Special training activity plan auxiliary generation system and method

Abstract

The invention provides an auxiliary generation system and method for a special training activity plan, and relates to the technical field of data processing, and the system comprises a multi-source data collection device which is used for continuously obtaining multi-source data of a trainer, and the multi-source data comprises real-time physiological monitoring data, historical training record data and preset training target parameters; the physiological adaptability modeling engine is in data connection with the multi-source data acquisition device and is used for constructing and dynamically updating a dynamic state model representing the individual physiological response characteristics of the trainee based on the multi-source data, and the dynamic state model quantitatively represents the physiological load bearing capacity, the fatigue accumulation state and the recovery rate of the trainee; the intelligent plan generator is connected with the physiological adaptability modeling engine and used for generating a personalized training scheme through a multi-target optimization algorithm according to the quantitative difference between the dynamic state model and preset training target parameters, and the personalized training scheme configures a target physiological intensity interval for each training unit; the invention discloses a real-time monitoring and adjusting device.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and in particular to a special training activity plan auxiliary generation system and method. BACKGROUND

[0002] In the field of special sports training, competitive sports and public fitness, formulating a scientific training plan is the key to improving sports performance and avoiding sports injuries.

[0003] The existing training plan generation method mainly has the following deficiencies: first, a large number of training plans are formulated based on universal templates or general theories. Such plans usually present a fixed cycle and intensity progression mode, and cannot fully consider individual differences of different trainers, such as initial physical fitness level, recovery ability, physiological characteristics, etc. This one-size-fits-all approach may not be effective for some trainers due to low intensity, and may increase fatigue accumulation and injury risk for another part of the trainers due to high load.

[0004] Second, some rely on personalized plans formulated by human coaches, which take into account individual situations to some extent, but the adjustment process has significant lag. Coaches usually adjust the plan based on weekly or monthly summaries and subjective feedback from the trainer. This approach cannot capture the immediate physiological state fluctuations of the trainer caused by daily routines, nutrition, stress, etc., and cannot intervene in real time during the training process. When the body has sent an overload signal, the adjustment of the plan is usually too late.

[0005] Therefore, how to overcome the rigidity and lag of the existing training plan, and effectively utilize multi-source physiological data to realize truly individualized, dynamic and intelligent training guidance is a technical problem to be solved in the field. SUMMARY

[0006] The present application provides a special training activity plan auxiliary generation system and method to solve the problem of how to overcome the rigidity and lag of the existing training plan, and effectively utilize multi-source physiological data to realize truly individualized, dynamic and intelligent training guidance in the prior art.

[0007] The present application provides a special training activity plan auxiliary generation system, comprising: A multi-source data acquisition device for continuously acquiring multi-source data of the trainer, the multi-source data comprising real-time physiological monitoring data, historical training record data and preset training target parameters; a physiological adaptability modeling engine, connected with the multi-source data acquisition device, configured to construct and dynamically update a dynamic state model representing physiological response characteristics of the individual trainer based on the multi-source data, the dynamic state model quantifying physiological load bearing capacity, fatigue accumulation state and recovery rate of the trainer; an intelligent plan generator, connected with the physiological adaptability modeling engine, configured to generate a personalized training plan through a multi-objective optimization algorithm according to a quantitative difference between the dynamic state model and a preset training target parameter, the personalized training plan configuring a target physiological intensity interval for each training unit; a real-time monitoring and adjuster, simultaneously connected with the multi-source data acquisition device and the intelligent plan generator, configured to compare real-time physiological monitoring data in a training execution process with the target physiological intensity interval, and generate an adjustment instruction based on the dynamic state model to adaptively adjust a subsequent training unit when a deviation beyond a threshold is detected; a feedback optimization processor, respectively establishing data channels with the physiological adaptability modeling engine, the intelligent plan generator and the real-time monitoring and adjuster, configured to iteratively optimize model parameters of the physiological adaptability modeling engine based on execution data of a complete training cycle, forming a system self-evolution closed loop.

[0008] According to the special training activity plan auxiliary generation system provided by the application, the physiological adaptability modeling engine extracts physiological response baseline features from historical training record data through a deep learning network, calculates fatigue accumulation state and recovery rate through a time series prediction model in combination with real-time physiological monitoring data, and fuses the physiological response baseline features, fatigue accumulation state and recovery rate to generate the dynamic state model. The dynamic state model is updated by a sliding window mechanism, and data weights in the window decay over time, ensuring the sensitivity of the model to recent physiological changes.

[0009] According to the special training activity plan auxiliary generation system provided by the application, the multi-objective optimization algorithm of the intelligent plan generator constructs a target function containing three optimization items: a training adaptability gain maximization item based on the dynamic state model prediction, a damage risk minimization item based on fatigue accumulation state calculation, and a training density optimization item based on recovery rate determination. The personalized training plan is obtained by solving the Pareto optimal solution of the target function, and the target physiological intensity interval of each training unit is dynamically determined according to the current parameters of the dynamic state model.

[0010] The real-time monitoring and adjuster establishes a physiological response prediction benchmark based on the dynamic state model, and calculates the deviation degree of the actual physiological monitoring data relative to the physiological response prediction benchmark in real time; When the deviation degree exceeds a first threshold value, the adjustment instruction reduces the intensity parameter of the current training unit; when the deviation degree exceeds a second threshold value, the adjustment instruction reconfigures the load distribution of the remaining training units; and when the deviation degree exceeds a third threshold value, the adjustment instruction terminates the high-intensity training and switches to a recovery mode. The first threshold value, the second threshold value and the third threshold value are dynamically adjusted according to the fatigue accumulation state in the dynamic state model.

