure channel, a subset of interaction nodes and a subset of interaction edges related to the target printing task in the reversible interaction dynamic subgraph are input as a local subgraph, an initial feature vector of each interaction node in the local subgraph is assigned, an adjacency relationship description is established based on the directed connection relationship between the interaction nodes, the node features of the local subgraph are updated layer by layer using a multi-layer graph aggregation structure, and a graph feature representation of the graph structure channel is obtained; In the physical property channel, the multi-modal fusion interaction physical property vector is divided into multiple physical property sub-vectors according to time physical property, behavior physical property, equipment load physical property and environment physical property, a physical property confidence coefficient is generated based on the missing condition, fluctuation amplitude and historical stability of each physical property sub-vector, and the physical property sub-vectors are screened and combined using a gate re-weighting structure driven by the physical property confidence coefficient, to obtain a physical property feature representation; In the heterogeneous collaborative convergence layer, the graph feature representation and the physical property feature representation are aligned and combined according to a preset fusion order, and a heterogeneous collaborative convergence operation is performed on the combined heterogeneous features, the heterogeneous collaborative convergence operation including joint weighting, non-linear transformation and inter-layer transmission of the graph features and the physical property features, to generate a fusion feature representation; The fusion feature representation is input into a policy vector output layer to generate a policy vector through linear mapping, the policy vector describing an interaction policy configuration corresponding to the current printing task.

7. The printing handheld terminal device interaction method based on the Internet of Things according to claim 1, characterized in that, The generation of the policy capsule execution sequence corresponding to the current task drives the printing module to perform a printing operation, including: According to the policy vector output by the improved GraphSAGE network, interaction policy configuration parameters are parsed, the interaction policy configuration parameters including a target interaction node set, an execution priority, a must-pass node marker, a skipable node marker, an exception handling marker and an energy weight parameter; For each interaction node in the target interaction node set, a corresponding policy capsule is generated, and interface template information, a trigger condition set, a jumpable output relationship set, an execution parameter set and the energy weight parameter are encapsulated in the same policy capsule to form a multi-capsule set to be issued, which is associated with a current task identifier, a device type identifier and a policy version number and then issued to a printing handheld terminal device through an Internet of Things platform; After receiving the multi-capsule set, the printing handheld terminal device performs adaptive screening according to the current task type, the real-time state of the device and the historical execution context, and removes policy capsules that do not match the task constraints or conflict with the device state to obtain an initial candidate policy capsule set; The compound policy capsules in the initial candidate policy capsule set are subjected to fission processing: Splitting is performed in the minimum functional unit of interface display, field verification, exception handling, user confirmation and security verification, and different condition branches are split into independent sub-capsules with the trigger condition as the boundary; When the energy weight parameter exceeds a preset threshold or the device state is in a restricted state, high-cost steps and key verification steps are split into independent sub-capsules and moved to a priority position in the execution sequence; The sub-capsules obtained by splitting are sequentially maintained and conflict is resolved according to the dependency relationship; Consistency verification is performed on the fission result to ensure consistent association of the sub-capsules in terms of version number, task identifier and trigger condition, and a candidate sub-capsule set is formed. According to the energy weight parameters, execution priority and trigger conditions in the candidate sub-capsule set, a strategy capsule execution sequence satisfying the task execution sequence constraint is constructed, which is mapped to an interface display step, a field verification step, an exception prompt step and a user confirmation step, a dynamic interaction process of the current printing task is completed, and the printing module is driven to perform a printing operation.

8. The printing handheld terminal device interaction method based on the Internet of Things according to claim 1, characterized in that, The updating of the dynamics parameters of the reversible interaction dynamic subgraph and the updating of the improved GraphSAGE network include: After completing the current printing task, the printing handheld terminal device packages the printing quality label corresponding to the interaction process, the printing result mark, the interaction event summary, the strategy capsule execution sequence, the exception record and the time consumption statistics into printing session feedback data; The Internet of Things platform analyzes and aligns the printing session feedback data, establishes a one-to-one correspondence between the strategy version number, the device type identifier, the task identifier and the strategy capsule execution sequence, stores the interaction event summary and the printing quality label in association according to the session, and generates a standardized feedback record; The Internet of Things platform updates the reversible interaction dynamic subgraph based on the standardized feedback record, recalculates the state statistics of the interaction nodes according to the number of entries, the number of successes and the number of failures of the interaction nodes, recalculates the transfer statistics of the interaction edges according to the number of triggers, the number of successful transfers and the number of failed transfers of the interaction edges, updates the reverse backtracking marker and the backtracking trigger threshold according to the exception record, generates an updated version of the reversible interaction dynamic subgraph and saves the version number; The Internet of Things platform updates the improved GraphSAGE network based on the standardized feedback record, updates the parameters of the graph structure channel and the physical property channel jointly using the training configuration corresponding to the strategy version number, generates a new strategy version and distributes it to the printing handheld terminal device.

9. A printing handheld terminal device interaction system based on Internet of Things, which executes the printing handheld terminal device interaction method based on Internet of Things in any one of claims 1 to 8, characterized in that, It includes the following modules: A data acquisition module for acquiring user interface operation events and multi-modal physical property data to form an interaction event sequence and a physical property data sequence; An event processing and subgraph construction module for time alignment and type normalization of the interaction event sequence to generate a reversible interaction dynamic subgraph; A physical property feature fusion module for extracting time physical property features, behavior physical property features, device load physical property features and environment physical property features from the physical property data sequence to generate a multi-modal fusion interaction physical property vector; A strategy generation module for constructing an improved GraphSAGE network, performing heterogeneous collaborative convergence, and generating a strategy vector; A strategy capsule execution module for generating a strategy capsule from the strategy vector and distributing it to the terminal, performing fission processing on the strategy capsule, and forming a strategy capsule execution sequence; A feedback updating module for receiving a printing quality label and printing session feedback data, updating the dynamics parameters of the reversible interaction dynamic subgraph and updating the improved GraphSAGE network.

Images