A Collaborative Design Method for Recording High-Fidelity Compression and Playback Based on Behavior Sequence Fusion
By constructing operation-dependent topology graphs and non-uniform rational B-spline parametric equations, the problems of large data storage volume and low playback fidelity in collaborative design systems are solved, achieving efficient data compression and synchronous playback.
Patent Information
- Application Number
- CN202611146421.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-30
- Publication Date
- 2026-11-17
AI Technical Summary
Existing technologies in collaborative design systems require large amounts of stored data and struggle to balance trajectory smoothness with discrete state synchronization timing, resulting in low playback fidelity and a lack of redundant data identification mechanisms, which increases storage consumption.
By constructing an operation-dependent topology graph based on behavior sequence fusion to remove redundant nodes, fitting non-uniform rational B-spline parametric equations, and combining interpolation analysis and clock management, high-fidelity compression and playback of discrete events and continuous data are achieved.
While reducing the amount of stored data, maintain system state consistency, ensure data playback timing synchronization and fidelity, and reduce space resource consumption.
Smart Images

Figure CN123115106A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of data processing technology, specifically to a method for high-fidelity compression and playback of collaborative design records based on behavioral sequence fusion. Background Technology
[0002] Collaborative design systems generate a large amount of operational data in multi-user concurrent operation environments. Saving and replaying operation records plays a crucial role in version tracking and state synchronization. The collaborative design process involves both continuous data corresponding to coordinate movement and discrete data corresponding to state changes. Existing solutions mostly use a frame-by-frame snapshot mode to uniformly save the mixed data, resulting in excessive storage consumption.
[0003] Existing technologies, when processing mixed data, treat continuous parameter changes and discrete state abrupt changes separately. During the playback and analysis stage, interpolation calculations are performed at a fixed frequency, resulting in an inability to simultaneously ensure the smoothness of the playback trajectory and the synchronization timing of discrete states. When the sampling frequency increases, continuous spatial data creates a storage and computing burden. When the sampling frequency decreases, the timing of discrete event insertion lacks accurate positioning, leading to state asynchrony and timing collisions within the analysis process, thus reducing playback fidelity.
[0004] Furthermore, the discrete operation sequence generated by the collaborative process contains intermediate states that do not constitute the final constraint. The existing architecture lacks a mechanism to identify and eliminate redundant process data. Storing all intermediate states increases storage consumption. At the same time, there is currently no mathematical mechanism to integrate discrete operation instructions into the continuous numerical analysis pipeline. It is impossible to establish a closed-loop relationship between the spatial data compression model and the action timing synchronization model, which cannot meet the high-fidelity playback requirements in high-concurrency scenarios. Summary of the Invention
[0005] To address the shortcomings of existing technologies, this invention provides a high-fidelity compression and playback method for collaborative design recording based on behavior sequence fusion. This method solves the technical problems of existing technologies, such as the large amount of data to be stored when recording and playing back concurrent collaborative operation data, and the difficulty in balancing trajectory smoothness and discrete state synchronization timing.
[0006] To achieve the above objectives, the present invention provides the following technical solution:
[0007] A collaborative design method for recording high-fidelity compression and playback based on behavior sequence fusion includes the following steps:
[0008] It receives concurrent collaborative operation data streams and divides them into discrete event sets and continuous event sets;
[0009] The discrete event set is mapped to an operation-dependent topology graph, and transitional nodes are removed to generate a reduced discrete topology event set.
[0010] Traverse the time domain of the continuous event set to retrieve the intersection events, extract the physical timestamps and map them to the normalized parameter domain, establish the topological event anchor points through time-series anti-collision offset processing, and establish the index association between the topological event anchor points and the corresponding discrete event payloads.
[0011] The continuous event set is fitted to a non-uniform rational B-spline parametric equation. The topological event anchor points are injected as heavy nodes into the node vector to induce a reduction in geometric continuity. Geometric split anchor points are injected based on the fitting residuals. The data used to reconstruct the non-uniform rational B-spline parametric equation, the reduced discrete topological event set, and the index association are persistently stored.
[0012] Interpolation analysis is performed along the non-uniform rational B-spline parametric equation. When the stepping reaches the derivative singularity point caused by the heavy node, the playback rendering clock is suspended. The discrete topology state synchronization instruction or no-operation skip rule is executed according to the anchor point type, and then smooth interpolation is resumed.
[0013] The technical principle of this invention is as follows: For the discrete state changes of concurrent collaborative data streams, a directed acyclic graph is introduced to construct a state-dependent topology. By comparing the in-degree and out-degree relationships, redundant intermediate transition operations are folded, and transitional state data that do not generate final system constraints are eliminated. For continuous geometric editing, it is parameterized into a non-uniform rational B-spline curve.
[0014] By utilizing the property that the number of nodes determines geometric continuity in the non-uniform rational B-spline algorithm, the triggering time of discrete events is mapped to the node vector of the continuous curve equation. By injecting nodes with the same number of curve degrees, singular points with discontinuous first derivatives are constructed inside the smooth curve.
[0015] During client-side playback interpolation parsing, the parsing engine uses derivative singularities as clock interrupt trigger sources. When parameters step to a node, the clock is suspended, allowing discrete topology events to be inserted into the rendering pipeline for execution. By fusing parametric equations with the topology graph, continuous degradation replaces system polling and timestamp comparison, compressing data volume and maintaining the consistency of discrete state synchronization and the timing of continuous spatial trajectory playback.
[0016] Furthermore, the step of receiving concurrent collaborative operation data streams and dividing them into discrete event sets and continuous event sets includes:
[0017] Extract the physical timestamp, operation attribute code, target object's globally unique identifier, and state change increment value carried within the original operation event stream;
[0018] Establish a business logic configuration dictionary, and divide the global heterogeneous operation event stream into the discrete event set and the continuous event set by looking up and comparing the operation attribute codes;
[0019] Aggregation and grouping processing is performed on the continuous event set, and the fixed time window parameter and the action termination event trigger sequence are combined to truncate and output an independent continuous operation sequence.