[0011] The feedback optimization processor adopts a reinforcement learning framework, takes the difference between the training target achievement degree and the expected target as a reward function, and takes an abnormal deviation event of the physiological monitoring data as a penalty term. The feature extraction weight of the physiological adaptability modeling engine and the objective function parameter of the intelligent plan generator are optimized through a policy gradient algorithm, so that the personalized training scheme generated by the system gradually converges to the optimal training strategy of the trainer.

[0012] The application also provides a special training activity plan auxiliary generation method based on any of the above systems, comprising: The real-time physiological monitoring data, the historical training record data and the preset training target parameters of the trainer are obtained through the multi-source data acquisition device; The physiological adaptability modeling engine is used to construct a dynamic state model based on the obtained data, comprising: The dynamic state model quantitatively represents the current physiological load bearing capacity, fatigue accumulation state and recovery rate of the trainer; The intelligent plan generator generates a personalized training scheme based on the quantitative difference between the dynamic state model and the preset training target parameters, and each training unit in the personalized training scheme is associated with a corresponding target physiological intensity interval; During the training execution process, the real-time monitoring and adjuster compares the real-time physiological monitoring data with the target physiological intensity interval, and generates an adjustment instruction for adaptive adjustment when the deviation exceeds a threshold value; After the training cycle ends, the feedback optimization processor optimizes the system model parameters based on the whole-cycle execution data, and the optimized system model parameters are used for scheme generation in the next training cycle to perform a continuous optimization cycle.

[0013] According to the special training activity plan auxiliary generation method provided by the application, the method further comprises: When generating personalized training schemes, training adaptive gain, impairment risk, and training density are used as optimization objectives, and the optimal solution is searched in the constraint space using a genetic algorithm. The constraint space is determined by the parameter boundaries of the dynamic state model. The constraint space includes the upper limit of physiological load tolerance, the safety threshold of fatigue accumulation state, and the minimum requirement for recovery rate.

[0014] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the special training activity plan auxiliary generation method as described above.

[0015] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the special training activity plan auxiliary generation method as described above.

[0016] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the method for assisting in the generation of a specific training activity plan as described above.

[0017] The system and method for generating specialized training activity plans provided by this invention constructs and updates a user's individualized dynamic state model in real time, ensuring that the generated training plan is highly matched to the trainee's immediate physiological state from the outset, achieving true deep personalization. More importantly, through real-time monitoring and adjustment, the system can intervene immediately in physiological states that deviate from safe or efficient ranges during training, transforming passive, lagging adjustments into proactive, forward-looking risk management, thereby maximizing training efficiency while ensuring training safety. Its core innovation lies in achieving the system's self-evolutionary capability through a feedback optimization processor; the system can learn and iterate from each training data point, continuously deepening its understanding of user physiological adaptability, ultimately transforming from a static tool into an intelligent training partner that grows alongside the user. In summary, this invention effectively connects isolated physiological data with training decisions, overcoming the rigidity and lag of existing technologies, and providing a safer, more efficient, and adaptively evolving new intelligent training paradigm. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating the special training activity plan generation system provided by the present invention; Figure 2 A schematic diagram of the process for the auxiliary generation method of the special training activity plan provided by the present invention; Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0021] Figure 1 This is a flowchart illustrating the specialized training activity plan generation system provided by the present invention, as shown below. Figure 1 As shown, the method includes the following: The multi-source data acquisition device 110 is used to continuously acquire multi-source data of the trainee, including: real-time physiological monitoring data, historical training record data, and preset training target parameters. The physiological adaptation modeling engine 120 is connected to the multi-source data acquisition device and is used to construct and dynamically update a dynamic state model that represents the physiological response characteristics of an individual trainee based on the multi-source data. The dynamic state model quantifies the trainee's physiological load tolerance, fatigue accumulation state, and recovery rate. The intelligent plan generator 130 is connected to the physiological adaptation modeling engine and is used to generate a personalized training plan through a multi-objective optimization algorithm based on the quantitative difference between the dynamic state model and the preset training target parameters. The personalized training plan configures a target physiological intensity range for each training unit. The real-time monitoring and adjustment unit 140 is also connected to the multi-source data acquisition device and the intelligent plan generator. It is used to compare the real-time physiological monitoring data during the training process with the target physiological intensity range. When a deviation exceeding the threshold is detected, an adjustment instruction is generated based on the dynamic state model to adaptively adjust the subsequent training units. The feedback optimization processor 150 establishes data channels with the physiological adaptive modeling engine, the intelligent plan generator, and the real-time monitoring and regulating unit, respectively, for iterative optimization of the model parameters of the physiological adaptive modeling engine based on the execution data of the complete training cycle, forming a self-evolving closed loop of the system.

[0022] In this application, the multi-source data acquisition device 110 functions as an input interface for the system to acquire multi-source heterogeneous data related to the trainee. This device can be implemented as an integrated software module that communicates with various external devices or data platforms, interacting with data through standardized data transmission protocols (e.g., Bluetooth Low Energy (BLE), ANT+, or application programming interfaces (APIs)).

