An adaptive scheduling method and system for voice SMS processing

Abstract

This invention relates to the field of voice SMS processing and discloses an adaptive scheduling method and system for voice SMS processing. The method includes extracting the task type, resource requirements, execution time constraints, and potential causal relationships for each instruction; constructing a causal graph to form a task state constraint set; using reinforcement learning, Bayesian inference, and fuzzy logic to calculate the dynamic priority weight of each task and generate a dynamic weight matrix; jointly allocating CPU, bandwidth, power supply, and execution time window to generate a preliminary resource mapping scheme; real-time acquisition of the execution status, resource usage, and environmental parameters of the preliminary resource mapping scheme; combining task deviation constraints to limit the task execution range to generate a complete resource mapping scheme; and establishing a prediction model based on historical task execution data, multi-source environmental information, and the complete resource mapping scheme to calculate potential constraint conflicts and generate an adaptively updated scheduling optimization strategy. This invention has the advantage of improved reliability.

Description

Technical Field

[0001] This invention relates to the field of voice SMS processing, specifically to an adaptive scheduling voice SMS processing method and system. Background Technology

[0002] With the widespread application of in-vehicle systems, smart homes, mobile terminals, and industrial control systems, users and systems often issue multiple parallel commands simultaneously. For example, a driver might simultaneously issue a voice command to start navigation and play music, or a home control command to turn on a water heater and play audio. Operators might simultaneously trigger the start or stop of multiple industrial devices. However, existing technologies typically use fixed priority or sequential scheduling methods to handle multiple commands, which has the following shortcomings: Insufficient task dependency identification: It cannot effectively identify causal dependencies between tasks, potentially leading to delays in the execution of safety-critical tasks; Insufficient constraint awareness: It lacks awareness of user hard constraints or preferred tasks, causing low-priority but user-focused tasks to be ignored; Inappropriate resource allocation: Under conditions of limited CPU, bandwidth, power, and time slices, task execution may exhibit unreasonable phenomena; Lack of real-time adjustment capabilities: The lack of real-time monitoring and adjustment mechanisms for task execution status and potential conflicts makes it difficult to guarantee execution performance and user experience in complex multi-tasking environments. Therefore, existing technologies cannot simultaneously address the needs of prioritizing safety-critical tasks, satisfying user hard constraints, and achieving reasonable execution under limited resources in a multi-command concurrent environment. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides an adaptive scheduling method and system for voice SMS processing, which has the advantage of improved reliability and solves the problems mentioned in the background.

[0004] To achieve the aforementioned goal of improving reliability, the present invention provides the following technical solution: an adaptive scheduling method for processing voice SMS messages, comprising the following steps: Extract the task type, resource requirements, execution time constraints, and potential causal relationships of each instruction, construct a cause-effect graph, and form a task state constraint set by labeling the safety criticality, user hard constraints, and possible deviations from the execution range of each task; Based on causal graphs and task state constraint sets, reinforcement learning, Bayesian inference, and fuzzy logic are used to calculate the dynamic priority weight of each task, and task deviation constraints are introduced into the weight calculation to generate a dynamic weight matrix. Based on the dynamic weight matrix, CPU, bandwidth, power supply and execution time window are jointly allocated. Combined with cause-effect graph and task state constraint set, the task execution order is determined and a preliminary resource mapping scheme is generated. The system collects the execution status, resource usage, and environmental parameters of the preliminary resource mapping scheme in real time. When any indicator exceeds the preset threshold, it generates a complete resource mapping scheme by reordering tasks, adjusting time slices, and reallocating resources, combined with task deviation constraints to limit the task execution range. Based on historical task execution data, multi-source environmental information, and a complete resource mapping scheme, a prediction model is established to calculate potential constraint conflicts and map the results to task deviation constraints, thereby generating an adaptive and updated scheduling optimization strategy.

[0005] Preferably, the process of forming the task state constraint set is as follows: The system synchronously receives voice and SMS commands through multi-source command parsing, performs semantic recognition processing on the voice commands, and extracts key fields from the SMS commands. Causal nodes are generated based on task type, resource requirements, and execution time constraints, and causal edges are established for potential dependencies between tasks. At the same time, a multi-dimensional constraint information set is formed by combining security criticality and user hard constraints. By combining the causal nodes, causal edges, task types and resource requirements, safety criticality labels, user hard constraints, and deviation execution range contained in the multidimensional constraint information set, the allowed flexible execution range is set through task deviation constraints, and this information is integrated to output a structured task state constraint set.

[0006] Preferably, the process of calculating the dynamic priority weight of each task using reinforcement learning, Bayesian inference, and fuzzy logic is as follows: The causal graph node features and multidimensional constraint information set are used as inputs and imported into a reinforcement learning network to train and predict the trend of task priority changes over time. Based on this, combined with Bayesian inference methods, the uncertainty between tasks and the probability of potential execution conflicts are quantitatively assessed, and a risk adjustment factor is generated. Based on fuzzy logic, the task priority of reinforcement learning prediction is nonlinearly mapped, and the risk adjustment factor is integrated into the priority weight calculation to generate the dynamic priority weight of each task. Output the weight vectors for all tasks, including the dynamic priority weight value, corresponding risk adjustment factor, task identifier, and allowable deviation range for each task, forming a preliminary weight set.

[0007] Preferably, the process of generating the dynamic weight matrix is ​​as follows: For each task in the initial weight set, extract the dynamic priority weight value and the corresponding risk adjustment factor, and calculate the adjustable boundary of the weight in combination with the task's possible deviation range. During the weight adjustment process, a weighted superposition and normalization method is used to merge the initial priority weights with the deviation from the constraint boundary. A consistency check is performed on the fused weight vector to detect and correct conflicts or over-limit situations caused by deviation from constraints, and a dynamic weight matrix containing task dynamic weight values, risk adjustment factors and deviation constraint information is generated.

