A multi-subsystem real-time coupling simulation method and system

Abstract

The application discloses a kind of multi-subsystem real-time coupling simulation method and system, it is related to computer simulation and interactive teaching demonstration technical field.The method is by collecting student operation event stream, calculating target pedal frequency, sprint intention value and actual pedal frequency, and power is combined.The wake flow resistance reduction rate is determined according to the queue geometry parameter of front rider, and the balanced speed is obtained by solving the power and resistance balance equation by numerical iteration.In the same simulation substep, phosphoric acid original reserve, anaerobic glycolysis reserve, blood lactic acid concentration, aerobic energy supply ratio, heart rate, recovery oxygen debt and energy supply substrate composition are updated in parallel, and multiple metabolic state quantities are mutually constrained according to preset cross-coupling relationship.Further, the rider lateral position state quantity and lane change instruction are introduced, and the wake flow resistance reduction rate is determined according to the wake flow cone composed of longitudinal distance and lateral offset;Non-player riders generate virtual operation event stream by difficulty-graded tactical decision, and reuse the same simulation kernel with students.The application can realize interactive cycling energy metabolism simulation of multiple subsystem numerical self-consistent, and is suitable for sports physiology teaching demonstration.

Description

Technical Field

[0001] This application relates to the field of computer simulation and interactive teaching demonstration technology, and in particular to a real-time coupled simulation method and system for multiple subsystems. Background Technology

[0002] In exercise physiology teaching, existing simulation methods mainly include: presenting the relationship between the energy supply system and intensity using static charts or videos; performing conversions using a single power-speed formula or a fixed heart rate curve; and relying on physical power vehicles to collect real physiological data. Static charts and videos can only passively display preset curves, and students cannot change the simulation process through their own operation; in the single formula scheme, each physiological quantity is independent and cannot reflect the mutual constraints between multiple energy supply subsystems; the physical power vehicle scheme relies on dedicated hardware sensors, which has high equipment costs and cannot directly present in vivo state quantities such as phosphagen reserve and anaerobic glycolysis reserve in the simulation interface.

[0003] A Chinese patent with publication number CN121641471A discloses a method, system, device, and storage medium for simulating exercise training based on metabolic capacity. This scheme includes collecting the basal metabolic parameters and exercise physiological data of the training subjects, calculating the metabolic response rate by combining training load parameters, and adaptively adjusting the training load parameters according to differences in metabolic response. However, this scheme uses preset training load parameters as the simulation-driving input. During the simulation, the state variables of each metabolic subsystem are not mutually constrained through preset cross-coupling relationships, and it does not include an interactive closed-loop mechanism that uses discrete user operation events to drive the simulation in real time and feeds back the simulation state to the user. Summary of the Invention

[0004] The purpose of this invention is to provide a real-time coupled simulation method and system for multiple subsystems to solve the following technical problems in the prior art: existing motion simulation schemes cannot drive multiple mutually coupled metabolic subsystems and cycling mechanics subsystems to perform parallel self-consistent calculations in real time by the discrete operation event flow of the trainee; there is a lack of preset cross-coupling constraints between various metabolic state variables; after the phosphagen reserve is depleted, it is impossible to realize hysteresis trigger-threshold unlocking modeling based on the reserve recovery amount; and it is impossible to form a complete interactive simulation closed loop of "operation input → multi-system state calculation → multi-modal feedback output" without external physical sensors.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: This invention provides a real-time coupled simulation method for multiple subsystems. Within each time-limited simulation substep, the following steps are sequentially performed on the rider: Collecting the discrete operation event stream generated by the rider; counting the discrete operation events in the first, second, and third time windows respectively to obtain the count value for each time window, and recording the time elapsed since the last operation event; calculating the target cadence based on the count values ​​of each time window using a piecewise mapping function; obtaining the sprint intention value based on the rate jumps of the count values ​​of each time window through gating and smoothing; calculating the actual cadence based on the target cadence using a first-order inertial relaxation method; calculating the steady-state power based on the actual cadence, and synthesizing the steady-state power with the starting cost term and the sprint surge term to obtain the composite power, wherein the sprint surge term is limited by anaerobic availability, which is synthesized from phosphagen reserves and anaerobic glycolysis reserves; acquiring the geometric parameters of the preceding continuous rider queue. The data includes the number of riders in succession, wheel spacing, lateral alignment, and pace difference. The wake drag reduction rate is determined based on the geometric parameters, and this rate only affects the air resistance term. The balanced speed is obtained by numerically iterating the power-drag balance equation based on the combined power and the effective frontal area corrected by the wake drag reduction rate. Within the same simulation substep, multiple metabolic state variables, including phosphagen reserve, anaerobic glycolysis reserve, blood lactate concentration, aerobic energy contribution, heart rate, oxygen debt recovery, and energy substrate composition, are updated in parallel. These metabolic state variables are mutually constrained by a preset cross-coupling relationship. When the phosphagen reserve drops to zero, it enters a locked state and operation input is prohibited until the phosphagen reserve recovers to a set proportional threshold. The multiple metabolic state variables and the balanced speed are output in a multimodal manner, and the output results are used as feedback for the trainee to adjust their operation, returning to the data acquisition step to form a closed loop. Through the above scheme, a discrete operation event flow is used as the sole driving input, and simultaneous updates are performed within each bounded simulation substep.