[0023] Real-time physiological monitoring data refers to physiological indicators collected continuously during training. Examples include heart rate (HR) and heart rate variability (HRV) data collected by heart rate monitoring devices; power output data collected by power measurement devices; and speed, distance, and altitude data collected by Global Positioning System (GPS) devices.

[0024] Historical training data refers to information about past training activities stored in a local or cloud database. This data includes, but is not limited to, the date, duration, distance or power completed for each training session, as well as training load metrics calculated from this data, such as training impulse (TRIMP).

[0025] Preset training target parameters are structured target information input by the trainee or coach through a user interface. Examples include the type and date of the target event, the expected athletic performance (such as the time to complete a specific distance), or the target for improving a specific physiological ability (such as the expected improvement in functional threshold power (FTP)).

[0026] The physiological adaptation modeling engine 120 processes raw data and constructs a mathematical model to characterize the physiological state of the trainee. This engine is connected to the multi-source data acquisition device 110 via data coupling, receiving the acquired data as model input.

[0027] A dynamic state model is a multidimensional data structure or mathematical model designed to comprehensively and quantitatively assess the physiological state of a trainee at a specific point in time. This model quantifies at least the following dimensions: Physiological load tolerance characterizes the upper limit of a trainee's safe tolerance to training stimuli in the current state. Its value is dynamic and is affected by recent training and recovery history.

[0028] The cumulative fatigue state characterizes the degree of physiological fatigue experienced by trainees due to training and other stressors.

[0029] Recovery rate characterizes a trainee's ability or speed to recover from a training load to baseline levels. In one specific embodiment, the engine may operate based on rule-based algorithms, such as calculating a moving average training load over a specific period (e.g., the past four weeks) to initially estimate physiological load tolerance.

[0030] The intelligent training plan generator 130 is designed to develop personalized training programs based on the output of the physiological adaptation modeling engine. It receives dynamic state model parameters and compares them with preset training target parameters to generate structured training instructions.

[0031] Quantitative difference refers to the gap between the current state and the target state in terms of quantifiable indicators.

[0032] For example, if the preset training objective is to increase the functional threshold power to 300 watts, while the current capability baseline value evaluated by the dynamic state model is 280 watts, then the quantization difference is 20 watts.

[0033] In multi-objective optimization algorithms, the formulation of a training plan requires balancing multiple conflicting objectives, such as maximizing training adaptive gains and minimizing physiological stress risks. In one implementation, the algorithm can be a constrained optimization model whose objective function is to maximize the expected training effect while controlling the physiological risk index within a preset safety threshold.

[0034] Personalized training programs output a structured set of instructions that includes specific execution parameters. For example, the program may explicitly define the target physiological intensity range for a specific training unit (such as interval training), such as "maintaining heart rate at 170-180 beats / minute" or "maintaining power output at 280-300 watts".

[0035] The function of the real-time monitoring and regulation unit 140 is to perform process control during training execution, ensure that training is carried out according to preset parameters, and intervene when abnormal physiological responses occur.

[0036] The regulator continuously compares the real-time physiological monitoring data continuously transmitted by the multi-source data acquisition device 110 with the target physiological intensity range preset by the intelligent plan generator 130.

[0037] Deviations exceeding a threshold are used to identify persistent and significant abnormal physiological responses. For example, a preset trigger condition could be "real-time heart rate exceeding the upper limit of the target range by 5 beats / minute for 2 consecutive minutes".

[0038] Once the adjustment mechanism is triggered, the regulator will generate an adjustment instruction based on the current dynamic state model (e.g., the model indicates a high level of fatigue accumulation). In a basic embodiment, the adjustment instruction could be "reduce the intensity parameter of subsequent unexecuted training units by 10%".

[0039] The feedback optimization processor 150 enables adaptive learning and model iteration for the system. This processor typically runs after a complete training cycle.

[0040] The processor integrates all relevant data within the cycle, including plan execution data, physiological response data, regulator intervention records, and final effect evaluation data. By analyzing the deviation between the predetermined plan and the actual results, the processor corrects the model parameters of the upstream modules. For example, if the analysis finds that the training intensity set by the system for a trainee is frequently reduced, it indicates that the initial model may have overestimated their physiological load tolerance. In this case, the processor will correct the relevant calculation parameters in the physiological adaptation modeling engine 120.

[0041] Through a continuous cycle of "plan-execution-feedback-correction", the system's modeling accuracy of the physiological response patterns of specific trainees will continue to improve, thereby generating more accurate and effective training plans in subsequent cycles.

[0042] In this application, the collaborative operation of various devices constructs a closed-loop intelligent optimization system. This system not only generates initial training plans based on individual characteristics, but more importantly, it can make real-time adjustments during training execution and perform model self-optimization after the macro-cycle ends. This mechanism achieves precise control over the training load, thereby improving training efficiency while effectively reducing the risk of sports injuries and overcoming the limitations of traditional static training plans.

[0043] Optionally, the physiological adaptive modeling engine extracts physiological response baseline features from historical training record data through a deep learning network, combines real-time physiological monitoring data to calculate fatigue accumulation state and recovery rate through a time-series prediction model, and fuses the physiological response baseline features, fatigue accumulation state and recovery rate to generate the dynamic state model. The dynamic state model is updated using a sliding window mechanism, where the weights of data within the window decay over time, ensuring the model's sensitivity to recent physiological changes.