[0008] Preferably, the process of jointly allocating CPU, bandwidth, power supply, and execution time window is as follows: The dynamic weight matrix is ​​mapped to a multi-dimensional resource allocation space to extract the dynamic priority weight, risk adjustment factor and deviation constraint range of each task, and at the same time, their respective CPU, bandwidth, power requirements and execution time windows are calculated. Construct a multi-objective optimization function, incorporating task priority weights, deviation constraints, risk adjustment factors, and system resource availability into the constraints, and perform constraint optimization on the resource allocation for each task; Based on heuristic scheduling and iterative optimization methods, resources are allocated first to high-weight, hard-constraint tasks, while dynamic adjustments are made to tasks that are allowed to deviate. During the optimization process, the CPU, bandwidth, power allocation values ​​and executable time range of each task are recorded to generate a complete set of resource allocation results.

[0009] Preferably, the process of generating a preliminary resource mapping scheme is as follows: Based on the task dependencies in the cause-effect graph, the resource allocation results are topologically sorted to establish a preliminary task execution sequence; By combining the task status constraint set, dynamic weight set, and risk adjustment factor, the initial task sequence is fine-tuned in terms of priority, and non-critical tasks are allowed to be flexibly adjusted within the range of deviation from the constraints. During the sorting process, based on heuristic scheduling and iterative optimization methods, the possible time overlap, resource conflict and priority competition between tasks are dynamically balanced, and a task sequence that takes into account both resource allocation and execution order is output. The adjusted task sequence is integrated with the resource allocation results to generate a preliminary resource mapping scheme that includes CPU, bandwidth, power allocation, executable time range, and priority labeling for each task.

[0010] Preferably, the process of task reordering, time slice adjustment, and resource reallocation is as follows: Real-time collection of the execution status, resource usage, and environmental parameters of the preliminary resource mapping scheme, including CPU load, bandwidth utilization, power consumption, and time slice usage, to construct multi-dimensional state features; The multidimensional state features are compared with preset thresholds. When any indicator exceeds the threshold, the task reordering mechanism is triggered. The execution order of high-priority tasks is fine-tuned based on dynamic priority weights, deviation constraints, and risk adjustment factors. By combining time slice adjustment strategies and resource allocation optimization methods, the resource consumption of low-priority tasks is dynamically optimized. By combining an iterative optimization algorithm with a dynamic task mapping table, the adjusted task order, resource usage, and time slice information are synchronously calculated and corrected to form an optimized task state.

[0011] Preferably, the process of generating a complete resource mapping scheme is as follows: Based on the optimized task state, the task deviation constraints are mapped to an adjustable execution time and resource allocation range; For high-risk or critical tasks, deviations are limited based on priority gating mechanisms; For tasks that can be flexibly scheduled, an iterative optimization algorithm is used to dynamically adjust the allocation of CPU, bandwidth, power supply, and time slices; During the iteration process, resource overlap and time overlap are handled through conflict detection and correction mechanisms, and a complete resource mapping scheme containing the final execution scope of the task, CPU, bandwidth, power allocation and constraint verification information is output.

[0012] Preferably, the process of generating an adaptively updated scheduling optimization strategy is as follows: Input historical task execution data, current environmental parameters, and complete resource mapping scheme into the prediction model to perform time-series modeling and risk assessment of the execution conflict probability, resource consumption fluctuations, and potential bottlenecks for each task. Based on the prediction results, the conflicting tasks are mapped to the deviation from the constraint range, and the dynamic update coefficient of the adjustable range of the task is calculated. For high-priority or safety-critical tasks, the deviation should be strictly limited; For low-priority or flexibly schedulable tasks, iterative optimization algorithms are used to dynamically adjust resource allocation, time-slice scheduling, and execution order. Based on reinforcement learning strategies combined with risk adjustment factors and historical experience data, task priority, resource consumption, and time slice parameters are jointly optimized to generate an adaptively updated scheduling optimization strategy.

[0013] An adaptive scheduling voice SMS processing system includes: Causal analysis module: Extracts task type, resource requirements, execution time and potential causal relationships, constructs causal graph and marks safety criticality, hard constraints and deviation range, forming a task state constraint set; Weight calculation module: Based on the causal graph and task state constraint set, it uses reinforcement learning, Bayesian inference and fuzzy logic to calculate the dynamic priority weight of the task and generate a dynamic weight matrix; Resource allocation module: It jointly allocates CPU, bandwidth, power and time window according to dynamic weight matrix, and determines the task execution order in combination with constraint set to generate resource mapping scheme; Status monitoring module: Real-time collection of task execution status, resource usage and environmental parameters. When the indicators exceed the threshold, task reordering, time slice adjustment and resource reallocation are performed to generate a complete resource mapping scheme. Strategy optimization module: Based on historical task execution data, multi-source environmental information and complete resource mapping scheme, establish a predictive model, calculate potential constraint conflicts, and generate adaptive and updated scheduling optimization strategies.

