A vector animation editing system supporting multi-user real-time collaboration

Abstract

The application belongs to the technical field of computers, and particularly relates to a vector animation editing system supporting real-time collaboration of multiple users. The system realizes operation conflict resolution, state strong consistency synchronization and low-delay rendering under high concurrency through multi-dimensional interactive perception, vector semantic conversion, distributed state collaboration, flow differential distribution and real-time acceleration rendering modules. The system effectively solves the problems of synchronization inconsistency, transmission redundancy and rendering lag by using vector clock ordering, viewport perception distribution, local damage redrawing and prediction compensation mechanism, and supports thousand-user-level collaborative editing and guarantees professional-level animation creation experience.

Description

Technical Field

[0001] This invention belongs to the field of computer technology, specifically relating to a vector animation editing system that supports real-time collaboration among multiple users. Background Technology

[0002] In the vast field of multimedia content creation and digital design, vector animation plays an irreplaceable role in scenarios such as web page interaction, mobile application interface design, and film and television animation production due to its advantages such as strong graphic expressiveness, low resource consumption, and support for arbitrary scaling. With the popularization of distributed office environments and the increasing demand for team collaboration, the creative mode is undergoing a transformation from isolated localized operations to highly integrated, real-time interactive cloud-based collaboration. This trend not only requires the underlying architecture to have powerful graphics rendering capabilities, but also needs to be built on the foundation of efficient data flow and multi-terminal synchronization mechanisms to support the extremely high requirements of modern digital production processes for response speed and creative flexibility.

[0003] Among these technologies, multi-user real-time collaborative vector animation editing is the core of efficient team creation. It aims to build an interactive environment that allows geographically dispersed participants to operate concurrently within the same creative sequence. This technology focuses on the real-time decomposition and distribution of complex vector path data, keyframe attributes, and layer cascading relationships, ensuring a high degree of consistency in the visual state and logical sequence displayed by all clients through millisecond-level instruction synchronization. In this process, the system not only handles basic graphics drawing interactions but also maintains precise alignment of the animation timeline across different network nodes, thereby achieving a seamless collaborative experience across regions and devices.

[0004] However, existing technologies face numerous challenges in achieving efficient, real-time collaborative vector editing. Traditional file-locking-based collaboration methods cannot meet the demands of instant interactive creation, leading to severe data contention and state conflicts among team members during concurrent operations. While existing systems attempt to introduce operation conversion or conflict resolution mechanisms, the parsing accuracy and semantic preservation capabilities of algorithms remain insufficient for the unique multi-layered nested structure and dynamic attribute relationships of vector animation, easily causing errors in graphic topology or confusion in timeline logic. Furthermore, the uncertainty of network latency often leads to mismatches between the preview states of individual terminals and the main server state. Coupled with the linear decline in system processing power under high-concurrency editing requests, large-scale animation rendering in multi-user environments often experiences stuttering or synchronization delays. These factors collectively result in user experience gaps and data inconsistencies during collaborative editing, becoming pressing technical challenges in the field of real-time vector animation editing. Summary of the Invention

[0005] The purpose of this invention is to provide a vector animation editing system that supports real-time collaboration among multiple users, in order to solve the core technical problems of existing technologies in large-scale concurrent editing environments, such as inconsistent vector data synchronization, frequent operation conflicts, significant real-time rendering delays under high load scenarios, and low data transmission efficiency in complex paths.

[0006] The technical solution of the present invention includes:

[0007] The multi-dimensional interactive sensing unit is used to monitor and capture the input device status of each collaborative terminal in real time, record physical operation events including click, drag, pressure touch, and rotation, and associate them with the system's absolute timestamp at the time of the operation to generate the original interaction sequence.

[0008] The vector semantic transformation engine receives the original interaction sequence output by the multi-dimensional interaction perception unit, and according to the built-in vector operator dictionary, parses the physical operation events into atomic operation operators with deterministic semantics, and extracts the unique identifier, attribute change vector, and coordinate transformation matrix of the affected vector object.

[0009] The distributed state coordination hub is used to build and maintain a global vector state tree based on hierarchical relationships. It receives atomic operation operators from different cooperating terminals and calculates the unique effective sequence of all operations on the global time axis by executing conflict resolution strategies and consistency verification logic based on causal relationships.

[0010] The streaming differential distribution node is used to extract the attribute difference value between the current state and the previous historical state according to the effective sequence determined by the distributed state coordination center, and distribute the differential data to all online collaborative terminals using streaming compression coding technology.

[0011] The real-time accelerated rendering module is used to decompress and restore atomic operation operators after receiving differential data from each collaborative terminal, call the graphics processor hardware acceleration interface to locally refresh the affected vector levels, and reconstruct a consistent animation visual appearance on each terminal.

[0012] As one embodiment of the present invention, the vector semantic transformation engine performs the following refinement steps when processing Bézier curve editing operations: extracting node coordinate information on the curve path, including the starting anchor point, the ending anchor point, and the relative displacement of the two control points in the two-dimensional plane; calculating the smoothness parameters of the path based on the tangent slope and curvature continuity constraints between nodes; encapsulating the above geometric parameters with the path's fill attributes, stroke width, and endpoint style to form a complete path transformation operator; during the encapsulation process, the engine performs fixed-point processing on the floating-point coordinates to ensure that different hardware platforms produce completely consistent geometric calculation results when executing the same operator.

[0013] Furthermore, the distributed state coordination hub employs a concurrent control algorithm based on vector clocks. When multiple concurrent modification requests for the same vector object are received, the coordination hub first checks the vector clock vector carried by each request to determine the order of the requests. For concurrent requests without a partial order relationship, the coordination hub calls a preset semantic priority matrix to sort them by weight according to the importance of the operation type and the user's administrative authority level. If the operation types are completely identical and the permissions are the same, a deterministic sorting logic based on the object hash value is introduced to ensure that these operations are applied in the same order on all replicas, thereby achieving a strongly consistent state transition.