[0044] In this application, feature extraction employs deep learning networks, such as convolutional neural networks (CNNs) or recurrent neural networks (RNNs) and their variants, to process large amounts of historical training data. Deep learning networks can automatically learn and extract baseline features of physiological responses that characterize individual-specific responses from high-dimensional, non-linear raw data. For example, the network can learn heart rate drift patterns or power decay curves for specific trainees under different types of training (such as endurance or burst training) from long-term heart rate and power data. These features constitute a relatively stable foundational part of the dynamic state model.

[0045] State calculation employs time-series predictive models, such as Long Short-Term Memory (LSTM) networks or Gated Recurrent Units (GRUs), to analyze recent real-time physiological monitoring data and training load data. These models effectively capture dynamic dependencies in time-series data. For example, by analyzing HRV, resting heart rate, and cumulative training load over the past few days, the model can calculate the current cumulative fatigue state index in real time and predict an individual's recovery rate.

[0046] Model fusion combines the baseline features of the physiological response obtained in the first two steps with the dynamically calculated state parameters to generate a unified dynamic state model. Fusion techniques may include dimensionality reduction through a fully connected layer after concatenating feature vectors, or assigning dynamic weights to different features using an attention mechanism.

[0047] To ensure that the model reflects both long-term adaptability and remains sensitive to recent changes, a sliding window mechanism is used in the data update process.

[0048] For example, the system can use a longer time window (e.g., 28 days) to update baseline features, while using a shorter time window (e.g., 3 days) to update fatigue and recovery status. Furthermore, the data within the window can be assigned weights that decay over time, i.e., time decay, so that the latest data has the greatest impact on the model's current state.

[0049] In this application, by introducing deep learning and time-series prediction models, the physiological adaptive modeling engine can more deeply explore the nonlinear relationships in the data and construct a dynamic state model that highly matches the individual's true physiological state. The combination of sliding window and time decay mechanisms ensures that the model has both stability and sensitivity, thereby significantly improving the scientific basis and accuracy of personalized training scheme formulation.

[0050] Optionally, the multi-objective optimization algorithm of the intelligent plan generator constructs an objective function containing three optimization terms: a term for maximizing training adaptability gain based on the prediction of the dynamic state model, a term for minimizing damage risk based on fatigue accumulation state calculation, and a term for optimizing training density based on recovery rate determination; The personalized training scheme is obtained by solving the Pareto optimal solution of the objective function, and the target physiological intensity range of each training unit is dynamically determined according to the current parameters of the dynamic state model.

[0051] In this application, the training plan generation process is formalized as a constrained optimization problem. The multi-objective optimization algorithm constructs an objective function containing the following three key optimization terms: The training adaptive gain maximization term aims to maximize the expected training effect. Its value is calculated by a predictive model that predicts the expected gain of key physiological indicators (such as functional threshold power and VO2 max) after training, based on the current dynamic state model and the training load to be planned.

[0052] The injury risk minimization term aims to control the risk of training-related physiological stress. Its value is directly related to the fatigue accumulation state parameter in the dynamic state model. For example, a risk function can be constructed where the probability of injury risk increases non-linearly when the fatigue accumulation index exceeds a preset threshold.

[0053] The training density optimization term aims to optimize the allocation ratio of training load to recovery time. Its value is determined by the recovery rate parameter. Individuals with higher recovery rates can be assigned higher training densities; conversely, those with lower recovery rates should have lower training densities to ensure adequate recovery.

[0054] By solving this objective function, the system can find a series of Pareto optimal solutions. Pareto optimality, or non-dominated solution, refers to a set of solutions where, for any one solution in the set, no other feasible solution can improve the value of another objective function without reducing the value of at least one objective function. The system can provide users with multiple options located on the Pareto optimal frontier, such as training strategies representing "high return, high risk," "balanced," and "low risk, robust," respectively.

[0055] Furthermore, the target physiological intensity range for each training unit is no longer a fixed value, but is dynamically determined based on the current parameters of the dynamic state model when the plan is generated. For example, when the dynamic state model indicates a high level of accumulated fatigue, the target heart rate range generated by the system will automatically be lowered, even for the same type of training task.

[0056] In this application, the training plan development process is transformed into a rigorous multi-objective mathematical optimization problem, enabling the generated plan to achieve a quantitative and scientific balance between training effectiveness, injury risk, and recovery pace. The dynamic determination mechanism of intensity zones further enhances the plan's immediate adaptability, ensuring that the stimulus for each training session matches the trainee's current physiological state.

[0057] Optionally, the real-time monitoring and regulating device establishes a physiological response prediction benchmark based on the dynamic state model and calculates the deviation of the actual physiological monitoring data from the physiological response prediction benchmark in real time. When the deviation exceeds the first threshold, the adjustment instruction reduces the intensity parameter of the current training unit; when the deviation exceeds the second threshold, the adjustment instruction reconstructs the load distribution of the remaining training units; when the deviation exceeds the third threshold, the adjustment instruction terminates high-intensity training and switches to recovery mode. The first threshold, the second threshold, and the third threshold are dynamically adjusted according to the fatigue accumulation state in the dynamic state model.

[0058] In this application, the physiological response prediction benchmark is as follows: The regulator first invokes the physiological adaptive modeling engine to predict a reasonable range of physiological responses in real time based on the current dynamic state model and the parameters of the training unit being executed. This range serves as the physiological response prediction benchmark. This benchmark is more individualized and immediate than a fixed target interval.

[0059] Deviation calculation can be achieved by the regulator continuously comparing real-time physiological monitoring data with the prediction benchmark and calculating the deviation.