[0020] Furthermore, the step of mapping the discrete event set to an operation-dependent topology graph and removing transitional nodes to generate a reduced discrete topology event set includes:
[0021] Map the operation events contained in the discrete event set to a directed acyclic graph data structure, and initialize an operation-dependent topology graph containing a set of vertices and a set of directed edges;
[0022] A depth-first traversal search algorithm is executed within the operation-dependent topology graph to compare the relationships between node state changes and detect redundant operation subgraphs in the topology network.
[0023] Based on the location-based redundant operation subgraph, the topology folding processing instruction is executed to clean up the exploration paths that have not generated final system constraints and generate the reduced discrete topology event set.
[0024] Furthermore, the step of executing topology folding processing instructions on the location-based redundant operation subgraph includes:
[0025] For the inverse operation subgraph, if there is no external dependency relationship between the intermediate transition vertices on the corresponding path, delete all intermediate transition vertices and their corresponding directed edges on the corresponding path.
[0026] For the subgraph of the covering operation, when there is no external dependency relationship between the intermediate assignment transition state vertex, the intermediate assignment transition state vertex is extracted, the initial in-degree event boundary at the front of the time axis and the final out-degree event boundary at the end of the time axis are retained, and a directed edge is established to directly jump from the initial in-degree event boundary to the final out-degree event boundary.
[0027] Furthermore, the step of traversing the temporal domain of the continuous event set to retrieve the intersection events includes:
[0028] Read the start and end physical timestamps contained in the continuous operation sequence to construct the time domain;
[0029] Traverse the reduced discrete topological event set. When the physical time node of the discrete event is within the closed interval of the time domain, and the globally unique identifier of the target object of the discrete event is consistent with the globally unique identifier of the target object of the continuous operation sequence, the discrete event is established as an intersection event.
[0030] Furthermore, the step of extracting the physical timestamp, mapping it to the normalized parameter domain, and establishing the topology event anchor point through temporal anti-collision offset processing includes:
[0031] Calculate and output the initial mapping parameters. If the numerical distance between the initial mapping parameters and the parameter values stored in the set of registered anchor points is not greater than the set minimum resolution threshold of the parameter domain, a timing collision is determined to have occurred.
[0032] Apply a temporal anti-collision offset equal to the minimum resolution threshold of the parameter domain to the initial mapping parameters, calculate the output offset adjustment parameters, and trigger a loop recursive verification instruction until the absolute value distance between the offset adjustment parameters and the parameter values stored in the registered anchor point set is greater than the minimum resolution threshold of the parameter domain.
[0033] The offset adjustment parameter, which is recursively verified, is established as the topological event anchor point.
[0034] Furthermore, the steps of fitting the continuous event set to a non-uniform rational B-spline parametric equation and injecting the topological event anchor points as heavy nodes into the node vector to induce a degradation in geometric continuity include:
[0035] For continuous operation sequences that are represented as discrete spatiotemporal point clouds in the spatial dimension, the non-uniform rational B-spline parametric equations are constructed, which are jointly driven by the control point sequence, node vectors, and weight factors.
[0036] The parameter values corresponding to the topological event anchor point are forcibly injected into the node vector according to the multi-node superposition mechanism, so that the final node multiplicity after injection at the topological event anchor point is consistent with the degree parameter of the non-uniform rational B-spline parameter equation.
[0037] Update the control points on both sides of the topological event anchor point respectively, so that the curve remains positionally continuous at the topological event anchor point, without applying the first derivative continuity constraint to both sides of the anchor point, thus forming a first derivative discontinuity feature at the topological event anchor point.
[0038] Furthermore, the step of injecting geometric splitting anchor points based on fitted residuals includes:
[0039] The shortest distance from each original sampling point inside the discrete spatiotemporal point cloud to the curve characterizing the non-uniform rational B-spline parametric equation is calculated as the fitting residual value, and the maximum fitting residual value is extracted from all fitting residual values.
[0040] When the maximum fitting residual is greater than the set tolerance threshold, the time parameter of the original sampling point corresponding to the maximum fitting residual is established as the geometric splitting anchor point.
[0041] The geometric splitting anchor points are injected into the node vector according to the heavy node superposition mechanism, triggering the equation splitting and reconstruction recursive calculation instruction until all fitting residual values are no greater than the tolerance threshold.
[0042] Furthermore, the step of suspending the playback rendering clock when performing interpolation analysis along the non-uniform rational B-spline parametric equation and stepping to the derivative singularity point caused by the heavy node includes:
[0043] Initialize the playback end rendering clock, extract the current physical timestamp output by the playback end rendering clock, and calculate the current interpolation parameters;
[0044] Spatial coordinate interpolation calculation is performed along the non-uniform rational B-spline parametric equation. The heavy node parameters in the front-open-back-closed interval formed by the previous interpolation parameter and the current interpolation parameter are retrieved, and the node multiplicity of each heavy node parameter in the node vector is read.
[0045] When the number of node multiplicity of any node parameter is the same as the degree parameter value of the non-uniform rational B-spline parameter equation, the current interpolation parameter pointer is positioned at the derivative singular point, and a logic interrupt is triggered by a software instruction to forcibly suspend the playback end rendering clock.
[0046] Furthermore, the step of executing discrete topology state synchronization instructions or no-operation skip rules based on anchor point type includes:
[0047] Extract the current interpolation parameter corresponding to the derivative singularity point as the query key, and perform an index query based on the index association in the reduced discrete topology event set;
[0048] When a specific discrete event payload is found to be associated, the current interpolation parameter is determined to belong to the topological event anchor point. The discrete event payload is then pushed into the main rendering thread, and the corresponding discrete topological state synchronization instruction is executed.
[0049] When no discrete event load is found, the current interpolation parameter is determined to belong to the geometric split anchor point. The no-operation skip rule is executed to skip the discrete state synchronous rendering pipeline and directly step the parameter pointer past the current derivative singularity.