[0006] The system incorporates cycling mechanics calculations and parallel updates of multiple metabolic state variables. The outputs of each subsystem maintain numerical consistency between physics and physiology under any operational sequence, thus forming a complete interactive simulation closed loop.

[0007] Furthermore, the first time window has a length of 0.8 s to 1.5 s, the second time window has a length of 2 s to 3 s, and the third time window has a length of 8 s to 12 s; the time step of the simulation substep does not exceed 0.2 s, and the time elapsed since the last operation event is limited before it can be used for state updates. By setting three time windows of different scales, the short window can capture the abrupt changes in sprint decisions, and the long window can represent the continuous load state, thereby making the analysis of operation intent more accurate; the upper limit constraint of the time step ensures the stability of numerical integration.

[0008] Preferably, the sprint intention value is detected as follows: when the rate of operation events within the first time window exceeds a set sprint determination threshold, sprint intention gating is triggered, and continuous sprint intention values ​​are obtained after first-order low-pass filtering; the segmented mapping of the target cadence includes three intervals: low-frequency, mid-frequency, and high-frequency, with different mapping slopes for each interval, and the high-frequency interval corresponding to the sprint interval. The segmented mapping ensures that low-frequency operations correspond to the aerobic endurance interval, and high-frequency operations correspond to the explosive sprint interval, and the gating mechanism prevents mid-frequency operations from being misjudged as sprints.

[0009] Furthermore, a valid wake chain is determined to be formed when there are at least three consecutive riders ahead, the wheel spacing of each wheel is less than a set wheel spacing threshold, the lateral offset is less than a set lateral threshold, and the pace difference is less than a set pace threshold. The wake drag reduction rate increases with the number of consecutive riders ahead and tends to saturate, and is capped after being corrected by three factors: wheel spacing coefficient, alignment coefficient, and pace coefficient. By performing multi-dimensional joint determination of the geometric parameters of the rider queue ahead, the physical mechanism of wake following can be quantitatively reproduced, and the influence of wheel spacing, lateral alignment, and pace matching on drag reduction effect in actual riding can be accurately reflected by the three correction coefficients.

[0010] Specifically, the wheel spacing threshold is 1.5 m; when the wheel spacing of any consecutive wheel is not less than the wheel spacing threshold or the number of riders in front is less than 3, the wake drag reduction rate is zero; the capped value of the wake drag reduction rate does not exceed 60%. The capped value design ensures that the drag reduction rate is within a physically reasonable range, avoiding extreme calculation results that exceed actual aerodynamic laws.