[0014] Compared with the prior art, the present invention provides an adaptive scheduling method and system for voice SMS processing, which has the following beneficial effects: This invention achieves full-process analysis and management of voice SMS processing tasks by constructing a causal graph and combining it with a task state constraint set. It can comprehensively evaluate the type, resource requirements, execution time, and potential causal relationships of each task, while considering the task's safety criticality, hard constraints, and deviations from the execution range to ensure efficient and safe task execution. Through reinforcement learning, Bayesian inference, and fuzzy logic calculation of dynamic priority weights, and the generation of a dynamic weight matrix, it achieves adaptive priority allocation for multiple tasks. It can rationally allocate various resources such as CPU, bandwidth, power supply, and execution time windows. Based on a preliminary resource mapping scheme, it collects task execution status in real time... By analyzing resource usage and environmental parameters, and through task reordering, time slice adjustment, and resource reallocation, combined with deviation constraints to limit the task execution range, a complete resource mapping scheme is generated to achieve closed-loop adaptive scheduling of the system. At the same time, a predictive model is established by using historical execution data, multi-source environmental information, and the complete resource mapping scheme to calculate potential constraint conflicts and dynamically map them to task deviation constraints, forming an adaptive and updated scheduling optimization strategy. This significantly improves the execution efficiency, resource utilization, and task reliability of multi-task scheduling, reduces the risks caused by resource conflicts or task deviations, and achieves efficient, flexible, and intelligent scheduling capabilities that are difficult to achieve with traditional static scheduling methods. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the method of the present invention; Figure 2 This is a schematic diagram of the structure of the present invention. Detailed Implementation

[0016] The technical solutions of 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.

[0017] Example 1: Please refer to Figure 1As shown in the figure, an adaptive scheduling voice SMS processing method according to an embodiment of the present invention includes the following steps: S1: Extract the task type, resource requirements, execution time constraints, and potential causal relationships for each instruction, construct a cause-effect graph, and form a task state constraint set by labeling the safety criticality, user hard constraints, and possible deviations from the execution range for each task.

[0018] The process of forming the task state constraint set in S1 is as follows: The system synchronously receives voice and SMS commands through multi-source command parsing, performs semantic recognition processing on the voice commands, and extracts key fields from the SMS commands. The interface module simultaneously receives voice commands and SMS commands from in-vehicle, home, mobile terminal, or industrial systems. For voice commands, the voice signal is first preprocessed, including noise suppression, endpoint detection, and feature extraction, such as MFCC or time-spectral features. Then, a trained semantic recognition model parses the voice content into structured task information, such as task type, resource requirements, and execution time constraints. For SMS commands, text standardization is performed to extract key fields, such as task number, operation type, priority, and execution window, and convert them into the same task feature format as the voice commands. Through this process, multi-source commands are uniformly represented as a task feature dataset, providing input for causal modeling.

[0019] Causal nodes are generated based on task type, resource requirements, and execution time constraints, and causal edges are established for potential dependencies between tasks. At the same time, a multi-dimensional constraint information set is formed by combining security criticality and user hard constraints. Based on the task feature dataset, corresponding causal nodes are generated for each task, and causal edges are constructed according to the potential dependencies between tasks. Dependencies may include constraints such as the requirement for prerequisite tasks to be completed, resource conflicts, and overlapping time windows. Furthermore, each task is marked with its safety criticality. For example, high-risk operations must be executed in a strict order, while low-risk tasks can be flexibly scheduled. At the same time, user hard constraints are marked, such as inviolable time nodes or resource restrictions. By integrating causal nodes, causal edges, task types, resource requirements, safety criticality, and user hard constraints, a multi-dimensional constraint information set is formed, providing a comprehensive reference for task execution constraints.

[0020] By combining the causal nodes, causal edges, task types and resource requirements, safety criticality labels, user hard constraints and deviation execution range contained in the multidimensional constraint information set, the allowed flexible execution range is set through task deviation constraints, and this information is integrated to output a structured task state constraint set. Based on a multi-dimensional constraint information set, the possible deviation range of each task's execution is analyzed, including time advancement or delay, resource allocation redundancy, and alternative execution sequences. By setting allowable flexible execution intervals through deviation constraints, and assigning priority weighting strategies to different tasks, safety-critical tasks or tasks with hard constraints are ensured to be executed first, while maximizing overall scheduling flexibility. Deviation constraints can be implemented by setting upper and lower limits, fault tolerance coefficients, or dynamically adjusting weights, and are associated with causal graph nodes and edges to ensure that task execution paths are adjustable within the constraint range. Causal nodes, causal edges, task types, resource requirements, safety criticality, user hard constraints, and deviation constraint information are integrated into a unified data structure, forming a structured task state constraint set. This constraint set uses a computable data format, such as JSON or a graph database structure, facilitating access and invocation by automatic scheduling algorithms to achieve dynamic scheduling optimization.

[0021] S2: Based on the causal graph and task state constraint set, reinforcement learning, Bayesian inference and fuzzy logic are used to calculate the dynamic priority weight of each task, and task deviation constraints are introduced in the weight calculation to generate a dynamic weight matrix.

[0022] The process of calculating the dynamic priority weight of each task in S2 using reinforcement learning, Bayesian inference, and fuzzy logic is as follows: The causal graph node features and multidimensional constraint information set are used as inputs and imported into a reinforcement learning network to train and predict the trend of task priority changes over time. The features of the causal graph nodes, including task type, resource requirements, execution time constraints, safety criticality, and other multi-dimensional constraint information, are used as input data and imported into a reinforcement learning network. The reinforcement learning network can adopt a policy network or Q-value network structure. By interacting with the environment to simulate the task execution process, the network is trained to study the trend of task priority changes over time. During training, the reward function can be set to indicators such as the successful completion of tasks according to safety criticality and user hard constraints, optimal resource utilization, and minimization of overall scheduling latency. After training, the reinforcement learning network can predict future changes in task priority, providing a preliminary basis for weight adjustment.

[0023] Based on this, combined with Bayesian inference methods, the uncertainty between tasks and the probability of potential execution conflicts are quantitatively assessed, and a risk adjustment factor is generated. Based on reinforcement learning prediction, Bayesian inference is used to quantify the potential execution conflicts and uncertainties among tasks. Through Bayesian networks or probabilistic graphical models, constraints such as task dependencies, resource conflicts, and execution time overlaps are transformed into conditional probabilities. The probability of conflict or failure of tasks under different scheduling conditions is calculated, and risk adjustment factors are generated based on the probability calculation results. This can quantify the risks that tasks may face during execution and provide supplementary information for dynamic priority weight calculation.