[0050] This invention provides a high-fidelity compression and playback method for collaborative design records based on behavior sequence fusion. It has the following beneficial effects:
[0051] 1. This invention generates a reduced discrete topology event set by constructing an operation-dependent topology graph based on a discrete event set and removing transitional nodes. This cleans up redundant process data while maintaining the consistency of the final state of the system, thereby reducing the amount of system storage data occupied.
[0052] 2. This invention fits a continuous set of events into a time parameter equation and injects topological event anchors as heavy nodes into the time parameter equation to induce a degradation of geometric continuity. It expresses continuous numerical changes as a parameter equation and constructs mathematical singularities within the equation. It builds a fusion storage model of discrete operation sequences and continuous spatiotemporal point clouds, thereby reducing the consumption of spatial data resources.
[0053] 3. This invention performs interpolation analysis along the time parameter equation and suspends the interpolation clock when stepping to the derivative singularity caused by the heavy node. It executes discrete topology state synchronization instructions and no-operation skip rules according to the singularity type, and then resumes smooth interpolation, injecting discrete actions into the continuous numerical analysis pipeline to ensure the timing synchronization fidelity of data playback. Attached Figure Description
[0054] Figure 1 This is a schematic diagram of the system timing interaction of the present invention;
[0055] Figure 2 This is a schematic diagram of the method flow of the present invention;
[0056] Figure 3 This is a comparison diagram of the sparse evolution of the adjacency matrix before and after folding the operation-dependent topology graph of the present invention.
[0057] Figure 4 This is the spatiotemporal feature mapping and temporal anti-collision offset processing judgment logic diagram of the present invention;
[0058] Figure 5 This invention relates to a two-dimensional spatiotemporal point cloud non-uniform rational B-spline fitting and multi-anchor point feature map;
[0059] Figure 6 This is a diagram illustrating the playback parsing pipeline derivative singularity interception and branch state routing judgment of the present invention.
[0060] Figure 7 This is the effect of reconstructing the three-dimensional continuous operation trajectory containing discrete semantic heavy nodes in the present invention. Detailed Implementation
[0061] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0062] See attached document Figure 1 This invention provides a high-fidelity compression and playback system for collaborative design recording based on behavior sequence fusion. The system includes:
[0063] The data stream parsing module is used to receive the concurrent raw operation event stream from the server, parse the event load characteristics, and divide the global heterogeneous operation event stream into discrete event sets and continuous event sets.
[0064] The topology reduction module is used to construct an operation-dependent topology graph based on a discrete event set, perform a depth-first traversal to detect inverse operation subgraphs and covering operation subgraphs, fold intermediate redundant nodes, and output the reduced discrete topology event set.
[0065] The spatiotemporal feature mapping module is used to retrieve matching and associated discrete events within the time domain of a continuous event sequence, extract physical timestamps, perform parameter domain normalization mapping, and generate topological event anchor points that have undergone time-series anti-collision offset processing.
[0066] The spatiotemporal parameter fitting module is used to fit the discrete spatiotemporal point cloud of a continuous event set into a non-uniform rational B-spline parametric equation, inject topological event anchors into the node vector as heavy nodes to induce geometric continuity degradation, and generate geometric split anchors based on the fitting residual threshold.
[0067] The joint parsing and playback module is used to perform parametric equation interpolation based on the rendering clock step, trigger interrupts by identifying mathematical singularities caused by heavy node parameters, and perform discrete topology state synchronization and breakpoint recovery according to the anchor point type.
[0068] See attached document Figure 2 This invention provides a high-fidelity compression and playback method for collaborative design records based on behavior sequence fusion, the method comprising the following steps:
[0069] S1. Receive concurrent collaborative operation data streams, parse operation event load information, and divide the data streams into discrete event sets representing sudden changes in system state and continuous event sets representing changes in parameter values based on operation attributes.
[0070] S2. Map the discrete event set into an operation-dependent topology graph consisting of design component attribute nodes and directed edges of operations. Detect closed-loop reciprocal operation paths and continuous coverage assignment operations through pattern matching. After removing transitional nodes, generate a reduced discrete topology event set.
[0071] S3. Traverse the temporal domain of the continuous event set, retrieve the intersection events that act on the same target object in the reduced discrete topological event set, extract the physical timestamps of the intersection events and map them to the normalized parameter domain, and establish the topological event anchor points after temporal anti-collision offset processing.
[0072] S4. The non-uniform rational B-spline algorithm is used to fit the continuous event set into a time parameter equation. The topological event anchor points are injected as heavy nodes into the equation node vector to induce a local geometric continuity degradation. When the maximum fitting residual exceeds the tolerance threshold, geometric split anchor points are injected. The control points, node vectors with heavy nodes and the reduced discrete topological event set are output and persistently stored.
[0073] S5. Initialize the playback clock and perform interpolation analysis along the parametric equation. When the analysis step reaches the heavy node and causes a derivative singularity, suspend the interpolation clock. Query and execute the corresponding discrete state instruction according to the singularity type. After execution, resume smooth interpolation across the derivative singularity until the playback ends.
[0074] In a specific implementation, step S1 provided by the present invention includes the following steps:
[0075] S101, Data Stream Parsing is set at the system front end to separate instantaneous state change behavior from continuous numerical change behavior.
[0076] The data stream parsing module receives the concurrent raw operation event stream from the server, performs data unpacking on the raw operation event stream, extracts the load characteristics carried inside, and configures the operation event as a five-tuple data structure containing a physical timestamp, user identity identifier, operation attribute code, target object globally unique identifier, and state change increment value.
[0077] Physical timestamps pinpoint the precise moment an event occurs; user identifiers isolate the hardware source of concurrent operations; operation attribute codes define the category of the operation behavior; a globally unique identifier for the target object points to the design component being operated on; and state change increments record the numerical changes in data.