[0014] Furthermore, the distributed state collaboration hub also includes a dynamic region locking subunit. This subunit divides the entire canvas into multiple logical spatial grids. When a collaborative terminal initiates editing of a specific vector object, the locking subunit immediately adds a temporary locking tag to the object and its smallest containing rectangle in the global state tree. During the lock's validity period, modification operators sent by other terminals targeting objects within the same region will be placed in a pending queue until the locking tag is released due to editing completion or heartbeat timeout. This effectively avoids frequent operation collisions within the microscopic space, improving the smoothness of collaborative editing.

[0015] In one embodiment of the present invention, the streaming differential distribution node executes an on-demand predictive distribution strategy. Internally, the node maintains a viewport state table for each collaborating terminal, recording in real-time the current viewing position and zoom level of each user on the canvas. When the global state tree changes, the distribution node prioritizes calculating changes to vector objects within the user's current viewport range and pushes them as high-priority differential packets. For changes to objects outside the viewport, the distribution node accumulates them in a background buffer and performs batch synchronization at a lower frequency. If a rapid movement of the user's viewport is detected, the distribution node predicts the area it will enter based on the movement trend and initiates streaming preloading of vector data for that area in advance.

[0016] Furthermore, the streaming differential distribution node employs an arithmetic coding algorithm based on attribute classification during the encoding process. It categorizes vector attributes into three main types: geometric attributes, visual attributes, and animation temporal attributes. For the statistical distribution characteristics of each attribute type, the coding probability model is dynamically adjusted. For motion parameters that are continuous in time series, second-order predictive coding is used, transmitting only the prediction residuals. This significantly compresses the transmission volume of animation keyframe data, enabling synchronization delays to be maintained within seconds even in narrowband network environments.

[0017] As one embodiment of the present invention, the real-time accelerated rendering module employs layered off-screen rendering and local damage detection technology. The rendering module divides vector animation into a background layer, a static layer, a dynamic editing layer, and an interactive feedback layer. Upon receiving differential synchronization data, the rendering module first locates the layer to which the affected vector object belongs and calculates the rectangle of the damaged area of ​​the object in the screen coordinate system. The rendering module only triggers pixel redrawing within the damaged area and uses a hardware mixing unit to synthesize the updated layer fragments with other unchanged off-screen caches. When rendering path shapes, the module uses a path pre-scanning algorithm to decompose complex Bézier paths into a series of basic triangular meshes and uses the vertex shader and pixel shader of the graphics processor for parallel filling, ensuring that even in complex scenes with tens of thousands of vector paths, a display refresh rate of 60 frames per second can still be maintained.

[0018] Furthermore, the real-time accelerated rendering module also integrates a prediction compensation rendering engine. When a momentary network fluctuation causes a delay in the arrival of differential data, the rendering engine uses existing historical motion vectors and operator semantics to predict the next frame state of the target object using linear or spline interpolation, and renders the predicted state in advance. When the actual differential data arrives, the engine corrects the predicted state to the actual state through a brief, smooth transition animation, thereby eliminating the stuttering caused by network fluctuations at the perceptual level.

[0019] As one embodiment of the present invention, it further includes a collaborative conflict backtracking management unit for maintaining a finite-length operation history stack in system memory. Each historical operation is associated with a complete state snapshot summary. When a user needs to perform an undo operation, the backtracking management unit does not simply reverse the execution of local operators, but requests undo authorization for that specific user from the distributed state collaboration center. The collaboration center replays all global operators from the undo point to the current time, removes the user's specific operation contribution, and generates a new compensation operator to distribute to each end. This backtracking mechanism based on global logs solves the logical confusion problem caused by traditional undo instructions in a multi-user interleaved operation environment.

[0020] Furthermore, the multi-dimensional interactive sensing unit supports feature extraction of multi-point touch trajectories. It uses a sliding window algorithm to denoise the original sampling points, identify higher-order interactive intentions such as closed wrapping, reciprocating swiping, and long-press selection, and encapsulates these intentions into macro operators. Macro operators occupy very little space during transmission, but at the receiving end, they can be expanded into a series of complex vector transformation instructions through preset macro scripts, further improving the transmission efficiency and response accuracy of interactive instructions.

[0021] In one embodiment of the present invention, the system establishes a peer-to-peer network topology based on logical regions during the initialization phase. When the number of online collaborative terminals exceeds a preset threshold of 128, the distributed state collaboration hub automatically divides the terminals into multiple collaboration clusters; each cluster elects a cluster head node responsible for handling minor conflicts and data aggregation within the cluster; cluster head nodes exchange large-scale states through the backbone hub; this hierarchical distributed management architecture enables the system to support the scalability of thousands of people simultaneously editing the same large-scale vector animation project online.

[0022] Furthermore, the vector semantic transformation engine also possesses graphical semantic understanding capabilities. When processing shape combinations or Boolean operations, the engine can identify the topological nesting relationships between the various closed paths involved in the operation and transform the complex Boolean operation process into a parameterized path generation formula. During synchronization, only the operation logic and the original path parameters are transmitted, rather than the complex result path after the operation. This not only ensures the mathematical accuracy of the path during scaling but also greatly reduces the transmission of redundant vertices.

[0023] Furthermore, the real-time accelerated rendering module employs a shader proxy-based rendering strategy when handling lighting and shadow filters and real-time blur effects. The system predefines a series of standardized vector filter operators, which are implemented by the rendering module calling locally optimized shader code on each device. This avoids frequent transmission of large-scale bitmap textures over the network, ensuring pixel-level consistency of filter effects across different devices.

[0024] As one embodiment of the present invention, the system also includes a resource consistency monitor. The monitor periodically calculates the global hash value of the local state tree of each cooperating terminal and compares it with the standard hash value maintained by the distributed state coordination center; if a state deviation is found in a terminal, the monitor will trigger a forced synchronization mechanism, and transmit the missing operator increment packet of the terminal in streaming mode, so that it returns to the globally consistent state in the shortest possible time.