[0060] A graded response mechanism refers to a regulator that incorporates a graded response strategy based on different deviation thresholds to achieve differentiated management of physiological abnormalities of varying severity.

[0061] First threshold (slight deviation): When the deviation exceeds the first threshold, a fine-tuning of the intensity is triggered. For example, the adjustment instruction lowers the intensity parameters (such as power target) of the current training unit by 10-20% to reduce physiological stress while maintaining training continuity.

[0062] Second threshold (moderate deviation): When the deviation exceeds the second threshold, load redistribution is triggered. For example, the adjustment instruction distributes the remaining load of the current training unit to subsequent training units, or replaces the current unit with a low-intensity recovery unit to avoid excessive physiological stress in a single instance.

[0063] Third threshold (severe deviation): When the deviation exceeds the third threshold, protective termination is triggered. This is considered a high-risk signal, and the adjustment command will immediately terminate the current high-intensity training and may guide the user to perform active recovery activities.

[0064] The dynamic thresholds are not fixed but are dynamically adjusted based on the cumulative fatigue state in the dynamic state model. When the model indicates that the user is in a state of high fatigue, these three thresholds will decrease accordingly, thereby improving the system's monitoring sensitivity and protection, and ensuring that appropriate risk management is provided under various physiological conditions.

[0065] In this application, by establishing a dynamic prediction benchmark and a graded response mechanism, the real-time adjustment function is upgraded from a simple threshold alarm to a refined risk management process. The dynamic adjustment capability of the threshold further enhances the system's security, ensuring that the system can provide the most suitable protection strategy based on the trainee's constantly changing fatigue state, thus achieving truly intelligent risk control.

[0066] Optionally, the feedback optimization processor adopts a reinforcement learning framework, using the difference between the training objective achievement and the expected objective as the reward function, and using abnormal deviation events of physiological monitoring data as penalty terms. By optimizing the feature extraction weights of the physiological adaptive modeling engine and the objective function parameters of the intelligent plan generator through the policy gradient algorithm, the personalized training scheme generated by the system gradually converges to the trainee's optimal training strategy.

[0067] In this application, the entire intelligent training optimization system is abstracted as a reinforcement learning agent. The state refers to the environmental information observed by the agent at the decision-making moment, i.e., the dynamic state model generated by the physiological adaptive modeling engine.

[0068] Actions refer to the decisions made by an agent based on its current state. Specifically, these can be personalized training plans generated by an intelligent plan generator or adjustment instructions generated by a real-time monitoring and regulating device.

[0069] Reward refers to the scalar feedback signal obtained by an agent from the environment after performing an action. The reward function in this embodiment is designed as follows: Positive rewards are derived from the difference between the degree to which the training objective is achieved and the expected objective. When the actual training effect exceeds expectations, a positive reward is given.

[0070] Negative rewards arise from abnormal deviations in physiological monitoring data. For example, each time a moderate or severe deviation in regulation is triggered, a negative reward is generated.

[0071] A policy refers to the mapping function from a state to an action of an agent, which corresponds to the model parameters and decision logic in the system.

[0072] In this application, after a training cycle, the processor collects all state-action-reward sequence data from that cycle. Subsequently, reinforcement learning algorithms such as policy gradient are used to update the system's policy.

[0073] Specifically, the algorithm adjusts the feature extraction weights of the physiological adaptive modeling engine and the objective function parameters of the intelligent plan generator, with the goal of maximizing long-term cumulative rewards when the system makes decisions in the future.

[0074] In this application, by introducing a reinforcement learning framework, the system's optimization process is transformed into an active exploration and learning process aimed at maximizing long-term returns. The system can autonomously learn highly individualized training strategies, even surpassing traditional experience, through interaction with the environment, ultimately converging to the optimal training mode for a specific trainee, achieving the highest level of intelligence.

[0075] Figure 2This is a schematic diagram of the process for the auxiliary generation method of the special training activity plan provided by the present invention, as shown below. Figure 2 As shown, it includes: Step 21: Acquire the trainee's real-time physiological monitoring data, historical training record data, and preset training target parameters through the multi-source data acquisition device; Step 22: Utilize the physiological adaptability modeling engine to construct a dynamic state model based on the acquired data. The dynamic state model quantitatively represents the trainee's current physiological load tolerance, fatigue accumulation state, and recovery rate. Step 23: The intelligent plan generator generates a personalized training scheme based on the quantitative difference between the dynamic state model and the preset training target parameters. Each training unit in the personalized training scheme is associated with a corresponding target physiological intensity range. Step 24: During the training process, the real-time monitoring and regulating device compares the real-time physiological monitoring data with the target physiological intensity range. When the deviation exceeds the threshold, it generates a regulation command for adaptive adjustment. Step 25: After the training cycle ends, the feedback optimization processor optimizes the system model parameters based on the full-cycle execution data. The optimized parameters are used for scheme generation in the next training cycle, forming a continuous optimization loop.

[0076] In this application, the user first sets their training goals through the system's interactive interface, such as the target event type and the desired competitive level. This constitutes the preset training goal parameters. The system automatically synchronizes the user's historical activity data on the associated sports platform through an authorized API interface, including the duration, intensity distribution, heart rate variability, and calculated training load (such as heart rate-based training impulse (TRIMP) or power-based training stress score (TSS). This constitutes the historical training record data. During each training session, the user wears a compatible physiological monitoring device (such as a heart rate monitor), and the system continuously acquires real-time physiological monitoring data such as heart rate and heart rate variability (HRV) through a multi-source data acquisition device.