[0078] S102, during the initialization phase, the system establishes a business logic configuration dictionary and establishes a discrete attribute code set and a continuous attribute code set within the configuration dictionary. The data flow parsing module compares the operation attribute codes by looking up the tables and divides the global heterogeneous operation event flow into a discrete event set and a continuous event set.
[0079] Discrete event sets represent the process of sudden changes in system state, including the operation of creating new components, destroying components, and switching Boolean constraint states. Continuous event sets represent the continuous high-frequency change process of parameter values on the time axis, including the operation of moving three-dimensional spatial coordinates and continuously adjusting the scaling ratio. Discrete event sets and continuous event sets have no overlapping elements, and their union constitutes a complete global heterogeneous operation event flow.
[0080] S103, the data stream parsing module performs aggregation and grouping processing on continuous event sets, grouping operation events with the same globally unique identifier and operation attribute code of the target object together, and truncating them with fixed time window parameters and action termination event trigger sequence. The action termination event is the physical release operation of the user's mouse button being lifted or the touch leaving the screen captured by the system front end. Its function is to provide clear truncation boundaries for continuous high-frequency data changes.
[0081] In this embodiment, the fixed time window parameter ranges from 200 milliseconds to 500 milliseconds, outputting multiple independent continuous operation sequences. The data stream parsing module extracts the state change increment value corresponding to each sampled physical timestamp and performs a forward accumulation operation combined with the initial position information to generate a multi-dimensional spatial coordinate set. The continuous operation sequence is represented as a discrete spatiotemporal point cloud morphology within the multi-dimensional space. The continuous operation sequence extraction expression is:
[0082] ;
[0083] In the formula: For the first A continuous sequence of operations; The sequence number parameter is the number of the consecutive operation sequence; This represents the total number of sampled frames at the time of truncation; This is the initial sampling physical timestamp; This is the second sampling physical timestamp; For the first Each sampling physical timestamp; This is the increment value of the state change corresponding to the initial sampling physical timestamp; This is the state change increment value corresponding to the second sampled physical timestamp; For the first The incremental value of the state change corresponding to each sampled physical timestamp.
[0084] S104. For the buffer storage processing of classified data, the existing double-ended queue construction and asynchronous writing methods are used. The data stream parsing module divides the discrete data buffer queue and the continuous data buffer queue in the system memory area to independently store the data packets and continuous operation sequences within the discrete event set.
[0085] See attached document Figure 3 In a specific implementation, step S2 provided by the present invention includes the following steps:
[0086] S201, the discrete behavior topology reduction process is set to transform discrete operations into a graph model, identify invalid operation branches according to the vertex association relationship of the graph model, the topology graph reduction module extracts the operation events contained in the discrete event set, maps the discrete operation events to a directed acyclic graph data structure, initializes the operation dependency topology graph in system memory, and configures the operation dependency topology graph as a topology network model containing a vertex set and a directed edge set.
[0087] Vertices represent the discrete state attributes of design component entities in the editing process and their corresponding design component entities. The topology graph reduction module extracts the operation attribute encoding and state change increment value from the discrete event set and applies them to the previous state of the design component entity to generate the final state and establish the final state as the vertex of the operation-dependent topology graph. Directed edges are associated with specific operation events, representing the state transition behaviors that occur sequentially along the physical time axis. The topology graph reduction module establishes the direction of directed edges based on the order of the physical timestamps of the operation events in the discrete event set, ensuring that the established operation-dependent topology graph does not contain a cycle. The allocation process of the graph data structure in the underlying memory space is carried out according to the existing adjacency list storage method.
[0088] S202, the topology graph reduction module executes a depth-first traversal search algorithm within the operation-dependent topology graph. The module extracts vertices with an in-degree of 0 within the operation-dependent topology graph as the starting point for traversal search, compares the relationship of node state changes, and detects redundant operation subgraphs in the topology network. The depth-first search algorithm belongs to the existing graph data structure traversal technology and is performed based on existing node access marking and stack backtracking methods.
[0089] Redundant operation subgraphs are divided into inverse operation subgraphs and covering operation subgraphs. Inverse operation subgraphs satisfy the following criteria: the initial state corresponding to the starting vertex is topologically equivalent to the final state corresponding to the ending vertex, forming a lifecycle closed loop, containing a continuous path from component instantiation to destruction. Covering operation subgraphs satisfy the following criteria: for the same discrete state attribute associated with the same vertex, multiple consecutive assignments occur on adjacent directed edges, and the out-degree value of the intermediate vertex on the continuous assignment path is always equal to 1.
[0090] S203, the topology graph reduction module executes topology folding instructions based on the redundant operation subgraph of the location. For the reciprocal operation subgraph, the system directly deletes all intermediate transition state vertices and their corresponding directed edges on the corresponding path. For the cover operation subgraph, the system adds external dependency checks. When there are no external dependencies on the intermediate assignment transition state vertices, the system extracts the intermediate assignment transition state vertices, retains the initial in-degree event boundary at the front of the time axis and the final out-degree event boundary at the end of the time axis, and establishes a cross-directioned edge that directly jumps from the initial in-degree event boundary to the final out-degree event boundary.
[0091] After the topology folding process is completed, the topology reduction module cleans up the exploration paths that have not generated final system constraints, generates and outputs the reduced discrete topology event set, and inputs the original discrete event set into the built-in topology folding processing function. It then extracts valid boundary events by combining the structural features of the inverse operation subgraph and the covering operation subgraph, and completes the removal of redundant operations. The reduction operation ensures that the global final state of the system before and after compression remains consistent.
[0092] See attached document Figure 4 In a specific implementation, step S3 provided by the present invention includes the following steps:
[0093] S301, the spatiotemporal feature mapping operation locks the discrete state change actions interspersed within the continuous operation process, establishes a time dimension synchronization relationship, the spatiotemporal feature mapping module extracts the continuous operation sequence temporarily stored in the continuous data buffer queue, reads the start physical timestamp and end physical timestamp contained in the continuous operation sequence, and constructs the time domain.