[0011] Furthermore, the numerical iterative solution employs a bisection method, performing a finite number of iterations within a preset speed search interval. The convergence condition is that the difference between drag power and effective power is less than a preset error threshold. When the rider is in a coasting state, the gliding deceleration is calculated based on the sum of rolling resistance and aerodynamic resistance. The bisection method allows for accurate calculation of the equilibrium speed under any drag reduction conditions, ensuring numerical consistency between mechanical and metabolic calculations. The separate handling of the coasting state accurately simulates the natural deceleration process after the rider stops pedaling.

[0012] Furthermore, the preset cross-coupling relationship includes at least one of the following: when the anaerobic glycolysis reserve is lower than a set depletion threshold, the phosphagen reserve is accelerated; when the blood lactate concentration increases, the upper limit of the aerobic energy supply ratio is lowered and the lactate clearance rate is reduced; when the oxygen debt is recovered, the heart rate and the aerobic energy supply ratio are maintained during the low-power phase; the sprint surge term is limited by the anaerobic availability. The above cross-coupling relationship makes the output of each metabolic subsystem mutually restrictive, which can reproduce the real physiological linkage phenomena such as "forced power reduction after sprinting depletes phosphagen reserves", "lactate accumulation lowers the upper limit of aerobic energy supply and inhibits lactate clearance", and "post-exercise oxygen debt maintains a higher heart rate during the low-power phase".

[0013] Preferably, the blood lactate concentration is updated as follows: the blood lactate production rate increases with the increase of anaerobic glycolysis flux and the sprint intention value; the blood lactate clearance rate is maximized within the active recovery power range and is inhibited during high-intensity exercise; the blood lactate concentration is limited within a preset physiological range. This update method accurately simulates the dynamic equilibrium process of blood lactate under different exercise intensities, ensuring that the simulated blood lactate concentration change pattern matches the actual exercise physiology.

[0014] Furthermore, the set percentage threshold ranges from 60% to 90%; during the locked state, the operational input in the acquisition step is cleared, and stepping operations are prohibited. This hysteresis-lock mechanism allows for precise modeling of the dynamic characteristics of the extreme fatigue state. The duration of the lock is endogenously determined by the physiological recovery rate of phosphagen reserves, rather than a fixed time penalty, thus realistically reproducing the dynamic characteristics of the "pull-out-recovery" process. Here, "pull-out" is defined as: a state where the athlete's metabolic energy supply, cardiopulmonary function, and neuromuscular function are insufficient to meet the current exercise intensity, resulting in the inability to maintain the target power or speed, necessitating a significant reduction in speed or even cessation of exercise.

[0015] In some embodiments, the method further includes: receiving an exogenous replenishment trigger command; storing the equivalent amount of the exogenous replenishment in an absorption pool variable; replenishing the anaerobic glycolysis reserve at a set absorption rate with a delay and increasing the proportion of blood glucose in the energy-supplying substrate composition; the number of times the exogenous replenishment is used in each simulation cycle does not exceed a set upper limit, and the time interval between two adjacent uses is not less than a set minimum interval. The exogenous replenishment, through the delayed absorption pool mechanism, simulates the delayed intestinal absorption physiological process of real nutrient replenishment, ensuring that the increase in the proportion of exogenous blood glucose for energy and the recovery of glycolysis reserves both exhibit delayed characteristics consistent with physiological laws.

[0016] Furthermore, the energy-supplying substrate composition includes five substrate pathways: phosphagen, muscle glycogen, blood glucose, free fatty acids, and mitochondrial lipids. The baseline proportion of each substrate is obtained by segmented interpolation based on exercise intensity, and then normalized to 100% after adding correction terms for lactate threshold pressure, sprint state, and exogenous supply. The proportion of phosphagen can transiently increase during sprinting. The normalized calculation of the five energy-supplying substrates enables the simulation system to present the dynamic changes of energy supply composition that cannot be directly measured in vivo in real time, providing an intuitive visualization basis for teaching exercise physiology.