[0024] Based on fuzzy logic, the task priority of reinforcement learning prediction is nonlinearly mapped, and the risk adjustment factor is integrated into the priority weight calculation to generate the dynamic priority weight of each task. By using fuzzy logic to perform nonlinear mapping on the task priorities predicted by reinforcement learning, the continuous and nonlinear characteristics of task priorities in high, medium, and low fuzzy intervals are handled. Risk adjustment factors are integrated into the nonlinear mapping results to achieve dynamic adjustment of task priority weights. Weighted superposition or fuzzy inference rules are adopted. For example, if a task has high risk and high priority, the weight is increased to the range X to ensure a balance between safety and flexibility. Dynamic priority weights are generated for each task, which can reflect the real-time risk and scheduling priority changes of task execution.

[0025] Output the weight vector of all tasks, including the dynamic priority weight value of each task, the corresponding risk adjustment factor, the task identifier and the deviation range of execution, to form a preliminary weight set; Each task's dynamic priority weight, corresponding risk adjustment factor, task identifier, and allowable deviation range are integrated to form a preliminary weight vector set. This set uses a computable data format, such as JSON, matrix, or graph database structure, for resource allocation, scheduling optimization, and task execution order planning. The output preliminary weight set supports real-time updates and can be dynamically adjusted according to task execution status, environmental changes, and constraints to achieve safe, efficient, and adaptive task scheduling management.

[0026] The process of generating the dynamic weight matrix in S2 is as follows: For each task in the initial weight set, extract its dynamic priority weight value and corresponding risk adjustment factor, and calculate the adjustable boundary of the weight in combination with the task's deviation range. From the initial weight set, the dynamic priority weight value and corresponding risk adjustment factor of each task are extracted one by one. Combined with the deviation range of each task's execution, such as the allowed time to be advanced or delayed, resource allocation redundancy and execution order flexibility, the adjustable boundary of the task weight is calculated, that is, the maximum and minimum value range of the weight under the constraints of safety and flexibility. The adjustment range of dynamic weight is established for each task, providing a basis for fusion and matrix generation.

[0027] During the weight adjustment process, a weighted superposition and normalization method is used to merge the initial priority weights with the deviation from the constraint boundary. The initial priority weight value of each task is fused with the deviation constraint boundary. The fusion process can adopt a weighted superposition method, that is, different weight coefficients are assigned to the initial weight and boundary weight according to the safety criticality and flexibility requirements of the task. At the same time, through normalization processing, the fusion result is uniformly mapped to a predetermined weight range, such as 0 to 1, to ensure that the task priority weights maintain the overall weight distribution balance while satisfying the deviation constraints, and to provide calculable and adjustable data for matrix generation.

[0028] A consistency check is performed on the fused weight vector to detect and correct conflicts or over-limit situations caused by deviation from constraints, and a dynamic weight matrix containing task dynamic weight values, risk adjustment factors and deviation constraint information is generated. For the fused weight vector, a consistency check is performed to detect weight conflicts, exceeding limits, or sorting anomalies that may be caused by deviations from constraints or boundary restrictions. For example, if the weight of a task exceeds the maximum allowable value due to boundary restrictions, it is corrected to the maximum boundary value. If multiple tasks have sorting conflicts due to weight adjustments, they are resolved through local reordering or weight fine-tuning. This can be achieved through a rule engine or constraint optimization algorithm to ensure that the dynamic weight of each task is reasonably distributed within a safe and executable range. The verified task dynamic weight values, corresponding risk adjustment factors, and deviation constraint information are integrated to generate a two-dimensional dynamic weight matrix. The rows of the matrix represent task numbers, and the columns represent task weights, risk factors, and deviation constraint information. This matrix can be directly used for resource allocation, execution order optimization, and real-time updates in scheduling algorithms. The output dynamic weight matrix supports incremental updates as the task execution status changes, ensuring that the scheduling system can dynamically adapt to the task execution environment and achieve safe, efficient, and adaptive task management.

[0029] S3: Based on the dynamic weight matrix, CPU, bandwidth, power supply and execution time window are jointly allocated. Combined with the cause-effect graph and task state constraint set, the task execution order is determined, and a preliminary resource mapping scheme is generated.

[0030] The process of jointly allocating CPU, bandwidth, power, and execution time window in S3 is as follows: The dynamic weight matrix is ​​mapped to a multi-dimensional resource allocation space to extract the dynamic priority weight, risk adjustment factor and deviation constraint range of each task, and at the same time, their respective CPU, bandwidth, power requirements and execution time windows are calculated. The dynamic weight matrix is ​​mapped to a multi-dimensional resource allocation space. Each task's corresponding row contains a dynamic priority weight value, a risk adjustment factor, and a deviation range from the execution range. At the same time, various resource requirements of the task are extracted, including CPU cycle requirements, bandwidth usage, power consumption, and execution time window. Through this mapping, each task forms a computable resource requirement vector in the resource allocation space, providing a basic data structure for joint allocation and optimization.

[0031] Construct a multi-objective optimization function, incorporating task priority weights, deviation constraints, risk adjustment factors, and system resource availability into the constraints, and perform constraint optimization on the resource allocation for each task; Based on the resource demand vector, a multi-objective optimization function is constructed. The optimization objectives include: maximizing the execution priority of high-weight tasks, minimizing resource conflicts, reducing the impact of risk adjustment factors, and ensuring optimal system resource utilization. The constraints include task deviation from the constraint range, risk adjustment factor limits, and availability of various system resources such as total CPU, bandwidth limit, and power supply capacity. This function transforms the resource allocation problem into a constrained optimization problem, providing a mathematical basis for heuristic or iterative optimization algorithms.