[0094] The spatiotemporal feature mapping module traverses the reduced discrete topology event set and executes the cross-addressing matching logic. The constraints of the cross-addressing matching logic are set as follows: the physical time node of the discrete event is within the closed interval of the time domain, and the globally unique identifier of the target object of the discrete event is consistent with the globally unique identifier of the target object of the continuous operation sequence. Discrete events that meet the matching conditions are established as intersection events.
[0095] S302, the spatiotemporal feature mapping module extracts the physical timestamps of the intersection event records. Based on existing linear numerical transformation methods, the spatiotemporal feature mapping module maps the physical timestamps associated with the intersection events to the normalized parameter domain, calculates and outputs the initial mapping parameters, and configures the normalized parameter domain as a dimensionless numerical space with values ranging from 0 to 1 within a closed interval. The linear mapping function expression is:
[0096] ;
[0097] In the formula: These are the initial mapping parameters; The physical timestamps for the records of the intersection events; The starting physical timestamp of the continuous operation sequence; This is the physical timestamp that marks the end of a continuous sequence of operations.
[0098] S303, when setting the temporal anti-collision offset to perform coordinate overlap verification when multiple concurrent events are mapped to a one-dimensional parameter space, the spatiotemporal feature mapping module establishes a set of registered anchor points in the system memory area and sets the minimum resolution threshold of the parameter domain.
[0099] The minimum resolution threshold of the parameter domain is determined based on the extreme value of floating-point precision supported by the underlying processor of the system. In this embodiment, the value range of the minimum resolution threshold of the parameter domain is configured to be 0.000001 to 0.00001. The spatiotemporal feature mapping module reads the initial mapping parameters, compares the absolute value distance between the initial mapping parameters and the parameter values stored in the registered anchor point set, and performs a monotonically increasing anti-collision verification mechanism.
[0100] S304, when the numerical distance between the initial mapping parameter and any remaining parameter value within the set of registered anchor points is greater than the minimum resolution threshold of the parameter domain, the spatiotemporal feature mapping module establishes the initial mapping parameter as a topological event anchor point.
[0101] If the numerical distance between the initial mapping parameter and the parameter value stored within the registered anchor point set is not greater than the minimum resolution threshold of the parameter domain, a temporal collision is determined to have occurred. In the event of a temporal collision, the spatiotemporal feature mapping module applies a temporal anti-collision offset to the initial mapping parameter. The value of the temporal anti-collision offset is set to the minimum resolution threshold of the parameter domain itself. The applied temporal anti-collision offset is used to calculate and output the offset adjustment parameter. The formula for calculating the anti-collision offset is:
[0102] ;
[0103] In the formula: Adjust the offset parameters; These are the initial mapping parameters; For the set of registered anchor points; The parameter values stored within the set of registered anchor points; This is the minimum resolution threshold for the parameter domain.
[0104] S305, after applying the temporal anti-collision offset to generate the offset adjustment parameter, the spatiotemporal feature mapping module triggers a cyclic recursive verification instruction for the offset adjustment parameter. During the cyclic application of the temporal anti-collision offset, the module compares the size relationship between the offset adjustment parameter and the upper limit value of the normalized parameter domain. If it is determined that the upper limit value is exceeded, a reverse search is performed and a reverse offset cancellation is performed. At the same time, a comparison with the lower limit value of the normalized parameter domain is added to prevent out-of-bounds errors. When performing offset cancellation, the order of multiple conflict events is maintained based on the physical timestamp order until the absolute value distance between the offset adjustment parameter and the parameter value stored in the registered anchor point set is greater than the minimum resolution threshold of the parameter domain.
[0105] The spatiotemporal feature mapping module establishes the offset adjustment parameters, which are recursively verified, as the topological event anchor points. It writes the parameter values corresponding to the topological event anchor points into the set of registered anchor points and establishes an index association between the topological event anchor points and the discrete event payloads carried by the intersection events.
[0106] See attached document Figure 5 In a specific implementation, step S4 provided by the present invention includes the following steps:
[0107] S401, the spatiotemporal parameter fitting module reads the continuous operation sequence temporarily stored in the continuous data buffer queue. The continuous operation sequence is represented as a discrete spatiotemporal point cloud in the spatial dimension. The spatiotemporal parameter fitting module extracts the physical timestamp of each sampling point in the discrete spatiotemporal point cloud and maps it to the normalized parameter domain as the fitting input parameter.
[0108] The spatiotemporal parameter fitting module constructs non-uniform rational B-spline parametric equations for discrete spatiotemporal point clouds. The non-uniform rational B-spline algorithm is a geometric modeling algorithm based on node vector control of local curve shapes, used to transform discrete point sets into analytical function expressions. The non-uniform rational B-spline parametric equations are jointly driven by three data structures: control point sequences, node vectors, and weight factors. The formula for the non-uniform rational B-spline parametric equations is as follows:
[0109] ;
[0110] In the formula: Calculate function values for parametric curves showing spatial location as a function of parameters; These are normalization parameters; The parameter is the traversal index of the summation item; Subtract 1 from the total number of control points; The degree parameter is the basis function of the non-uniform rational B-spline; The first parameter calculated based on the same order of node vectors. One non-uniform rational B-spline basis function; For the first Weight factors corresponding to each control point; For the first Spatial coordinates of the control points.
[0111] S402, the dynamic constraint injection principle breaks the smooth and continuous characteristics of conventional fitting curves. It constructs derivative discontinuities at associated time nodes, which act as forced breakpoints for triggering discrete events. The spatiotemporal parameter fitting module extracts topological event anchor points, and then forcibly injects these anchor points into the node vector associated with the non-uniform rational B-spline parametric equation. The injection operation adopts a multi-node superposition mechanism.