[0017] This invention also provides a multi-subsystem real-time coupled simulation system, including a memory and a processor, and further comprising: an operation acquisition module, used to acquire discrete operation event streams generated by trainees, count the discrete operation events in a first time window, a second time window, and a third time window respectively, obtain the count value of each time window, and record the duration since the last operation event; an intent parsing module, signal-connected to the operation acquisition module, used to calculate the target pedal frequency based on the count values ​​of each time window through a piecewise mapping function, and obtain the sprint intent value based on the rate jump of the count values ​​of each time window through gating and smoothing; a mechanics calculation module, signal-connected to the intent parsing module, used to calculate the actual pedal frequency based on the target pedal frequency using a first-order inertial relaxation method, and synthesize the steady-state power, the starting cost term, and the sprint inrush term to obtain the composite power, wherein the sprint inrush term is limited by anaerobic availability, which is synthesized from phosphate reserves and anaerobic glycolysis reserves; and a resistance solving module, signal-connected to the mechanics calculation module. The calculation module is connected to the drag calculation module and is used to acquire the geometric parameters of the continuous rider queue ahead and determine the wake drag reduction rate that only affects the air resistance term. Based on the combined power and the effective frontal area corrected by the wake drag reduction rate, the balance equation between power and drag is solved numerically through iteration to obtain the balance speed. The metabolic coupling module is connected to the drag solution module and is used to update multiple metabolic state variables in parallel within the same simulation substep, including phosphagen reserve, anaerobic glycolysis reserve, blood lactate concentration, aerobic energy supply ratio, heart rate, recovery oxygen debt, and energy substrate composition. The metabolic state variables are mutually constrained according to a preset cross-coupling relationship. When the phosphagen reserve drops to zero, it enters a locked state to prohibit operation input until the phosphagen reserve recovers to a set proportional threshold and then unlocks. The rendering feedback module is connected to the metabolic coupling module and is used to output the multiple metabolic state variables and the balance speed in a multimodal manner. The output results are transmitted as feedback signals to the operation acquisition module to form a closed loop. Each functional module corresponds strictly to each step of the simulation method. The signal connection relationship between modules ensures the orderly flow of data and the real-time transmission of calculation results from each subsystem.

[0018] In some embodiments, a server-based authoritative scheduling module is further included, communicatively connected to the operation acquisition module, the resistance solving module, and the metabolic coupling module. This module is used to uniformly advance the simulation substeps of multiple users at fixed time intervals. The server-based authoritative scheduling module calculates the wake relation matrix based on the real position coordinates of each user and sends state snapshots to each user's client. Non-player riders use a pace strip mechanism to anchor the current speed of the last user. The server-based authoritative scheduling module ensures time synchronization of the simulation substeps of each user in a multi-user scenario and enables multiple users to share aerodynamic constraints in the same physical scene by uniformly calculating the wake relation matrix.

[0019] As a preferred embodiment, it also includes a behavior telemetry module, which is signal-connected to the operation acquisition module for use in fixing the sampling interval.

[0020] The system records each user's operation events and metabolic status data at intervals, and folds them into statistics for non-overlapping time windows, storing them in the telemetry buffer. The output of the behavior telemetry module is used only for teaching analysis and does not participate in simulation calculations. The behavior telemetry module and the simulation calculation link are isolated from each other, ensuring that telemetry data acquisition does not affect the real-time performance of the simulation engine.

[0021] Compared with the prior art, the present invention has the following beneficial effects: 1. By updating seven metabolic state variables in parallel within each bounded simulation substep—phosphagen reserve, anaerobic glycolysis reserve, blood lactate concentration, aerobic energy supply ratio, heart rate, recovery oxygen debt, and energy substrate—and by pre-setting cross-coupling relationships to constrain each state variable, the output of each subsystem maintains physical and physiological numerical consistency under any operation sequence. This overcomes the shortcomings of existing single-formula schemes where physiological quantities are independent and cannot reflect cross-system linkage.

[0022] 2. By using the discrete operation event flow of the trainee as the sole driving input and outputting the simulation state in a multimodal manner and feeding it back to the acquisition step, a complete interactive simulation closed loop of "operation input → multi-system state calculation → multimodal feedback output" is formed. This allows trainees to change the simulation process in real time through their own operations and observe the changes in physiological indicators in real time, solving the technical problem that existing static charts and video solutions cannot provide a real interactive closed loop.