[0032] Based on heuristic scheduling and iterative optimization methods, resources are allocated first to high-weight, hard-constraint tasks, while dynamic adjustments are made to tasks that are allowed to deviate. A heuristic scheduling method combined with iterative optimization is used to allocate resources for tasks. Based on the dynamic weight matrix, CPU, bandwidth, power and execution time windows are allocated to high-weight tasks and hard-constrained tasks first to ensure that critical tasks meet the constraints. For tasks that are allowed to deviate, dynamic resource adjustment is performed according to their deviation range and risk adjustment factor. The overall resource allocation is optimized through iteration, so that the system can improve resource utilization efficiency while ensuring safety and priority. During the iteration process, local search, genetic algorithm or constraint optimization solver can be used to achieve an approximate optimal allocation scheme.

[0033] During the optimization process, the CPU, bandwidth, power allocation values ​​and executable time range of each task are recorded to generate a complete set of resource allocation results. During the resource optimization iteration process, the CPU allocation value, bandwidth allocation amount, power allocation value, and executable time range of each task are recorded in real time. This information is integrated into a dynamic resource allocation result set, using a computable data structure such as a table, matrix, or JSON format. This facilitates real-time retrieval, dynamic updates, and conflict correction during optimization iterations. The final resource allocation value of each task is confirmed, and these final values ​​are updated to the same set, thereby generating a complete resource allocation result set. This set includes the final CPU, bandwidth, power allocation amount, and executable time range of each task, ensuring that it can be directly used for task scheduling, execution monitoring, and dynamic adjustment. Through this dynamic recording and final confirmation mechanism, the scheduling system can achieve task execution status monitoring, resource allocation visualization, and multi-task joint management, thereby supporting safe, efficient, and adaptive task scheduling and resource management.

[0034] The process of generating a preliminary resource mapping scheme in S3 is as follows: Based on the task dependencies in the cause-effect graph, the resource allocation results are topologically sorted to establish a preliminary task execution sequence; Using the generated resource allocation result set and causal graph as input, the tasks are topologically sorted according to the dependency relationships between task nodes in the causal graph. During the topological sorting process, it is ensured that all tasks with pre-dependent relationships are logically executed before subsequent tasks, thereby establishing a preliminary task execution sequence that satisfies causal constraints. This ensures that the task execution order structurally conforms to the dependency relationships between tasks, providing a basis for adjustments based on priority and resource constraints.

[0035] By combining the task status constraint set, dynamic weight set, and risk adjustment factor, the initial task sequence is fine-tuned in terms of priority, and non-critical tasks are allowed to be flexibly adjusted within the range of deviation from the constraints. Based on the initial task execution sequence, a task state constraint set, a dynamic weight set, and a risk adjustment factor are further introduced to fine-tune the priority of task execution order. For tasks with high safety criticality or hard user constraints, the stability of their execution order is maintained. For non-critical tasks, the execution order is allowed to move forward or backward within their allowable deviation range. This can improve scheduling flexibility and overall execution efficiency without breaking causal constraints.

[0036] During the sorting process, based on heuristic scheduling and iterative optimization methods, the possible time overlap, resource conflict and priority competition between tasks are dynamically balanced, and a task sequence that takes into account both resource allocation and execution order is output. During task sorting and priority adjustment, heuristic scheduling and iterative optimization methods are adopted to dynamically balance possible time overlaps, resource conflicts and priority competition. When multiple tasks are detected competing for CPU, bandwidth or power resources in the same time window, the task order or execution time window is iteratively adjusted according to dynamic weights, risk adjustment factors and deviation constraints until resource feasibility and constraints are met, and a task execution sequence that takes into account both resource availability and execution order rationality is output.

[0037] The adjusted task sequence is integrated with the resource allocation results to generate a preliminary resource mapping scheme that includes CPU, bandwidth, power allocation, executable time range and priority labeling for each task. The adjusted task execution sequence is integrated with the corresponding resource allocation results to generate a preliminary resource mapping scheme. This scheme includes the CPU allocation value, bandwidth allocation amount, power allocation value, executable time range and priority labeling information for each task. The preliminary resource mapping scheme is represented by a computable data structure and can be directly used for scheduling execution, real-time monitoring and further optimization and adjustment, providing a clear and practical execution reference for the dynamic scheduling system.

[0038] S4: Real-time collection of the execution status, resource usage, and environmental parameters of the preliminary resource mapping scheme. When any indicator exceeds the preset threshold, a complete resource mapping scheme is generated by reordering tasks, adjusting time slices, and reallocating resources, combined with task deviation constraints to limit the task execution range.

[0039] The process of task reordering, time slice adjustment, and resource reallocation in S4 is as follows: Real-time collection of the execution status, resource usage, and environmental parameters of the preliminary resource mapping scheme, including CPU load, bandwidth utilization, power consumption, and time slice usage, to construct multi-dimensional state features; During the execution of the initial resource mapping scheme, the execution status, resource usage, and environmental parameters of each task are collected in real time. The collected data includes, but is not limited to, CPU load, bandwidth utilization, power consumption level, and time slice usage. The collected data is then processed for time synchronization, anomaly filtering, and normalization, and integrated into a multi-dimensional state feature vector representing the current scheduling state, providing real-time input for reordering and optimization decisions.