[0112] Inside the node vector, the spatiotemporal parameter fitting module copies and superimposes the parameter values corresponding to the topological event anchor point. During node insertion, the control point sequence is updated synchronously to maintain the curve shape. This ensures that the final node multiplicity at the parameter value is consistent with the degree parameter of the non-uniform rational B-spline parametric equation. The spatiotemporal parameter fitting module updates the control points on both sides of the topological event anchor point, ensuring that the curve remains positionally continuous at the topological event anchor point. However, no first-order derivative continuity constraint is applied to both sides of the anchor point, resulting in a first-order derivative discontinuity feature at the topological event anchor point.
[0113] S403, the geometric split anchor point generation is set to truncate and segment the fit at the error limit position. The spatiotemporal parameter fitting module calculates the shortest distance from each original sampling point inside the discrete spatiotemporal point cloud to the non-uniform rational B-spline parametric equation characterization curve, and establishes the shortest distance as the fitting residual value. The spatiotemporal parameter fitting module extracts the maximum fitting residual from all fitting residual values.
[0114] The system memory pre-sets a tolerance threshold, which is determined based on the extreme value of the physical distance mapped from the screen space resolution of the front-end rendering device to the three-dimensional world coordinate system. In this embodiment, the tolerance threshold ranges from 0.01 mm to 0.1 mm. The spatiotemporal parameter fitting module compares the relationship between the maximum fitting residual and the tolerance threshold.
[0115] When the maximum fitting residual is greater than the tolerance threshold, the spatiotemporal parameter fitting module establishes the time parameter of the original sampling point corresponding to the maximum fitting residual as the geometric split anchor point. The spatiotemporal parameter fitting module also injects the geometric split anchor point into the node vector according to the heavy node superposition mechanism, triggering the equation splitting and reconstruction recursive calculation instruction. The equation splitting and reconstruction recursive calculation instruction is a process of splitting the current curve segment into two independent sub-curves and recalculating the control point position based on the node insertion algorithm. Its function is to forcibly break the fitting interval at the error exceeding the limit to reduce the local geometric residual, until all fitting residual values are not greater than the tolerance threshold.
[0116] S404, the spatiotemporal parameter fitting module performs persistent storage processing on the fitted data. The persistent storage data structure is configured to contain four independent data blocks.
[0117] The first data block stores the control point sequence and associated weight factors output from the inverse calculation of the non-uniform rational B-spline parametric equation; the second data block stores the node vector with topological event anchor points and geometric split anchor points; the first and second data blocks together constitute the data used to reconstruct the non-uniform rational B-spline parametric equation; the third data block stores the reduced discrete topological event set output by the topology reduction module and the index association between the topological event anchor points and the corresponding discrete event loads; the fourth data block stores the continuous operation sequence identifier, the target object's globally unique identifier, the operation attribute code, the start physical timestamp, the end physical timestamp, and the degree parameter of the non-uniform rational B-spline parametric equation. The spatiotemporal parameter fitting module generates a serialized file based on the data structure and outputs it to an external physical storage medium for compressed recording.
[0118] See attached document Figure 6 In a specific implementation, step S5 provided by the present invention includes the following steps:
[0119] S501, the joint parsing and playback module reads the serialized file from persistent storage, deserializes and reconstructs the non-uniform rational B-spline parametric equations and reduces the discrete topological event set in system memory, and restores the index association between the topological event anchor points and the corresponding discrete event loads. The joint parsing and playback module initializes the playback-end rendering clock, which is configured as a master-frequency driven incrementing timer. The joint parsing and playback module extracts the current physical timestamp output by the playback-end rendering clock, maps the current physical timestamp to the normalized parameter space, and calculates the current interpolation parameters. The normalized interpolation mapping formula is:
[0120] ;
[0121] In the formula: The current interpolation parameter; The current physical timestamp; The physical timestamp of the start of the playback sequence; The physical timestamp for the end of the playback sequence.
[0122] S502, the singularity interception mechanism utilizes the mathematical derivative breakage characteristics caused by multiple nodes to force a synchronization interruption within the continuous rendering pipeline. The joint analysis and playback module performs spatial coordinate interpolation calculations along the non-uniform rational B-spline parametric equation based on the current interpolation parameters.
[0123] During the parsing step, the joint parsing playback module saves the previous interpolation parameter and calculates the current interpolation parameter, and retrieves all the heavy node parameters in the front-open-back-closed interval formed by the previous interpolation parameter and the current interpolation parameter; when there are one or more heavy node parameters in the interval, the number of node weights of each heavy node parameter in the node vector is read one by one in ascending order of parameter value;
[0124] When the number of node multiplicity of any node parameter is the same as the degree parameter value set by the non-uniform rational B-spline parametric equation, the underlying parser determines that the current interpolation parameter interval crosses and hits the corresponding derivative singular point. Then, it positions the current interpolation parameter pointer to the derivative singular point and triggers a logic interrupt through software instructions, forcibly suspending the playback end rendering clock.
[0125] S503 distinguishes between pure geometric error truncation and discrete operations with business semantics in the branch routing execution process. When the rendering clock is suspended at the playback end, the joint parsing playback module extracts the current interpolation parameters corresponding to the derivative singular points as the query key. Based on the index association, the index query is performed within the reduced discrete topology event set, and the branch routing determination logic is executed.
[0126] When a query finds that the current interpolation parameter is associated with a specific discrete event payload, it determines that the current interpolation parameter belongs to the topology event anchor point. The joint parsing and playback module pushes the discrete event payload into the main rendering thread, executes the corresponding discrete topology state synchronization instruction, and completes the change of the design component attribute state.
[0127] When no discrete event load is found, the current interpolation parameter is determined to be a geometric split anchor point. The joint analysis playback module executes the no-operation skip rule for the geometric split anchor point. The no-operation skip rule is configured to skip the discrete state synchronous rendering pipeline, without making any attribute data modification, and directly step the parameter pointer to cross the current derivative singularity.