[0077] Before the training cycle begins, the physiological adaptation modeling engine performs in-depth analysis of the acquired historical training data. Using machine learning algorithms (e.g., recurrent neural networks), the model extracts the user's individual physiological response patterns from long-term training and physiological response data, such as the heart rate's response speed to different intensities of training load and recovery curve characteristics, thereby constructing a baseline model of physiological load tolerance representing their long-term adaptive level. Simultaneously, by combining recent days' resting heart rate and HRV data, the model assesses the user's current fatigue accumulation and recovery rate. These three core parameters together constitute the user's initial dynamic state model.

[0078] The intelligent plan generator receives a dynamic state model and preset training target parameters. By comparing the quantitative difference between the current ability evaluated by the model and the target level, the system identifies key physiological abilities that need to be improved. Subsequently, the system initiates a multi-objective optimization algorithm whose objective function aims to maximize the expected training adaptive gain while constraining the physiological risk index calculated based on the fatigue accumulation state below a safe threshold.

[0079] After the algorithm solves the problem, it generates a personalized training plan containing multiple training phases. Each key training unit in the plan clearly defines a target physiological intensity range. For example, the target heart rate range for a high-intensity interval training session is set within a specific high-intensity range.

[0080] During this high-intensity interval training session, the real-time monitoring and regulation system activated. Later in the training, the system detected that the user's real-time heart rate remained consistently above the upper limit of the preset target physiological intensity range, indicating that their current physiological stress might have exceeded the planned expectations. The system determined that this deviation had exceeded the preset mild risk threshold.

[0081] To avoid excessive fatigue, the system automatically generates and issues adjustment instructions, such as lowering the intensity target for the next interval set or appropriately extending the active recovery time between sets. Users execute training according to the adjusted instructions, thereby avoiding the potential risk of overtraining while ensuring the effectiveness of the training stimulus.

[0082] At the end of a complete training microcycle (e.g., one week), the feedback optimization processor is activated. It integrates all training execution data, physiological response data, and all auto-adjustment records from that cycle. Processor analysis revealed that the system triggered downward intensity adjustments multiple times during this cycle, suggesting that the initial model may have slightly overestimated the user's recovery rate.

[0083] Based on this analysis, the processor modifies the model parameters used to calculate the recovery rate in the physiological adaptation modeling engine through optimization algorithms (e.g., gradient descent).

[0084] When the system generates a training plan for the next micro-cycle, it will call this parameter-optimized model. Therefore, the newly generated plan will more accurately reflect the user's true recovery ability; for example, it may automatically increase the number of rest days between high-intensity training sessions or reduce the total load on consecutive training days. Through this periodic, continuous optimization cycle, the system's modeling accuracy of the user's physiological characteristics continuously improves.

[0085] This application achieves significant technical benefits by constructing a complete closed loop encompassing data acquisition, modeling, planning, monitoring, adjustment, and feedback optimization. First, it enables highly personalized training plans, generating unique solutions based on an individual's historical physiological response patterns and current state, rather than using a generic template. Second, it possesses real-time adaptive adjustment capabilities, intelligently intervening during training based on immediate physiological feedback, effectively preventing training loads from exceeding an individual's safe tolerance level, thereby significantly reducing the risk of sports injuries. Most importantly, this method, through periodic feedback optimization, endows the system with self-learning and self-evolutionary capabilities. This means the system can continuously improve the accuracy of modeling individual user physiological characteristics, allowing the precision and effectiveness of training guidance to continuously improve over time, ultimately helping users maximize their athletic potential while ensuring safety.

[0086] Optionally, the method further includes: When constructing the dynamic state model, baseline features of physiological responses reflecting individual differences are extracted from historical training record data; Then, the current fatigue accumulation status is assessed by combining real-time physiological monitoring data, and the recovery rate is calculated based on recent recovery performance. The baseline characteristics of the theoretical response, the current fatigue accumulation state, and the recovery rate are weighted and fused to generate a unified dynamic state model. The update frequency of the dynamic state model is synchronized with the training execution frequency, ensuring that the model parameters reflect the latest physiological state of the trainee.

[0087] The method further includes: When generating personalized training schemes, training adaptive gain, impairment risk, and training density are used as optimization objectives, and the optimal solution is searched in the constraint space using a genetic algorithm. The constraint space is determined by the parameter boundaries of the dynamic state model. The constraint space includes the upper limit of physiological load tolerance, the safety threshold of fatigue accumulation state, and the minimum requirement for recovery rate.

[0088] In this application, the process of constructing a dynamic state model is itself a hierarchical and dynamic process.

[0089] First, the system extracts baseline features of physiological responses that reflect individual differences from a large amount of historical training data. This constitutes a baseline model that represents the long-term, stable physiological characteristics of an individual, much like establishing a user's physiological "fingerprint".

[0090] Subsequently, the system combines real-time physiological monitoring data to assess the current state of fatigue accumulation and calculates the recovery rate based on recent recovery performance. These two dynamic parameters together capture the short-term fluctuations in the user's body.

[0091] Ultimately, the system uses a weighted fusion algorithm to generate a unified dynamic state model by combining relatively stable baseline physiological response characteristics with dynamic current fatigue accumulation and recovery rate. This model's update frequency is synchronized with the user's training execution frequency, ensuring that the model parameters accurately reflect the trainee's latest physiological state and providing a data foundation for subsequent decision-making.