[0023] 3. By triggering the locking state when the phosphate reserve is reduced to zero and releasing the lock only when the reserve is restored to a set proportional threshold, the dynamic characteristics of the extreme fatigue state can be accurately modeled. The duration of the lock is endogenously determined by the physiological recovery rate of the phosphate reserve. Unlike the fixed time penalty, it can realistically reproduce the 6.5 s of hysteresis dynamics in the "pull-up-recovery" process, while the existing schemes cannot express this mechanism.

[0024] 4. By making a multi-dimensional joint judgment based on the geometric parameters of the continuous rider queue ahead (number of riders, wheel spacing, lateral alignment, and pace difference), and after correction by three factors including wheel spacing coefficient, alignment coefficient, and pace coefficient, the wake drag reduction rate is determined. Moreover, this drag reduction rate only applies to the air resistance term, which realizes the quantitative modeling of the wake following physical mechanism and ensures the self-consistency of the mechanical calculation. Existing motion simulation schemes do not include such a real-time drag correction mechanism based on queue spatial relationship.

[0025] 5. By interpolating the composition of the five energy-supplying substrates (phosphagen, muscle glycogen, blood glucose, free fatty acids, and mitochondrial lipids) according to exercise intensity and normalizing them after superimposing multiple corrections, and by visualizing in real time state quantities such as phosphagen reserves and anaerobic glycolysis reserves that cannot be directly measured in the body, the technical problem that existing physical power vehicle schemes cannot present the body's reserves and energy supply composition has been solved, providing an intuitive mechanism demonstration method for teaching exercise physiology. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the overall architecture of a multi-subsystem real-time coupled simulation method, including lane change input, lateral wake cone drag reduction, and non-player rider tactical decision-making modules.

[0027] Figure 2 This is a schematic diagram of the real-time coupled response curves of multiple physiological subsystems in a teaching scenario.

[0028] Figure 3 A schematic diagram of the stacking structure of the energy supply substrate that evolves in real time with the intensity of motion and time.

[0029] Figure 4 This is a schematic diagram of the characteristics of the wake-riding drag reduction model.

[0030] Figure 5 This is a schematic diagram showing the dynamic changes in phosphate precursor reserves and anaerobic glycolysis reserves, as well as the burst lock-in range.

[0031] Figure 6 Layout diagram of the teaching interaction interface Figure 7 This is a schematic diagram of the transverse cross section of the wake cone and the entry and exit of the wake shielding zone during lane changes.

[0032] Figure 8 A flowchart illustrating the tactical decision-making process for non-player riders, categorized by difficulty level. Detailed Implementation

[0033] The present invention will now be described in further detail with reference to the accompanying drawings. The present invention provides a real-time coupled simulation method, system, and computer-readable storage medium for multiple subsystems. Its core improvement lies in: using the discrete operation event flow of the trainee as the sole driving input; transforming the operational intent into mechanical parameters through multi-time-window counting and piecewise mapping; solving the equilibrium vehicle speed in the drag equation containing wake geometric constraints using a numerical iterative method; and updating seven types of metabolic state variables in parallel within each bounded simulation substep, with mutual constraints through preset cross-coupling relationships. This solves the technical problems of existing motion simulation schemes, such as the lack of interactive closed loops, the absence of cross-system coupling constraints, and the inability to calculate in vivo reserve states.

[0034] Example 1 Figure 1 This is a schematic diagram of the overall architecture of the real-time coupled simulation method for multiple subsystems provided in Embodiment 1 of the present invention. Figure 1 As shown, this method executes steps S1 to S6 sequentially for each rider within each simulation substep with a limited time step, and forms a closed loop through the feedback from S6 to S1.

[0035] This embodiment uses a Node.js backend to run the simulation engine, while the frontend collects click events and renders the dashboard. Core mechanical constants are managed uniformly by the simulation engine: system mass 88 kg (rider 80 kg + vehicle 8 kg), air density 1.226 kg / m³, effective frontal area coefficient CdA 0.28, rolling resistance coefficient Crr 0.0045, and transmission efficiency η 0.96. The simulation substep time step dt does not exceed 0.2 s, and the input time step is limited before it can be used for state updates.