[0128] S504: After the discrete topology state synchronization instruction is executed and the state change is confirmed to be effective, along with the geometric splitting anchor point triggering no-operation skip, the joint analysis playback module performs the interrupt recovery operation. The joint analysis playback module releases the playback end rendering clock from the suspended state. The joint analysis playback module crosses the derivative singularity point, extracts the node vector fragment corresponding to the next adjacent parameter subdomain, restores the smooth interpolation step calculation of the non-uniform rational B-spline parametric equation, and the analysis playback process continues to loop recursively until the playback end rendering clock reaches the playback sequence termination physical timestamp, completing the high-fidelity reproduction of all collaborative design records.
[0129] See attached document Figure 7 To aid in understanding the technical solution of this invention, an application example of three-dimensional mechanical component collaborative design recording compression and playback is provided below.
[0130] The system receives concurrent collaborative operation data streams, extracts and records continuous operation sequences of spatial translation trajectories of three-dimensional mechanical parts. The starting physical timestamp of the continuous operation sequence constitutes the starting point of the time domain, and the ending physical timestamp of the continuous operation sequence constitutes the ending point of the time domain. Within the closed interval of the time domain, there are initial discrete state change events and termination discrete state change events. The operation attribute configuration of the initial discrete state change event is assigned to a temporary material, and the operation attribute configuration of the termination discrete state change event is assigned to the final material.
[0131] When the initial discrete state change event and the final discrete state change event perform consecutive assignment actions on the same discrete state attribute of the same three-dimensional mechanical component, the system establishes the initial discrete state change event and the final discrete state change event as a covering operation subgraph. The system extracts the intermediate assignment transition state vertices associated with the initial discrete state change event, retains the final discrete state change event, and generates a reduced discrete topology event set.
[0132] The system reads the physical timestamp of the termination discrete state change event and substitutes it into the normalized interpolation mapping formula to calculate and output the initial mapping parameters. If the initial mapping parameters are more than the minimum resolution threshold of the parameter domain and there is no collision, the system establishes the initial mapping parameters as the topology event anchor points.
[0133] The system constructs a non-uniform rational B-spline parametric equation and sets the degree parameter. The system injects topological event anchor points into the node vector and performs copying and superposition to generate heavy nodes. In the heavy node injection completion step, the system calculates the shortest distance from the original sampling points inside the discrete spatiotemporal point cloud to the curve represented by the non-uniform rational B-spline parametric equation and extracts the maximum fitting residual value.
[0134] The system compares the maximum fitting residual with the set tolerance threshold and determines that it is greater than the tolerance threshold. The system extracts the original sampling point time parameters associated with the maximum fitting residual and establishes the original sampling point time parameters as geometric split anchor points. The system injects the geometric split anchor points into the node vector according to the heavy node superposition mechanism and triggers the equation splitting and reconstruction calculation until all fitting residual values are no greater than the tolerance threshold.
[0135] During the playback phase, the system performs step-by-step parametric equation interpolation. If the current interpolation parameter value is consistent with the topology event anchor point parameter value, it determines that the number of node repetitions is consistent with the number parameter value and determines that the derivative singularity point is hit. The system triggers a logic interruption and queries and matches within the reduced discrete topology event set to assign the final material command. The system executes the discrete topology state synchronization command and resumes smooth interpolation step-by-step calculation in the completed state.
[0136] If the current interpolation parameter value is consistent with the geometric split anchor point parameter value, and the derivative singularity is hit again and the system does not query the matching discrete event load, the system determines that the current interpolation parameter belongs to the geometric split anchor point and performs a no-operation skip rule to directly step the parameter pointer past the derivative singularity to complete the analytical pipeline connection. In this way, while ensuring the synchronization and fidelity of discrete action states in the collaborative design process, the storage occupancy of continuous trajectory data is greatly compressed, achieving high-fidelity compression and playback.
[0137] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A high-fidelity compression and playback method for collaborative design records based on behavior sequence fusion, characterized in that, The method includes the following steps: It receives concurrent collaborative operation data streams and divides them into discrete event sets and continuous event sets; The discrete event set is mapped to an operation-dependent topology graph, and transitional nodes are removed to generate a reduced discrete topology event set. Traverse the time domain of the continuous event set to retrieve the intersection events, extract the physical timestamps and map them to the normalized parameter domain, establish the topological event anchor points through time-series anti-collision offset processing, and establish the index association between the topological event anchor points and the corresponding discrete event payloads. The continuous event set is fitted to a non-uniform rational B-spline parametric equation, and the topological event anchor points are injected as heavy nodes into the node vector to induce geometric continuity degradation. Geometric split anchor points are injected based on the fitting residual. Persistent storage is used to reconstruct the data of the non-uniform rational B-spline parametric equation, the reduced discrete topological event set, and the index association; Interpolation analysis is performed along the non-uniform rational B-spline parametric equation. When the stepping reaches the derivative singularity point caused by the heavy node, the playback rendering clock is suspended. The discrete topology state synchronization instruction or no-operation skip rule is executed according to the anchor point type, and then smooth interpolation is resumed.
2. The method for high-fidelity compression and playback of collaborative design records based on behavior sequence fusion according to claim 1, characterized in that, The step of receiving concurrent collaborative operation data streams and dividing them into discrete event sets and continuous event sets includes: Extract the physical timestamp, operation attribute code, target object's globally unique identifier, and state change increment value carried within the original operation event stream; Establish a business logic configuration dictionary, and divide the global heterogeneous operation event stream into the discrete event set and the continuous event set by looking up and comparing the operation attribute codes; Aggregation and grouping processing is performed on the continuous event set, and the fixed time window parameter and the action termination event trigger sequence are combined to truncate and output an independent continuous operation sequence.
3. The method for high-fidelity compression and playback of collaborative design records based on behavior sequence fusion according to claim 1, characterized in that, The steps of mapping the discrete event set to an operation-dependent topology graph and removing transitional nodes to generate a reduced discrete topology event set include: Map the operation events contained in the discrete event set to a directed acyclic graph data structure, and initialize an operation-dependent topology graph containing a set of vertices and a set of directed edges; A depth-first traversal search algorithm is executed within the operation-dependent topology graph to compare the relationships between node state changes and detect redundant operation subgraphs in the topology network. Based on the location-based redundant operation subgraph, the topology folding processing instruction is executed to clean up the exploration paths that have not generated final system constraints and generate the reduced discrete topology event set.