[0025] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0026] 1. This invention fundamentally solves the conflict problem in high-concurrency editing by multiple users by constructing a distributed state collaboration framework based on atomic operation operators and vector clocks; by transforming complex physical interactions into operator streams with deterministic semantics and performing unified sorting and conflict resolution on the global timeline, it ensures that the vector animation state seen by all participants is highly consistent, greatly improving the determinism and reliability of team collaboration.

[0027] 2. This invention introduces streaming differential distribution and viewport-based on-demand synchronization mechanisms, which significantly reduces the dependence of collaborative editing on network bandwidth. By transmitting only attribute differential values ​​and using compression encoding optimized for vector features, the synchronization delay between each end can be controlled in the millisecond range even when carrying complex vector projects containing tens of thousands of nodes, providing users with an ultimate real-time experience that is close to local editing.

[0028] 3. This invention utilizes hardware-accelerated rendering, layered off-screen compositing, and predictive compensation algorithms to overcome the performance bottleneck of real-time synchronous display of large-scale vector animations; local damage detection technology reduces invalid redraws, while predictive compensation rendering effectively resists the negative visual impact of network jitter, ensuring the smoothness and silky feel of the animation preview, and meeting the stringent requirements of professional animation creation for real-time feedback.

[0029] 4. This invention optimizes the interaction logic in the collaboration process by introducing spatial grid locking and a global log-based undo / replay mechanism; dynamic region locking reduces mutual exclusion interference at the operation level, while the globally consistent undo mechanism solves the backtracking problem when multiple operations are intertwined, making the collaborative editing process more in line with human intuitive thinking and significantly improving the creation efficiency of large-scale projects.

[0030] 5. The hierarchical and clustered network architecture and resource consistency monitoring adopted in this invention endow the system with strong scalability and robustness; the system can automatically adjust the topology according to the user scale and has self-repair capability when occasional synchronization errors occur, supporting diverse application scenarios from small team discussions to large community collaborative creation, and has extremely high engineering practical value. Attached Figure Description

[0031] Figure 1 This is a schematic diagram of the overall technical solution architecture proposed in this invention;

[0032] Figure 2 This is a schematic diagram of the core principle framework of the distributed state coordination hub in this invention. Detailed Implementation

[0033] Example 1

[0034] Please refer to Figure 1 and Figure 2This invention provides a vector animation editing system that supports real-time collaboration among multiple users. Its core logic lies in constructing a dynamic interactive environment capable of operating across regions and terminals while maintaining strong consistency. The system's architecture employs a highly decoupled distributed structure with separate front-end and back-end processes. It replaces the original pixel-level data transmission with precisely defined atomic operation operator streams, fundamentally solving the bandwidth consumption and latency conflicts associated with large-scale vector data synchronization. In this embodiment, the system's overall operation begins with the accurate capture of user intent by each collaborating terminal. Subsequently, through standardized semantic conversion, global state coordination, efficient differential distribution, and hardware-accelerated rendering, it ultimately achieves millisecond-level seamless synchronous collaboration among multiple users on the same canvas.

[0035] The system comprises several core functional modules, including: a multi-dimensional interactive perception unit, a vector semantic transformation engine, a distributed state collaboration hub, streaming differential distribution nodes, and a real-time accelerated rendering module. These modules communicate with each other via a pre-defined high-speed data bus and asynchronous message queues, ensuring high reliability and real-time performance of data flow.

[0036] The multi-dimensional interaction sensing unit is the primary interface between the system and the physical world, and its main task is to achieve high-fidelity capture of user operation behavior. This unit is integrated into the underlying driver layer of each collaborative terminal, deeply interfacing with the operating system kernel's input subsystem to monitor and capture the input device status of each collaborative terminal in real time. In the specific execution logic, the multi-dimensional interaction sensing unit establishes a high-frequency sampling thread, with the sampling frequency set to 120 Hz to ensure the capture of even the most subtle changes in user strokes or clicks. The physical operation events recorded by this unit include, but are not limited to, clicks, drags, pressure touches, rotations, and multi-finger zooms. Each captured raw physical event is immediately encapsulated into a data structure and associated with the system's absolute timestamp at the time of the operation. This absolute timestamp is obtained through a synchronous network time protocol, with an accuracy down to the microsecond level, thus providing a unique time reference for subsequent causal relationship judgments by the distributed state coordination center. The encapsulated raw interaction sequence includes the device serial number, operation type code, raw coordinate offset, pressure sensing value, and timestamp information, which is pushed to the vector semantic conversion engine in real time.

[0037] To improve the transmission efficiency of interactive commands, the multi-dimensional interactive sensing unit also supports feature extraction of multi-touch trajectories. It utilizes a sliding window algorithm to denoise the original sampling points, with the sliding window length typically set to 5 to 10 sampling periods. By calculating the curvature changes and velocity vectors of the sampling points within the window, the unit can identify higher-order interactive intents such as closed loops, reciprocating swipes, and long-press selection, and encapsulate these complex intents into macro operators with a higher semantic level. These macro operators occupy minimal space during transmission, but at the receiving end, they can be expanded into a series of complex vector transformation commands using a pre-defined macro script.

[0038] The vector semantic transformation engine receives the raw interaction sequence output by the multi-dimensional interaction perception unit. Its core function is to map physical behaviors to mathematical expressions. Internally, the engine contains a built-in dictionary of vector operators, predefining all possible vector editing behaviors and their corresponding mathematical models. Upon receiving the raw interaction sequence, the engine first activates the parser to parse the physical operation events into atomic operation operators with deterministic semantics. For each parsed operator, the engine extracts the unique identifier of the affected vector object, the attribute change vector, and the coordinate transformation matrix.