[0036] Step S1 is the discrete operation acquisition layer, executed by the operation acquisition module 101. The operation acquisition module 101 acquires the discrete operation event stream generated by the student through clicks, touch screens, or buttons, counts the discrete operation events in the first, second, and third time windows respectively, obtains the count values ​​tapCount1, tapCount2, and tapCount3 for each time window, and records the time since the last operation event (secondsSinceLastTap).

[0037] Specifically, the first time window has a length of 1.2 s, the second time window has a length of 2.5 s, and the third time window has a length of 10 s. Short windows can capture sudden changes in sprint decisions, while long windows can represent continuous load states, thus making the analysis of operational intentions more accurate. The time elapsed since the last operational event is limited before it can be used for subsequent state updates to avoid numerical integration instability.

[0038] Step S2 is the intent parsing layer, executed by the intent parsing module 102. The intent parsing module 102 calculates the desired cadence based on the count values ​​of each time window using a piecewise mapping function, and obtains the sprint intent value sprintIntent based on the rate jumps of the count values ​​of each time window through gating and smoothing. The value of sprintIntent ranges from 0 to 1.

[0039] Specifically, the target cadence is mapped to three segments based on the third time window count value tap10: when tap10 is no more than 20, the target cadence is approximately tap10 multiplied by 4.25; when tap10 is greater than 20 but no more than 40, the target cadence is 85 plus the difference between tap10 and 20 multiplied by 1.50; when tap10 is greater than 40, the target cadence is 115 plus the difference between tap10 and 40 multiplied by 4.00, in rpm. The high-frequency segment corresponds to the burst sprint range.

[0040] The sprint intent value is detected as follows: when the rate of operation events within the first time window exceeds the set sprint judgment threshold (approximately 5 times / second), sprint intent gating is triggered, and continuous sprint intent values ​​are obtained after first-order low-pass filtering. This gating mechanism ensures that threshold operations of 3 to 4 times / second will not be misjudged as genuine sprints, thereby guaranteeing the accuracy of intent parsing.

[0041] Step S3 is the vehicle mechanics layer, executed by the mechanics calculation module 103. The mechanics calculation module 103 calculates the actual cadence based on the target cadence using a first-order inertial relaxation method, reflecting the inertial characteristic that "cadence is not instantaneously equal to intention", which includes the asymmetric rate of acceleration and deceleration and the neuromuscular penalty.

[0042] The mechanical calculation module 103 synthesizes the steady-state power steadyPower, the startCost term startCost, and the sprintSurge term to obtain the synthesized power P. Among them, the sprintSurge term is limited by the anaerobic availability, which is synthesized from the phosphagen reserve and the anaerobic glycolysis reserve; it is also limited by finger fatigue, and the power upper limit is about 1500 W. Specifically, the synthesis power is calculated as P=clamp(42+255×cadenceIntensity^2.12+120×speedDrag+365×accelerationStress×tapIntensity+790×sprintIntent^1.10×anaerobicAvailability×neuromuscularPenalty,0,1500); anaerobic availability=clamp(0.58×phosphagen power support+0.42×glycolysis power support,0.30,1); finger fatigue is accumulated by finger load, and the neuromuscular fatigue coefficient is reduced by up to 14%.

[0043] Step S4 is the aerodynamics and numerical solution layer, executed by the drag solution module 104. The drag solution module 104 first obtains the geometric parameters of the continuous rider queue ahead, including the number of riders ahead, wheel spacing, lateral alignment, and pace difference, and determines the wake drag reduction rate based on these parameters. This drag reduction rate only applies to the air resistance term.

[0044] Specifically, the wake chain geometry determination rules are as follows: a valid wake chain is formed only when there are at least three consecutive riders moving in the same direction ahead, and the wheel spacing along the queue is less than 1.5 m, laterally aligned, and with similar paces. Base drag reduction rates of 46%, 53%, 57%, and 59% are given for 3, 4, 5, and 6 or more riders ahead, respectively, and then capped at 60% after adjustments based on three factors: the spacing factor, the alignment factor (0.72 plus 0.28 multiplied by the average alignment), and the pace factor (1 minus 0.25 multiplied by the speed difference correction term).