[0092] This high-precision, high-time-efficiency dynamic state model then became the core basis for generating personalized training schemes.

[0093] The process of generating the training scheme is formalized as a complex multi-objective constrained optimization problem. The system takes training adaptive gain, impairment risk, and training density as the core optimization objectives, and uses advanced global optimization algorithms such as genetic algorithms to search for the optimal combination of training schemes within a specific constraint space.

[0094] The constraint space is completely defined by the parameters of the dynamic state model generated in the previous step.

[0095] Specifically, the constraint space includes the upper limit of physiological load tolerance, the safety threshold of fatigue accumulation state, and the minimum requirement for recovery rate, all determined by the model.

[0096] All candidate training schemes generated by the genetic algorithm during the iteration process must strictly adhere to these boundary conditions determined by the individual's current physiological state; any scheme that exceeds the boundary will be eliminated.

[0097] This embodiment achieves a fundamental shift from experience-driven to data-driven approaches by tightly integrating hierarchical dynamic modeling with individualized constraint-based optimization search. It goes beyond simply generating a plan; it dynamically seeks the optimal path that maximizes long-term gains within a safe flight envelope defined by a precise individual model. This deeply coupled mechanism ensures that the training scheme, while pursuing ultimate performance, is fundamentally safe and feasible, thereby generating highly complex and individually optimal training strategies that surpass human coaching experience.

[0098] Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention, such as... Figure 3As shown, the electronic device may include a processor 310, a communications interface 320, a memory 330, and a communication bus 340, wherein the processor 310, communications interface 320, and memory 330 communicate with each other via the communication bus 340. The processor 310 can call logical instructions in the memory 330 to execute a method for assisting in the generation of a specialized training activity plan. This method includes acquiring real-time physiological monitoring data, historical training record data, and preset training target parameters of the trainee through the multi-source data acquisition device. The physiological adaptation modeling engine is used to construct a dynamic state model based on the acquired data. The dynamic state model quantitatively represents the trainee's current physiological load tolerance, fatigue accumulation state, and recovery rate. The intelligent plan generator generates a personalized training scheme based on the quantitative difference between the dynamic state model and the preset training target parameters. Each training unit in the personalized training scheme is associated with a corresponding target physiological intensity range. During the training process, the real-time monitoring and regulating device compares the real-time physiological monitoring data with the target physiological intensity range. When the deviation exceeds the threshold, it generates a regulation command for adaptive adjustment. After the training cycle ends, the feedback optimization processor optimizes the system model parameters based on the full-cycle execution data. The optimized system model parameters are used to generate the scheme for the next training cycle, so as to carry out continuous optimization loop.

[0099] Furthermore, the logical instructions in the aforementioned memory 330 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0100] On the other hand, the present invention also provides a computer program product, the computer program product including a computer program, the computer program being able to be stored on a non-transitory computer-readable storage medium, the computer program being executed by a processor, the computer being able to execute the special training activity plan auxiliary generation method provided by the above methods, the method including: acquiring real-time physiological monitoring data of trainees, historical training record data and preset training target parameters through the multi-source data acquisition device; The physiological adaptation modeling engine is used to construct a dynamic state model based on the acquired data. The dynamic state model quantitatively represents the trainee's current physiological load tolerance, fatigue accumulation state, and recovery rate. The intelligent plan generator generates a personalized training scheme based on the quantitative difference between the dynamic state model and the preset training target parameters. Each training unit in the personalized training scheme is associated with a corresponding target physiological intensity range. During the training process, the real-time monitoring and regulating device compares the real-time physiological monitoring data with the target physiological intensity range. When the deviation exceeds the threshold, it generates a regulation command for adaptive adjustment. After the training cycle ends, the feedback optimization processor optimizes the system model parameters based on the full-cycle execution data. The optimized system model parameters are used to generate the scheme for the next training cycle, so as to carry out continuous optimization loop.

[0101] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method for generating a special training activity plan assisted by the methods described above, the method comprising: acquiring real-time physiological monitoring data of the trainee, historical training record data, and preset training target parameters through the multi-source data acquisition device; The physiological adaptation modeling engine is used to construct a dynamic state model based on the acquired data. The dynamic state model quantitatively represents the trainee's current physiological load tolerance, fatigue accumulation state, and recovery rate. The intelligent plan generator generates a personalized training scheme based on the quantitative difference between the dynamic state model and the preset training target parameters. Each training unit in the personalized training scheme is associated with a corresponding target physiological intensity range. During the training process, the real-time monitoring and regulating device compares the real-time physiological monitoring data with the target physiological intensity range. When the deviation exceeds the threshold, it generates a regulation command for adaptive adjustment. After the training cycle ends, the feedback optimization processor optimizes the system model parameters based on the full-cycle execution data. The optimized system model parameters are used to generate the scheme for the next training cycle, so as to carry out continuous optimization loop.