[0040] The multidimensional state features are compared with preset thresholds. When any indicator exceeds the threshold, the task reordering mechanism is triggered. The constructed multidimensional state features are compared with a preset set of thresholds, which may include resource utilization limits, time slice over-utilization ratios, and energy consumption warning values. When any indicator exceeds the corresponding threshold, it is determined that there is a potential performance or risk problem in the current scheduling state, and a task reordering mechanism is triggered. This triggering process can be implemented through a rule engine or event detection module to ensure that scheduling adjustments are timely and targeted.

[0041] The execution order of high-priority tasks is fine-tuned based on dynamic priority weights, deviation constraints, and risk adjustment factors. During the task reordering process, the dynamic priority weight, deviation constraint range, and risk adjustment factor of each task are comprehensively considered to fine-tune the task execution order. For high-priority or safety-critical tasks, their position in the execution sequence is improved to ensure that they receive execution resources first. For non-critical tasks that are allowed to deviate, their execution order is adjusted within the deviation constraint range to alleviate resource pressure and improve overall scheduling efficiency.

[0042] By combining time slice adjustment strategies and resource allocation optimization methods, the resource consumption of low-priority tasks is dynamically optimized. Based on task reordering, and further combined with time slice adjustment strategies and resource allocation optimization methods, the CPU time slice, bandwidth usage, and power allocation of low-priority or high-resource-consuming tasks are dynamically optimized. Adjustment methods may include shortening the time slice length, delaying the execution time window, or reducing the resource allocation ratio, thereby achieving a dynamic balance of resource utilization without affecting the execution of critical tasks.

[0043] By combining iterative optimization algorithms with dynamic task mapping tables, the adjusted task order, resource consumption, and time slice information are synchronously calculated and corrected to form an optimized task state. Through iterative optimization algorithms, combined with dynamic task mapping tables, the adjusted task order, resource usage, and time slice allocation information are synchronously calculated and consistency corrected. If new conflicts or over-limit situations are detected during the correction process, iterative adjustments will continue until the task state constraints and resource feasibility requirements are met, ultimately forming an optimized task state, which is then used for scheduling execution and status monitoring to achieve continuous adaptive task management.

[0044] The process of generating a complete resource mapping scheme in S4 is as follows: Based on the optimized task state, the task deviation constraints are mapped to an adjustable execution time and resource allocation range; Using the optimized task status as input, the deviation constraint information corresponding to each task is extracted, and the deviation constraint is mapped to an adjustable execution time range and resource allocation range. During the mapping process, the allowable time advance or delay interval and the resource increase or decrease range of the task are associated with the current execution status to form an adjustable range for each task in terms of execution time, CPU, bandwidth, power and other resource dimensions, thereby providing clear boundaries for differentiated adjustments.

[0045] For high-risk or critical tasks, deviations are limited based on priority gating mechanisms; For high-risk or safety-critical tasks, a priority gating mechanism is introduced to limit the deviation range. Based on the dynamic priority weight of the task and the risk adjustment factor, a strict deviation threshold is set to prevent excessive adjustments in execution time or resource allocation for critical tasks, thereby ensuring the stability and security of critical task execution.

[0046] For tasks that can be flexibly scheduled, an iterative optimization algorithm is used to dynamically adjust the allocation of CPU, bandwidth, power supply, and time slices; For non-critical tasks that can deviate from the execution range, an iterative optimization algorithm is used to dynamically adjust the allocation of CPU, bandwidth, power and time slices. During the adjustment process, the current resource utilization, dynamic task weight and overall scheduling goal are comprehensively considered. Within the allowable deviation range, the task execution time and resource occupation are optimized and allocated to improve the overall resource utilization efficiency and scheduling flexibility of the system.

[0047] During the iteration process, resource overlap and time overlap are handled through conflict detection and correction mechanisms, and a complete resource mapping scheme containing the final execution scope of the task, CPU, bandwidth, power allocation and constraint verification information is output. During the iterative process of resource and time adjustment, a conflict detection and correction mechanism is used to identify potential resource and time overlaps between tasks in real time. Once a conflict is detected, the relevant tasks are corrected and adjusted according to causal constraints, priority rules, and deviation constraints until all constraints are met. The final output is a complete resource mapping scheme that includes the final execution time range, CPU allocation value, bandwidth allocation amount, power allocation value, and corresponding constraint verification information for each task. This scheme can be directly used for task execution control, status monitoring, and scheduling optimization.

[0048] S5: Based on historical task execution data, multi-source environmental information, and a complete resource mapping scheme, establish a prediction model, calculate potential constraint conflicts, map the results to task deviation constraints, and generate an adaptively updated scheduling optimization strategy.

[0049] The process of generating adaptive update scheduling optimization strategies in S5 is as follows: Input historical task execution data, current environmental parameters, and complete resource mapping scheme into the prediction model to perform time-series modeling and risk assessment of the execution conflict probability, resource consumption fluctuations, and potential bottlenecks for each task. Using historical task execution data, current environmental parameters, and a complete resource mapping scheme as input, the prediction model performs time-series modeling of the task execution process. The historical task execution data includes task completion time, resource usage trajectory, and conflict occurrence records. The current environmental parameters include CPU load, bandwidth utilization, and power reserve. Based on the above information, the prediction model conducts risk assessments on the execution conflict probability, resource usage fluctuation trend, and potential scheduling bottlenecks of each task, providing a predictive basis for scheduling strategy updates.

[0050] Based on the prediction results, the conflicting tasks are mapped to the deviation from the constraint range, and the dynamic update coefficient of the adjustable range of the task is calculated. Based on the prediction results, tasks with a high probability of conflict or risk of resource fluctuation are mapped to their corresponding deviation constraint range, and the dynamic update coefficient of the adjustable range of the task is calculated to characterize the flexibility of the task in terms of execution time, resource allocation and order adjustment. It can be dynamically adjusted with changes in predicted risk, thereby achieving adaptive update of deviation constraints.