[0039] When handling Bézier curve editing operations, the vector semantic transformation engine executes refined transformation logic. It first extracts the node coordinate information on the curve path, including the start anchor point, end anchor point, and the relative displacement of the two control points in the 2D plane. To ensure the smoothness of the path during scaling or deformation, the engine calculates the path's smoothness parameters based on the tangent slope and curvature continuity constraints between nodes. These geometric parameters are then encapsulated with the path's fill attributes, stroke width, and endpoint styles to form a complete path transformation operator. During this encapsulation process, the engine performs a crucial fixed-point conversion step: mapping all floating-point coordinates to a large integer space, for example, by multiplying by 65536 and rounding to convert it to a 32-bit fixed-point number. This mechanism ensures that different hardware platforms produce completely consistent geometric calculation results when executing the same operator, eliminating the risk of rendering inconsistencies caused by differences in floating-point precision.

[0040] Furthermore, the vector semantic transformation engine possesses graphical semantic understanding capabilities. When processing shape combinations or Boolean operations, the engine can identify the topological nesting relationships between the various closed paths involved in the operation and transform the complex Boolean operation process into a parameterized path generation formula. During synchronization, the system only transmits the computational logic encoding and the original path parameters, rather than the thousands of redundant vertices generated after the operation.

[0041] The distributed state coordination hub is the logical brain of the entire system, responsible for maintaining the consistency of the global state. It maintains a hierarchical global vector state tree on a cloud server or edge computing node. This state tree fully records the parent-child relationships, hierarchy depth, attribute states, and historical change records of all vector objects on the current canvas. The coordination hub receives atomic operation operators from different collaborating terminals and initiates concurrent control processes.

[0042] To achieve strong consistency in a multi-user, high-concurrency environment, the distributed state coordination hub employs a concurrency control algorithm based on vector clocks. The vector clocks record the degree of influence of each cooperating terminal on the global state and its historical logical order.

[0043] ;

[0044] The above formula describes the distributed state coordination center receiving data from the terminal. When processing operation messages, the logic for updating the local vector clock array is as follows. Representative subject A locally maintained counter for its own sequence of operations is incremented whenever it generates a new atomic operator. Representative subject Other subjects perceived Operation progress. When a vector clock is used. When the operation message arrives at the central hub, the central hub performs a maximum value operation to merge the logical times of different terminals, thereby establishing a partial order relationship between operations.

[0045] By implementing a causal conflict resolution strategy and consistency verification logic, the coordination hub can calculate a unique effective sequence of all operations on the global timeline. When multiple concurrent modification requests for the same vector object are received, the coordination hub first checks the vector clock carried by each request to determine the order of the requests. For concurrent requests without a partial order relationship, the coordination hub invokes a preset semantic priority matrix to sort them by weight according to the importance of the operation type and the user's administrative authority level. If the operation types are completely identical and the permissions are the same, a deterministic sorting logic based on the object hash value is introduced to ensure that these operations are applied in the same order on all replicas.

[0046] The distributed state coordination hub also includes a dynamic region locking subunit. This subunit divides the entire canvas into multiple logical spatial grids, typically 256 pixels by 256 pixels in size. When a collaborating terminal initiates editing of a specific vector object, the locking subunit immediately adds a temporary lock tag to the object and its smallest containing rectangle in the global state tree. During the lock's validity period, modification operators sent by other terminals targeting objects within the same region are placed in a pending queue. This effectively avoids frequent operation collisions within the microscopic space.

[0047] The streaming differential distribution node is responsible for pushing the processed state change information to each endpoint in real time. Based on the effective sequence determined by the distributed state coordination center, it extracts the attribute difference values ​​between the current state and the previous historical state. This process employs a highly optimized streaming compression coding technique.

[0048] The streaming differential distribution node executes an on-demand, predictive distribution strategy. Internally, each node maintains a viewport state table for each collaborating terminal, recording in real-time the user's current viewing position and zoom level on the canvas. When the global state tree changes, the distribution node prioritizes calculating changes to vector objects within the user's current viewport and pushes them as high-priority differential packets. For changes to objects outside the viewport, the distribution node accumulates them in a background buffer and performs batch synchronization at a lower frequency. If a rapid movement of the user's viewport is detected, the distribution node predicts the area it will enter based on the movement trend and initiates streaming preloading of vector data for that area in advance.

[0049] During the encoding process, the streaming differential distribution node employs an arithmetic coding algorithm based on attribute classification. It categorizes vector attributes into three main types: geometric attributes, visual attributes, and animation temporal attributes. The coding probability model is dynamically adjusted based on the statistical distribution characteristics of each attribute type. For motion parameters that are continuous in the time series, second-order predictive coding is used.

[0050]

[0051] The formula describes the compression logic of dynamic attributes by the streaming differential distribution nodes. Wherein, The attribute value represents the current moment. and These represent the historical attribute values ​​from the previous two time points. The second-order prediction residuals are calculated. The distribution node only transmits the residual value. Since the residual usually approaches 0 in smooth motion, this greatly compresses the transmission volume of animation keyframe data.

[0052] The real-time accelerated rendering module is integrated into each collaborative terminal and is responsible for converting received logical instructions into visual images. This module employs layered off-screen rendering and local damage detection technology. It divides vector animation into a background layer, a static layer, a dynamic editing layer, and an interactive feedback layer. Upon receiving differential synchronization data, the rendering module first decompresses and restores the atomic operation operators, locates the layer to which the affected vector object belongs, and calculates the rectangle of the damaged area of ​​the object in the screen coordinate system.

[0053] The rendering module only triggers pixel redrawing within the damaged area and uses a hardware blending unit to composite the updated layer fragments with other unchanged off-screen caches. When rendering path shapes, the module uses a path pre-scan algorithm to decompose complex Bézier paths into a series of basic triangular meshes and uses the graphics processor's vertex and pixel shaders for parallel filling.

[0054] The real-time accelerated rendering module also integrates a prediction-compensated rendering engine. When network jitter causes a delay in the arrival of differential data, the rendering engine interpolates and predicts the next frame's state of the target object based on existing historical motion vectors and operator semantics, and renders it in advance. When the actual differential data arrives, the engine corrects the predicted state to the actual state through a brief, smooth transition animation, thus eliminating screen stuttering.