4. The method for high-fidelity compression and playback of collaborative design records based on behavior sequence fusion according to claim 3, characterized in that, The steps for executing topology folding instructions on the location-based redundant operation subgraph include: For the inverse operation subgraph, if there is no external dependency relationship between the intermediate transition vertices on the corresponding path, delete all intermediate transition vertices and their corresponding directed edges on the corresponding path. For the subgraph of the covering operation, when there is no external dependency relationship between the intermediate assignment transition state vertex, the intermediate assignment transition state vertex is extracted, the initial in-degree event boundary at the front of the time axis and the final out-degree event boundary at the end of the time axis are retained, and a directed edge is established to directly jump from the initial in-degree event boundary to the final out-degree event boundary.
5. The method for high-fidelity compression and playback of collaborative design records based on behavior sequence fusion according to claim 1, characterized in that, The steps for traversing the temporal domain of the continuous event set to retrieve the intersection events include: Read the start and end physical timestamps contained in the continuous operation sequence to construct the time domain; Traverse the reduced discrete topological event set. When the physical time node of the discrete event is within the closed interval of the time domain, and the globally unique identifier of the target object of the discrete event is consistent with the globally unique identifier of the target object of the continuous operation sequence, the discrete event is established as an intersection event.
6. The method for high-fidelity compression and playback of collaborative design records based on behavior sequence fusion according to claim 1, characterized in that, The step of extracting physical timestamps, mapping them to the normalized parameter domain, and establishing topological event anchor points through temporal anti-collision offset processing includes: Calculate and output the initial mapping parameters. If the numerical distance between the initial mapping parameters and the parameter values stored in the set of registered anchor points is not greater than the set minimum resolution threshold of the parameter domain, a timing collision is determined to have occurred. Apply a temporal anti-collision offset equal to the minimum resolution threshold of the parameter domain to the initial mapping parameters, calculate the output offset adjustment parameters, and trigger a loop recursive verification instruction until the absolute value distance between the offset adjustment parameters and the parameter values stored in the registered anchor point set is greater than the minimum resolution threshold of the parameter domain. The offset adjustment parameter, which is recursively verified, is established as the topological event anchor point.
7. The method for high-fidelity compression and playback of collaborative design records based on behavior sequence fusion according to claim 1, characterized in that, The steps of fitting the continuous event set to a non-uniform rational B-spline parametric equation and injecting the topological event anchor points as heavy nodes into the node vector to induce a degradation in geometric continuity include: For continuous operation sequences that are represented as discrete spatiotemporal point clouds in the spatial dimension, the non-uniform rational B-spline parametric equations are constructed, which are jointly driven by the control point sequence, node vectors, and weight factors. The parameter values corresponding to the topological event anchor point are forcibly injected into the node vector according to the multi-node superposition mechanism, so that the final node multiplicity after injection at the topological event anchor point is consistent with the degree parameter of the non-uniform rational B-spline parameter equation. Update the control points on both sides of the topological event anchor point respectively, so that the curve remains positionally continuous at the topological event anchor point, without applying the first derivative continuity constraint to both sides of the anchor point, thus forming a first derivative discontinuity feature at the topological event anchor point.
8. The method for high-fidelity compression and playback of collaborative design records based on behavior sequence fusion according to claim 1, characterized in that, The step of injecting geometric splitting anchor points based on fitted residuals includes: The shortest distance from each original sampling point inside the discrete spatiotemporal point cloud to the curve characterizing the non-uniform rational B-spline parametric equation is calculated as the fitting residual value, and the maximum fitting residual value is extracted from all fitting residual values. When the maximum fitting residual is greater than the set tolerance threshold, the time parameter of the original sampling point corresponding to the maximum fitting residual is established as the geometric splitting anchor point. The geometric splitting anchor points are injected into the node vector according to the heavy node superposition mechanism, triggering the equation splitting and reconstruction recursive calculation instruction until all fitting residual values are no greater than the tolerance threshold.
9. The method for high-fidelity compression and playback of collaborative design records based on behavior sequence fusion according to claim 1, characterized in that, The steps of performing interpolation analysis along the non-uniform rational B-spline parametric equations and suspending the playback rendering clock when stepping to the derivative singularity point caused by the heavy node include: Initialize the playback end rendering clock, extract the current physical timestamp output by the playback end rendering clock, and calculate the current interpolation parameters; Spatial coordinate interpolation calculation is performed along the non-uniform rational B-spline parametric equation. The heavy node parameters in the front-open-back-closed interval formed by the previous interpolation parameter and the current interpolation parameter are retrieved, and the node multiplicity of each heavy node parameter in the node vector is read. When the number of node multiplicity of any node parameter is the same as the degree parameter value of the non-uniform rational B-spline parameter equation, the current interpolation parameter pointer is positioned at the derivative singular point, and a logic interrupt is triggered by a software instruction to forcibly suspend the playback end rendering clock.
10. The method for high-fidelity compression and playback of collaborative design records based on behavior sequence fusion according to claim 1, characterized in that, The steps of executing discrete topology state synchronization instructions or no-operation skip rules based on anchor point type include: Extract the current interpolation parameter corresponding to the derivative singularity point as the query key, and perform an index query based on the index association in the reduced discrete topology event set; When a specific discrete event payload is found to be associated, the current interpolation parameter is determined to belong to the topological event anchor point. The discrete event payload is then pushed into the main rendering thread, and the corresponding discrete topological state synchronization instruction is executed. When no discrete event load is found, the current interpolation parameter is determined to belong to the geometric split anchor point. The no-operation skip rule is executed to skip the discrete state synchronous rendering pipeline and directly step the parameter pointer past the current derivative singularity.