[0051] For high-priority or safety-critical tasks, the deviation should be strictly limited; For high-priority or safety-critical tasks, their deviation range is strictly limited based on their dynamic priority weights and risk adjustment factors. By setting a minimum adjustment range or a fixed execution window, the stability of the execution order and resource allocation of such tasks is ensured during the scheduling optimization process, preventing the introduction of security risks or execution uncertainties due to frequent adjustments.

[0052] For low-priority or flexibly schedulable tasks, iterative optimization algorithms are used to dynamically adjust resource allocation, time-slice scheduling, and execution order. For low-priority tasks or tasks with flexible scheduling characteristics, an iterative optimization algorithm is used to dynamically adjust their resource allocation, time slice scheduling, and execution order. Taking into account dynamic update coefficients, current resource availability, and overall scheduling objectives, the task execution parameters are optimized within the allowable deviation range to alleviate predicted resource bottlenecks and conflict risks.

[0053] Based on the reinforcement learning strategy combined with risk adjustment factors and historical experience data, task priority, resource consumption and time slice parameters are jointly optimized to generate an adaptive and updated scheduling optimization strategy. Based on the above optimizations, a reinforcement learning strategy is introduced. The risk adjustment factor, historical experience data, and current prediction results are used as state inputs to jointly optimize task priority, resource consumption parameters, and time slice allocation. The reinforcement learning model updates the scheduling strategy through continuous interaction and feedback, enabling the system to adaptively adjust scheduling decisions according to different execution scenarios. Finally, an adaptively updated scheduling optimization strategy is generated and used for subsequent task scheduling and execution.

[0054] Example 2: Please refer to Figure 2 As shown, an adaptive scheduling voice SMS processing system includes: Causal analysis module: Extracts task type, resource requirements, execution time and potential causal relationships, constructs causal graph and marks safety criticality, hard constraints and deviation range, forming a task state constraint set; Weight calculation module: Based on the causal graph and task state constraint set, it uses reinforcement learning, Bayesian inference and fuzzy logic to calculate the dynamic priority weight of the task and generate a dynamic weight matrix; Resource allocation module: It jointly allocates CPU, bandwidth, power and time window according to dynamic weight matrix, and determines the task execution order in combination with constraint set to generate resource mapping scheme; Status monitoring module: Real-time collection of task execution status, resource usage and environmental parameters. When the indicators exceed the threshold, task reordering, time slice adjustment and resource reallocation are performed to generate a complete resource mapping scheme. Strategy optimization module: Based on historical task execution data, multi-source environmental information and complete resource mapping scheme, establish a prediction model, calculate potential constraint conflicts, and generate adaptive and updated scheduling optimization strategies.

[0055] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0056] 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 method for adaptively scheduling voice SMS processing, characterized in that, Includes the following steps: Extract the task type, resource requirements, execution time constraints, and potential causal relationships of each instruction, construct a cause-effect graph, and form a task state constraint set by labeling the safety criticality, user hard constraints, and possible deviations from the execution range of each task; Based on causal graphs and task state constraint sets, reinforcement learning, Bayesian inference, and fuzzy logic are used to calculate the dynamic priority weight of each task, and task deviation constraints are introduced into the weight calculation to generate a dynamic weight matrix. Based on the dynamic weight matrix, CPU, bandwidth, power supply and execution time window are jointly allocated. Combined with cause-effect graph and task state constraint set, the task execution order is determined and a preliminary resource mapping scheme is generated. The system collects the execution status, resource usage, and environmental parameters of the preliminary resource mapping scheme in real time. When any indicator exceeds the preset threshold, it generates a complete resource mapping scheme by reordering tasks, adjusting time slices, and reallocating resources, combined with task deviation constraints to limit the task execution range. Based on historical task execution data, multi-source environmental information, and a complete resource mapping scheme, a prediction model is established to calculate potential constraint conflicts and map the results to task deviation constraints, thereby generating an adaptive and updated scheduling optimization strategy.

2. The adaptive scheduling voice SMS processing method according to claim 1, characterized in that, The process of forming the task state constraint set is as follows: The system synchronously receives voice and SMS commands through multi-source command parsing, performs semantic recognition processing on the voice commands, and extracts key fields from the SMS commands. Causal nodes are generated based on task type, resource requirements, and execution time constraints, and causal edges are established for potential dependencies between tasks. At the same time, a multi-dimensional constraint information set is formed by combining security criticality and user hard constraints. By combining the causal nodes, causal edges, task types and resource requirements, safety criticality labels, user hard constraints, and deviation execution range contained in the multidimensional constraint information set, the allowed flexible execution range is set through task deviation constraints, and this information is integrated to output a structured task state constraint set.

3. The adaptive scheduling voice SMS processing method according to claim 2, characterized in that, The process of calculating the dynamic priority weights of each task using reinforcement learning, Bayesian inference, and fuzzy logic is as follows: The causal graph node features and multidimensional constraint information set are used as inputs and imported into a reinforcement learning network to train and predict the trend of task priority changes over time. Based on this, combined with Bayesian inference methods, the uncertainty between tasks and the probability of potential execution conflicts are quantitatively assessed, and a risk adjustment factor is generated. Based on fuzzy logic, the task priority of reinforcement learning prediction is nonlinearly mapped, and the risk adjustment factor is integrated into the priority weight calculation to generate the dynamic priority weight of each task. Output the weight vectors for all tasks, including the dynamic priority weight value, corresponding risk adjustment factor, task identifier, and allowable deviation range for each task, forming a preliminary weight set.