[0055] Furthermore, the real-time accelerated rendering module employs a shader proxy-based rendering strategy when handling lighting filters and real-time blur effects. The system predefines a series of standardized vector filter operators, which are implemented by the rendering module calling locally optimized shader code on each end, thereby avoiding the transmission of large-scale bitmap textures over the network.

[0056] During the initialization phase, the system establishes a peer-to-peer network topology based on logical regions. When the number of online collaborative terminals exceeds a preset threshold of 128, the distributed state coordination hub automatically divides the terminals into multiple collaborative clusters. Within each cluster, a cluster head node is elected to handle minor conflicts and data aggregation within the cluster. Large-scale state exchanges between cluster head nodes are conducted through the backbone hub.

[0057] Finally, the system also includes a collaborative conflict backtracking management unit and a resource consistency monitor. The backtracking management unit maintains an operation history stack, allowing logically clear undo operations to be performed in a multi-user interleaved environment. The resource consistency monitor periodically calculates the global hash value of the local state tree and compares it with the cloud standard value; if a deviation is found, it forces synchronization.

[0058] Example 2

[0059] Based on Example 1, this example provides a more in-depth engineering description of the macro-operator encapsulation mechanism of the multi-dimensional interactive sensing unit and the dynamic region locking strategy of the distributed state coordination hub, aiming to further reduce the synchronization load in extremely complex operation scenarios and improve the accuracy of conflict avoidance.

[0060] When performing trajectory feature extraction, the multi-dimensional interactive perception unit integrates a feature extraction sub-engine, which constructs a dynamic coordinate vector sequence through a high-frequency sampling stream. To accurately identify higher-order interactive intentions such as closed loops and reciprocating swipes, the sub-engine employs a pattern recognition algorithm based on a Hidden Markov Model. During the sliding window sampling process, the sub-engine calculates the instantaneous velocity vector and centripetal acceleration of the trajectory in real time. When it detects that the velocity vector completes a continuous 360-degree directional shift within a short period and the Euclidean distance between the starting and ending points is less than a preset closure threshold, the sub-engine determines that the trajectory segment is a closed loop intention. At this point, the multi-dimensional interactive perception unit no longer distributes hundreds or thousands of original coordinate points, but instead encapsulates them into a macro-operator containing the center coordinates, average radius, rotation direction, and closure confidence. The binary message length of this macro-operator is typically no more than 64 bytes, compressing the volume by more than 90% compared to the original coordinate stream.

[0061] For reciprocating swipe gestures, the sub-engine determines the direction by monitoring the frequency of positive and negative transformations of the coordinate projection along the main axis. When more than three directional reversals are detected within 300 milliseconds, the system automatically identifies it as an erasing or smearing intent and encapsulates it into a macro operator that defines the bounding box of the influence range and the frequency amplitude. These macro operators, after entering the vector semantic conversion engine, are directly mapped to specific brush effects or path deformation instructions. The macro operator also carries an adaptive step size parameter, which allows the receiving end to dynamically adjust the path subdivision number when the macro operator expands based on its own graphics processor performance, thereby obtaining smooth visual feedback on devices with different performance levels.

[0062] In this embodiment, the dynamic region locking subunit of the distributed state collaboration hub is refined into a management engine with spatial indexing capabilities. The management engine maintains a quadtree spatial index based on an R-tree structure in memory, categorizing all vector objects on the canvas according to their geometric centroids. When a collaborative terminal initiates a selection operation on object A, the dynamic region locking subunit not only locks object A itself but also performs a spatial query to find all other vector objects that intersect with the smallest containing rectangle of object A. The subunit calculates a dynamically locked influence region based on the hierarchical depth and spatial distance of these objects.

[0063] The structure of the locked tag is designed as a composite field containing the owner identifier, locking level, locking start timestamp, and expiration duration. Locking levels are divided into shared locks and exclusive locks: when a user only views attributes or performs lightweight drag-and-drop, a shared lock is applied, allowing other users to edit across different attribute dimensions; when a user enters a deep modification mode such as path node editing or Boolean operations, an exclusive lock is applied, completely prohibiting any other terminal from writing to that area.

[0064] To address the potential deadlock issue caused by locking, the dynamic area locking subunit introduces a heartbeat detection and forced release mechanism. Each terminal holding a lock tag must send a sustain signal to the coordination hub every 500 milliseconds. If the hub does not receive a signal within two consecutive heartbeat cycles, it determines that the terminal's connection is abnormal or in a suspended state. The locking subunit will automatically roll back the terminal's unfinished operations and distribute a lock release notification globally, reactivating operators in the suspended queue. This locking mechanism based on spatial grids and dynamic thresholds ensures that multiple users can maintain a high degree of operational continuity while collaborating in the same local area, without experiencing lag due to frequent locking conflicts.

[0065] Furthermore, the vector semantic transformation engine adds a path compression step based on fractal dimension when handling fixed-point transformations. Before converting the physical trajectory into a Bézier curve operator, the engine evaluates the trajectory's complexity. If the trajectory's local curvature changes smoothly, the engine uses the Douglas-Puk algorithm to thin out nodes, eliminating redundant intermediate anchor points. While preserving geometric features, this reduces the path operator load by an average of 40%. The fixed-point transformed data uses variable-length byte encoding, using single-byte storage for small offsets and 4-byte storage for large-scale transformations, further optimizing the transmission efficiency of the binary protocol.

[0066] Example 3

[0067] This embodiment focuses on describing the network topology evolution of the system under a large-scale user concurrency environment and the internal implementation details of the collaborative conflict backtracking management unit.

[0068] When the system detects that the number of currently online collaborative terminals exceeds the critical point of 128, the distributed state collaboration hub automatically initiates a topology reconstruction process, transforming from a flat star network to a hierarchical clustered architecture. Within each generated collaborative cluster, the system elects a cluster head node based on the terminal's computing performance, uplink bandwidth, and average network latency. The cluster head node possesses some of the functions of the collaboration hub, capable of handling minor editing conflicts between users within the cluster.