[0102] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0103] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0104] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A system for assisting in the generation of specialized training activity plans, characterized in that, include: A multi-source data acquisition device is used to continuously acquire multi-source data from trainees, including: real-time physiological monitoring data, historical training record data, and preset training target parameters. A physiological adaptation modeling engine is connected to the multi-source data acquisition device and is used to construct and dynamically update a dynamic state model that represents the physiological response characteristics of an individual trainee based on the multi-source data. The dynamic state model quantifies the trainee's physiological load tolerance, fatigue accumulation state, and recovery rate. An intelligent plan generator, connected to the physiological adaptation modeling engine, is used to generate personalized training schemes based on the quantitative difference between the dynamic state model and the preset training target parameters through a multi-objective optimization algorithm. The personalized training scheme configures a target physiological intensity range for each training unit. The real-time monitoring and adjustment device is connected to the multi-source data acquisition device and the intelligent plan generator. It is used to compare the real-time physiological monitoring data during the training process with the target physiological intensity range. When a deviation exceeding the threshold is detected, the adjustment command is generated based on the dynamic state model to adaptively adjust the subsequent training units. The feedback optimization processor establishes data channels with the physiological adaptive modeling engine, the intelligent plan generator, and the real-time monitoring and regulation unit, respectively, to iteratively optimize the model parameters of the physiological adaptive modeling engine based on the execution data of the complete training cycle, forming a self-evolving closed loop of the system.

2. The specialized training activity plan generation system according to claim 1, characterized in that, The physiological adaptive modeling engine extracts physiological response baseline features from historical training data through a deep learning network, combines real-time physiological monitoring data with a time-series prediction model to calculate fatigue accumulation state and recovery rate, and fuses the physiological response baseline features, fatigue accumulation state and recovery rate to generate the dynamic state model. The dynamic state model is updated using a sliding window mechanism, where the weights of data within the window decay over time, ensuring the model's sensitivity to recent physiological changes.

3. The specialized training activity plan generation system according to claim 1 or 2, characterized in that, The multi-objective optimization algorithm of the intelligent plan generator constructs an objective function containing three optimization terms: a term to maximize the training adaptability gain based on the prediction of the dynamic state model, a term to minimize the damage risk based on the fatigue accumulation state calculation, and a term to optimize the training density based on the recovery rate. The personalized training scheme is obtained by solving the Pareto optimal solution of the objective function, and the target physiological intensity range of each training unit is dynamically determined according to the current parameters of the dynamic state model.

4. The specialized training activity plan generation system according to claim 1, characterized in that, The real-time monitoring and regulation device establishes a physiological response prediction benchmark based on the dynamic state model and calculates the deviation of the actual physiological monitoring data from the physiological response prediction benchmark in real time. When the deviation exceeds the first threshold, the adjustment instruction reduces the intensity parameter of the current training unit; When the deviation exceeds the second threshold, the adjustment instruction reconstructs the load distribution of the remaining training units; when the deviation exceeds the third threshold, the adjustment instruction terminates high-intensity training and switches to recovery mode. The first threshold, the second threshold, and the third threshold are dynamically adjusted according to the fatigue accumulation state in the dynamic state model.

5. The specialized training activity plan generation system according to claim 1, characterized in that, The feedback optimization processor adopts a reinforcement learning framework, using the difference between the training objective achievement and the expected objective as the reward function, and using abnormal deviation events of physiological monitoring data as the penalty term. By optimizing the feature extraction weights of the physiological adaptive modeling engine and the objective function parameters of the intelligent plan generator through the policy gradient algorithm, the personalized training scheme generated by the system gradually converges to the trainee's optimal training strategy.

6. A method for assisting in the generation of specialized training activity plans based on the specialized training activity plan generation system according to any one of claims 1 to 5, characterized in that, include: The multi-source data acquisition device acquires the trainee's real-time physiological monitoring data, historical training record data, and preset training target parameters. The physiological adaptation modeling engine is used to construct a dynamic state model based on the acquired data. The dynamic state model quantitatively represents the trainee's current physiological load tolerance, fatigue accumulation state, and recovery rate. The intelligent plan generator generates a personalized training scheme based on the quantitative difference between the dynamic state model and the preset training target parameters. Each training unit in the personalized training scheme is associated with a corresponding target physiological intensity range. During the training process, the real-time monitoring and regulating device compares the real-time physiological monitoring data with the target physiological intensity range. When the deviation exceeds the threshold, it generates a regulation command for adaptive adjustment. After the training cycle ends, the feedback optimization processor optimizes the system model parameters based on the full-cycle execution data. The optimized system model parameters are used to generate the scheme for the next training cycle, so as to carry out continuous optimization loop.

7. The method for assisting in the generation of specialized training activity plans according to claim 6, characterized in that, The method further includes: When constructing the dynamic state model, baseline features of physiological responses reflecting individual differences are extracted from historical training record data; Then, the current fatigue accumulation status is assessed by combining real-time physiological monitoring data, and the recovery rate is calculated based on recent recovery performance. The baseline characteristics of the theoretical response, the current fatigue accumulation state, and the recovery rate are weighted and fused to generate a unified dynamic state model. The update frequency of the dynamic state model is synchronized with the training execution frequency, ensuring that the model parameters reflect the latest physiological state of the trainee.

8. The method for assisting in the generation of specialized training activity plans according to claim 6 or 7, characterized in that, The method further includes: When generating personalized training plans, training adaptive gain, impairment risk, and training density are used as optimization objectives, and the optimal solution is searched in the constraint space using a genetic algorithm. The constraint space is determined by the parameter boundaries of the dynamic state model. The constraint space includes the upper limit of physiological load tolerance, the safety threshold of fatigue accumulation state, and the minimum requirement for recovery rate.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method for assisting in the generation of a special training activity plan as described in any one of claims 6 to 8.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method for assisting in the generation of a special training activity plan as described in any one of claims 6 to 8.

Images