4. The adaptive scheduling voice SMS processing method according to claim 3, characterized in that, The process of generating the dynamic weight matrix is ​​as follows: For each task in the initial weight set, extract the dynamic priority weight value and the corresponding risk adjustment factor, and calculate the adjustable boundary of the weight in combination with the task's possible deviation range. During the weight adjustment process, a weighted superposition and normalization method is used to merge the initial priority weights with the deviation from the constraint boundary. A consistency check is performed on the fused weight vector to detect and correct conflicts or over-limit situations caused by deviation from constraints, and a dynamic weight matrix containing task dynamic weight values, risk adjustment factors and deviation constraint information is generated.

5. The adaptive scheduling voice SMS processing method according to claim 4, characterized in that, The process of jointly allocating CPU, bandwidth, power, and execution time window is as follows: The dynamic weight matrix is ​​mapped to a multi-dimensional resource allocation space to extract the dynamic priority weight, risk adjustment factor and deviation constraint range of each task, and at the same time, their respective CPU, bandwidth, power requirements and execution time windows are calculated. Construct a multi-objective optimization function, incorporating task priority weights, deviation constraints, risk adjustment factors, and system resource availability into the constraints, and perform constraint optimization on the resource allocation for each task; Based on heuristic scheduling and iterative optimization methods, resources are allocated first to high-weight, hard-constraint tasks, while dynamic adjustments are made to tasks that are allowed to deviate. During the optimization process, the CPU, bandwidth, power allocation values ​​and executable time range of each task are recorded to generate a complete set of resource allocation results.

6. The adaptive scheduling voice SMS processing method according to claim 5, characterized in that, The process of generating a preliminary resource mapping scheme is as follows: Based on the task dependencies in the cause-effect graph, the resource allocation results are topologically sorted to establish a preliminary task execution sequence; By combining the task status constraint set, dynamic weight set, and risk adjustment factor, the initial task sequence is fine-tuned in terms of priority, and non-critical tasks are allowed to be flexibly adjusted within the range of deviation from the constraints. During the sorting process, based on heuristic scheduling and iterative optimization methods, the possible time overlap, resource conflict and priority competition between tasks are dynamically balanced, and a task sequence that takes into account both resource allocation and execution order is output. The adjusted task sequence is integrated with the resource allocation results to generate a preliminary resource mapping scheme that includes CPU, bandwidth, power allocation, executable time range, and priority labeling for each task.

7. The adaptive scheduling voice SMS processing method according to claim 6, characterized in that, The process of task reordering, time slice adjustment, and resource reallocation is as follows: Real-time collection of the execution status, resource usage, and environmental parameters of the preliminary resource mapping scheme, including CPU load, bandwidth utilization, power consumption, and time slice usage, to construct multi-dimensional state features; The multidimensional state features are compared with preset thresholds. When any indicator exceeds the threshold, the task reordering mechanism is triggered. The execution order of high-priority tasks is fine-tuned based on dynamic priority weights, deviation constraints, and risk adjustment factors. By combining time slice adjustment strategies and resource allocation optimization methods, the resource consumption of low-priority tasks is dynamically optimized. By combining an iterative optimization algorithm with a dynamic task mapping table, the adjusted task order, resource usage, and time slice information are synchronously calculated and corrected to form an optimized task state.

8. The adaptive scheduling voice SMS processing method according to claim 7, characterized in that, The process of generating a complete resource mapping scheme is as follows: Based on the optimized task state, the task deviation constraints are mapped to an adjustable execution time and resource allocation range; For high-risk or critical tasks, deviations are limited based on priority gating mechanisms; For tasks that can be flexibly scheduled, an iterative optimization algorithm is used to dynamically adjust the allocation of CPU, bandwidth, power supply, and time slices; During the iteration process, resource overlap and time overlap are handled through conflict detection and correction mechanisms, and a complete resource mapping scheme containing the final execution scope of the task, CPU, bandwidth, power allocation and constraint verification information is output.

9. The adaptive scheduling voice SMS processing method according to claim 8, characterized in that, The process of generating an adaptively updated scheduling optimization strategy is as follows: Input historical task execution data, current environmental parameters, and complete resource mapping scheme into the prediction model to perform time-series modeling and risk assessment of the execution conflict probability, resource consumption fluctuations, and potential bottlenecks for each task. Based on the prediction results, the conflicting tasks are mapped to the deviation from the constraint range, and the dynamic update coefficient of the adjustable range of the task is calculated. For high-priority or safety-critical tasks, the deviation should be strictly limited; For low-priority or flexibly schedulable tasks, iterative optimization algorithms are used to dynamically adjust resource allocation, time-slice scheduling, and execution order. Based on reinforcement learning strategies combined with risk adjustment factors and historical experience data, task priority, resource consumption, and time slice parameters are jointly optimized to generate an adaptively updated scheduling optimization strategy.

10. An adaptive scheduling voice SMS processing system, applied to the method described in any one of claims 1-9, characterized in that, include: Causal analysis module: Extracts task type, resource requirements, execution time and potential causal relationships, constructs causal graph and marks safety criticality, hard constraints and deviation range, forming a task state constraint set; Weight calculation module: Based on the causal graph and task state constraint set, it uses reinforcement learning, Bayesian inference and fuzzy logic to calculate the dynamic priority weight of the task and generate a dynamic weight matrix; Resource allocation module: It jointly allocates CPU, bandwidth, power and time window according to dynamic weight matrix, and determines the task execution order in combination with constraint set to generate resource mapping scheme; Status monitoring module: Real-time collection of task execution status, resource usage and environmental parameters. When the indicators exceed the threshold, task reordering, time slice adjustment and resource reallocation are performed to generate a complete resource mapping scheme. Strategy optimization module: Based on historical task execution data, multi-source environmental information and complete resource mapping scheme, establish a prediction model, calculate potential constraint conflicts, and generate adaptive and updated scheduling optimization strategies.

Images