[0069] The conflict resolution process within a cluster is completed at the edge, without going through the main cloud hub. For example, when user A and user B in the same cluster simultaneously attempt to modify the fill color of the same vector, the cluster head node directly makes a decision based on the vector clock it receives locally, and uploads the decision result as a merge operator to the main hub. The main hub is only responsible for maintaining state synchronization between different clusters. This hierarchical distributed management architecture greatly alleviates the concurrent pressure on the central server, enabling the system to support thousands of users simultaneously editing the same large vector animation project online.

[0070] The collaborative conflict backtracking management unit, as a high-level consistency guarantee mechanism of the system, focuses on maintaining a non-linear operation history grid. Unlike the traditional linear undo stack, this history grid records the dependencies between each operator and other operators. Each historical operation is associated with a lightweight state snapshot digest, which contains the hash checksums of the affected objects before and after the operation.

[0071] When a user triggers an undo command, the rollback management unit does not perform a simple reverse operation. In a multi-user environment, the current state may have been modified by subsequent user actions, and a simple reverse operation would lead to logical overwriting or graph fragmentation. Instead, the rollback management unit submits an undo request to the distributed state coordination center. Upon receiving the request, the coordination center uses global log replay technology to re-execute all subsequent operators in a virtual replica, starting from the global version number corresponding to the undo point, but uniquely excludes the operator for which the user requested undoing.

[0072] By comparing the replayed state with the current true state, the coordination center calculates a compensation operator. The compensation operator is essentially a correction instruction that guides affected vector objects to a correct, consistent state. The compensation operator is assigned the highest distribution priority and synchronized to all online terminals via streaming differential distribution nodes. This backtracking mechanism based on global logs completely solves the logical chaos caused by undo commands in multi-user interleaved operation environments, ensuring that each user's undo action is executed accurately and consistently.

[0073] The resource consistency monitor runs in the background as a low-priority thread, periodically performing a depth-first traversal of the local state tree. During this traversal, it uses the Malek hash algorithm to generate local hash values ​​for each vector layer and combines these hash values ​​into a global state fingerprint. If the local monitor finds that its own fingerprint does not match the standard fingerprint published by the coordination hub for three consecutive synchronization cycles, it determines that a state deviation has occurred locally.

[0074] At this point, the monitor will not request a full synchronization, as the full data could be tens of megabytes. Instead, it will send an incremental request with a range of missing version numbers. Based on the request, the distributed state coordination center retrieves its maintained operator archive database and packages all missing atomic operation operators for that terminal into a single streaming incremental packet for forced synchronization. This incremental packet employs stronger error-correcting coding during distribution to ensure accurate delivery even in poor network conditions. Through this self-healing mechanism, the system guarantees that even if individual terminals experience sporadic data loss or software malfunctions, they can automatically return to a globally consistent state within seconds.

[0075] In this embodiment, the real-time accelerated rendering module further integrates a shader-based path pre-scanning algorithm. When rendering complex Boolean operation results or paths with tens of thousands of vertices, the rendering module utilizes the GPU's computational shaders to simultaneously calculate the path's monotonicity partitioning and trapezoidal coverage across thousands of parallel cores. This pre-scanning technique offloads the complex topology analysis tasks originally handled by the central processing unit (CPU) to the GPU, significantly reducing the geometric preparation time per frame. Combined with local damage detection, the system can maintain a refresh rate of 60 frames per second while reducing power consumption by more than 50%, which has significant practical value for collaborative editing scenarios on mobile devices.

[0076] The predictive compensation rendering engine employs a spline interpolation prediction model to handle network jitter. It records not only position but also the velocity, angular velocity, and scaling rate of each vector. By constructing a second-order differential equation of motion, the engine can predict the object's trajectory within the next 50 milliseconds. When the actual difference data arrives, if the deviation between the predicted and actual positions is within 10 pixels, the engine initiates a 150-millisecond smoothing correction animation, using a cubic Bézier curve to smoothly guide the predicted point to the actual point, making the network jitter virtually imperceptible to the user. This multi-layered synchronization and compensation mechanism collectively constitutes the system's ultimate real-time collaborative experience.

[0077] Example 4

[0078] In a vector animation editing system that supports real-time collaboration among multiple users, this embodiment provides a more in-depth engineering refinement of the path generation logic of the vector semantic conversion engine and the hierarchical management strategy of the real-time accelerated rendering module for real-time synchronization and efficient rendering of large-scale complex vector graphics.

[0079] The vector semantic transformation engine introduces a curvature-adaptive parametric compression algorithm when processing high-density vector paths. When a user draws a free path using a pressure-sensitive drawing device, the multi-dimensional interactive sensing unit generates a large number of raw sampling points. Instead of directly converting these points into straight line segments, the engine activates a real-time fitter. The fitter dynamically identifies straight line segments, arc segments, and general Bézier segments in the path by calculating the discrete curvature between consecutive sampling points.

[0080] For identified straight line segments, the engine retains only the coordinates of the start and end points; for circular arc segments, it converts them into the center point, radius, and start and end angles; for complex free curves, it uses the least squares method to find the optimal distribution of control points. During this transformation, the engine performs rigorous coordinate quantization, mapping the coordinate values ​​to a predefined grid system. This parameterized representation not only significantly reduces the amount of data that needs to be transmitted but also provides more semantic information for subsequent distributed state coordination. For example, when two users simultaneously attempt to smooth the same path, the coordination hub can directly perform mean averaging within the parameter space without dealing with massive vertex conflicts, significantly improving the mathematical accuracy of conflict resolution.

[0081] In this embodiment, the distributed state coordination hub enhances its logical robustness to undo and redo operations. Internally, the coordination hub maintains a global operation sequence log, which is persistently stored in a distributed in-memory database. Each log entry contains not only atomic operation operators but also a hash of the state snapshot before the operation. When the coordination conflict backtracking management unit initiates an undo request, the coordination hub utilizes differential state reconstruction technology to perform a partial rollback only on the affected specific node branches.

[0082] To avoid a chain reaction caused by undoing operations, the system introduces a conflict detection model based on dependency graphs. If user A's undo operation is a logical prerequisite for user B's subsequent operation—for example, user A creates a shape, and user B colors it—then when user A undoes the creation operation, the system automatically marks user B's coloring operation as invalid and notifies user B's terminal. This deep logical association management ensures that the logical integrity of the canvas is never compromised under high-intensity multi-user collaboration.

[0083] In this embodiment, the real-time accelerated rendering module employs a more advanced layer management and compositing strategy. In addition to the basic background, static, and dynamic layers, the rendering module introduces an off-screen texture cache pool. Based on the activity frequency of vector objects, the cache pool automatically pre-renders complex vector groups that have not changed for a long time as bitmap textures and stores them in video memory. When the user moves the viewport or performs simple non-destructive transformations, the system directly calls these pre-rendered textures for rapid compositing, without needing to recalculate the geometry of thousands of vectors.

[0084] Once the streaming differential distribution node sends modification instructions for these pre-rendered objects, the real-time accelerated rendering module immediately triggers damage detection, performing local redrawing only on invalid texture blocks. To balance memory usage and rendering performance, the module implements a cache replacement strategy based on the least recently used principle. For textures outside the current viewport that have been inactive for a long time, the system automatically releases them from video memory and performs asynchronous pre-rendering using a background thread the next time the viewport is accessed.

[0085] Furthermore, the streaming differential distribution node adds an adaptive synchronization mode based on bandwidth estimation for unstable network environments on mobile devices. The node monitors the round-trip latency and packet loss rate of each terminal in real time. When network bandwidth is sufficient, the system distributes a high-precision operator stream containing a complete geometric description; when network quality degrades, the system automatically switches to a low-precision mode, transmitting only critical path nodes and core attribute changes, and having the local rendering module automatically complete some minor animation transition effects based on semantics. This flexible synchronization mechanism ensures that users can still see consistent static results in weak network environments and automatically recover lost visual details after network recovery.

[0086] Through a series of in-depth optimizations and engineering implementations in data acquisition, semantic conversion, collaborative control, differential distribution, and hardware-accelerated rendering, this invention successfully constructs a vector animation collaborative editing platform capable of supporting thousands of users with low latency and strong consistency. The system not only solves the conflict problem under high concurrency but also breaks through the performance bottleneck of vector animation in real-time network synchronization through extreme differential compression and hardware rendering optimization, providing professional animation teams with an extremely efficient collaborative creation tool.

[0087] Example 5

[0088] This embodiment further explores the forced recovery process of the resource consistency monitor under extreme abnormal conditions, as well as the in-depth mining of complex gesture semantics by the multi-dimensional interactive perception unit.

[0089] In this embodiment, the resource consistency monitor is granted higher self-healing privileges. In addition to regular hash comparisons, the monitor also monitors the memory usage and graphics processor load of each terminal in real time. If a terminal is found to have a backlog of synchronization queues due to local resource exhaustion, the monitor will proactively trigger a degradation command. Under this command, the terminal will temporarily suspend complex local predictive compensation rendering and instead adopt a more robust, acknowledgment-based slow synchronization mode until the local buffer is cleared.

[0090] Following a catastrophic disconnection and reconnection, the resource consistency monitor executes a phased recovery strategy. First, it synchronizes global metadata, including layer structure, object lists, and their latest version numbers. Next, it synchronizes the state of visible objects within the current viewport. Finally, it silently synchronizes incremental operators outside the viewport in the background. This phased strategy ensures that users see the latest screen content and regain editing capabilities within the first second of reconnection, without waiting for massive amounts of historical data to fully load.

[0091] In this embodiment, the multi-dimensional interactive sensing unit is specifically optimized for operation with a professional stylus. It can capture the stylus's tilt angle, rotation angle, and pressure sensitivity change rate. This high-dimensional data is mapped in real-time to the dynamic width, transparency, and stroke texture offset of the vector path. To prevent the stress caused by high-dimensional data synchronization, the sensing unit introduces a threshold filtering algorithm based on feature changes. A new attribute change operator is only generated when the tilt angle change exceeds 2 degrees or the pressure sensitivity change exceeds 5%. This perception-based filtering technology reduces redundant sampling synchronization by approximately 60% while preserving stroke details.

[0092] Through this series of meticulous technical implementations, the system achieves exceptional interactive sensitivity and system stability while ensuring strong consistency, perfectly supporting the stringent requirements of large-scale vector animation creation for real-time performance, accuracy, and scalability. The technical solution proposed in this invention not only possesses significant technological advancements but also demonstrates high maturity and application potential in engineering implementation, enabling its widespread application in various digital scenarios such as online education, remote collaborative design, and large-scale community collaborative creation. The organic integration and closed-loop operation between modules ensure that the entire system can cope with extremely complex real-world network environments and high-intensity user interaction pressure, providing solid technical support for modern distributed collaborative work and creation.

[0093] In summary, this invention constructs a closed-loop, self-evolving multi-user real-time collaborative vector animation editing system through the collaborative work of a multi-dimensional interactive perception unit, a vector semantic conversion engine, a distributed state collaboration hub, streaming differential distribution nodes, a real-time accelerated rendering module, and various auxiliary units. The system not only overcomes the consistency challenge in high-concurrency environments but also achieves low bandwidth consumption and high frame rate display through extreme algorithm optimization, significantly improving collaboration efficiency and user experience, and possessing immense commercial value and broad market prospects.

Claims

1. A vector animation editing system supporting real-time collaboration among multiple users, characterized in that, include: The multi-dimensional interactive sensing unit is used to monitor and capture the input device status of each collaborative terminal in real time, record physical operation events including click, drag, pressure touch, and rotation, and associate them with the system's absolute timestamp at the time of the operation to generate the original interaction sequence. The vector semantic transformation engine is used to receive the original interaction sequence output by the multi-dimensional interaction perception unit, and according to the built-in vector operator dictionary, parse the physical operation events into atomic operation operators with deterministic semantics, and extract the unique identifier, attribute change vector and coordinate transformation matrix of the affected vector object. The distributed state coordination hub is used to build and maintain a global vector state tree based on hierarchical relationships. It receives atomic operation operators from different cooperating terminals and calculates the unique effective sequence of all operations on the global time axis by executing conflict resolution strategies and consistency verification logic based on causal relationships. The streaming differential distribution node is used to extract the attribute difference value between the current state and the previous historical state according to the effective sequence determined by the distributed state coordination center, and distribute the differential data to all online collaborative terminals using streaming compression coding technology. The real-time accelerated rendering module is used to decompress and restore atomic operation operators after receiving differential data from each collaborative terminal, call the graphics processor hardware acceleration interface to locally refresh the affected vector levels, and reconstruct a consistent animation visual appearance on each terminal.

2. The vector animation editing system supporting real-time multi-user collaboration according to claim 1, characterized in that, The vector semantic transformation engine is used to: extract node coordinate information on the curve path, including the starting anchor point, ending anchor point, and relative displacement of control points in the two-dimensional plane, when processing Bézier curve editing operations; calculate the smoothness parameter of the path based on the tangent slope and curvature continuity constraints between nodes; encapsulate the smoothness parameter with the path's fill attribute, stroke width, and endpoint style to form a complete path transformation operator; and perform fixed-point processing on the floating-point coordinates during the encapsulation process.

3. The vector animation editing system supporting real-time multi-user collaboration according to claim 1, characterized in that, The distributed state coordination hub is used to: employ a concurrent control algorithm based on vector clocks; when multiple concurrent modification requests for the same vector object are received, check the vector clock vector carried by each request to determine the order of the requests; For concurrent requests that do not have a partial order relationship, a preset semantic priority matrix is ​​invoked to sort them by weight according to the importance of the operation type and the user's administrative authority level; if the operation types are completely identical and the permissions are the same, a deterministic sorting logic based on object hash values ​​is introduced.

4. The vector animation editing system supporting real-time multi-user collaboration according to claim 1, characterized in that, The distributed state collaboration hub also includes a dynamic region locking subunit; the dynamic region locking subunit is used to: divide the entire canvas into multiple logical spatial grids; when a collaboration terminal initiates editing behavior on a specific vector object, it adds a temporary locking tag to the object and its smallest containing rectangle in the global state tree; During the lockout period, modification operators sent by other terminals targeting objects within the same area are placed in the pending queue.

5. The vector animation editing system supporting real-time multi-user collaboration according to claim 1, characterized in that, The streaming differential distribution node is used to: execute an on-demand predictive distribution strategy; and maintain a viewport status table for each collaborating terminal to record the current viewing position and zoom level of each user on the canvas in real time. When the global state tree changes, changes to vector objects within the user's current viewport are calculated first and pushed as high-priority differential packets; for changes to objects outside the viewport, they are accumulated in the background buffer for batch synchronization. If a rapid movement of the user's viewport is detected, the area to be entered is predicted based on the movement trend, and streaming preloading of the area vector data is started in advance.

6. The vector animation editing system supporting real-time multi-user collaboration according to claim 1, characterized in that, The streaming differential distribution node is used to: apply an arithmetic coding algorithm based on attribute classification during the encoding process; divide vector attributes into three categories: geometric attributes, visual attributes, and animation temporal attributes; dynamically adjust the coding probability model based on the statistical distribution characteristics of each type of attribute; and for motion parameters that are continuous in the time series, adopt second-order predictive coding and transmit only the prediction residual.

7. The vector animation editing system supporting real-time multi-user collaboration according to claim 1, characterized in that, The real-time accelerated rendering module is used to: employ layered off-screen rendering and local damage detection technology; divide vector animation into background layer, static layer, dynamic editing layer, and interactive feedback layer; upon receiving differential synchronization data, locate the layer to which the affected vector object belongs, and calculate the rectangle of the damaged area of ​​the object in the screen coordinate system; trigger pixel redrawing only within the damaged area, and use the hardware mixing unit to synthesize the updated layer fragment with other unchanged off-screen caches; when rendering path shapes, use a path pre-scanning algorithm to decompose complex Bézier paths into a series of basic triangular meshes, and use the vertex shader and pixel shader of the graphics processor for parallel filling.

8. The vector animation editing system supporting real-time collaboration among multiple users according to claim 1, characterized in that, The real-time accelerated rendering module also integrates a prediction compensation rendering engine; the prediction compensation rendering engine is used to: when the network environment experiences instantaneous jitter that causes the differential data to arrive late, interpolate and predict the next frame state of the target object based on the existing historical motion vectors and operator semantics, and render the predicted state in advance; when the actual differential data arrives, the predicted state is corrected to the actual state through a brief smooth transition animation.

9. The vector animation editing system supporting real-time collaboration among multiple users according to claim 1, characterized in that, It also includes a collaborative conflict backtracking management unit; the collaborative conflict backtracking management unit is used to: maintain a finite-length operation history stack in system memory, each historical operation is associated with a complete state snapshot summary; when a user needs to perform an undo operation, it requests undo authorization for that specific user from the distributed state collaboration center; the distributed state collaboration center removes the user's specific operation contribution by replaying all global operators from the undo point to the current time, and generates a new compensation operator to distribute to each end.

10. The vector animation editing system supporting real-time multi-user collaboration according to claim 1, characterized in that, It also includes a resource consistency monitor; the resource consistency monitor is used to: periodically calculate the global hash value of the local state tree of each cooperating terminal and compare it with the standard hash value maintained by the distributed state coordination center; If a terminal is found to have deviated from its state, a forced synchronization mechanism is triggered, which uses streaming to transmit the missing operator increment packets of that terminal to bring it back to a globally consistent state.

Images