[0045] The resistance solution module 104 then solves the balance equation between power and resistance. The resistance-power equation is: The effective frontal area CdA_eff is equal to CdA multiplied by 1 minus the wake drag reduction ratio. A bisection method is used for 28 iterations within the speed range of 0 to 150 km / h, with the convergence condition being that the difference between drag power and effective power is less than 0.1 W, to obtain the equilibrium speed. When the rider is coasting, the coasting deceleration is calculated as the sum of rolling resistance and aerodynamic drag.

[0046] Step S5 is the metabolic coupling layer, executed by the metabolic coupling module 105. Within the same simulation substep, the metabolic coupling module 105 updates the following seven types of metabolic state variables in parallel, with each metabolic state variable constrained by a preset cross-coupling relationship.

[0047] The first type of state quantity is phosphagen reserve (cpReserve). Phosphagen reserve is depleted according to oxygen deficit, accelerated stress, sprint surge, and glycolytic depletion signals, and is restored in stages according to exercise intensity. When phosphagen reserve drops to zero, the metabolic coupling module 105 enters a locked state (cpLocked), during which operation input is prohibited and cleared; when phosphagen reserve recovers to 80%, the locked state is released, and trampling operation resumes.

[0048] The second type of state variable is the anaerobic glycolytic reserve (glyReserve). The anaerobic glycolytic reserve is consumed based on the benchmark of "depletion in approximately 60 seconds at full intensity," and replenished gradually during low-intensity or inertial periods. Specifically, the glycolytic depletion rate (glyDrain) is calculated as 100 divided by 60 and then multiplied by the exercise load to the power of 1.12.

[0049] The third state quantity is blood lactate concentration. Blood lactate concentration equals the amount produced minus the amount eliminated; the production rate increases with the increase of glycolytic flux and sprint intention value, while the elimination rate is the largest in the active recovery power range and is inhibited during high-intensity exercise; blood lactate concentration is limited to the physiological range of 1.0 to 15.5 mmol / L.

[0050] The fourth type of state quantity is aerobic fraction. The aerobic fraction is given as a target fraction according to the work intensity, and is adjusted downward by oxygen deficit, glycolytic reserve, acceleration and sprint penalty items; during the recovery period, a higher aerobic fraction is maintained even at low power. The fifth type of state quantity is heart rate, which is composed of the immediate drive component, the slow component (cardiovascular drift on the scale of tens of seconds), and the post-exercise maintenance component, with a resting heart rate of 80 bpm and a maximum heart rate of 200 bpm.

[0051] The sixth type of state quantity is the recovery oxygen debt epoc, used to track internal metabolic work after strenuous exercise, thus reproducing the physiological phenomenon of "low power but still high heart rate and aerobic energy supply ratio". The seventh type of state quantity is the energy substrate composition, which includes five substrates: phosphagen, muscle glycogen, blood glucose, free fatty acids, and mitochondrial lipids. It is interpolated piecewise according to exercise intensity and normalized to 100% after multiple corrections.

[0052] The pre-defined cross-coupling relationships include at least one of the following: when anaerobic glycolysis reserves fall below a set depletion threshold, phosphagen reserves are accelerated; when blood lactate concentration increases, the upper limit of aerobic energy contribution is lowered and the lactate clearance rate is reduced; when oxygen debt accumulation is restored, heart rate and aerobic energy contribution are maintained during low-power phases; the sprint surge term is limited by anaerobic availability. These cross-coupling relationships cause the outputs of each metabolic subsystem to mutually constrain each other.

[0053] Step S6 is the multimodal rendering layer, executed by the rendering feedback module 106. The rendering feedback module 106 renders all metabolic state quantities and equilibrium vehicle speed into track animation and a teaching dashboard. Multimodal output includes speed, cadence, power-to-body ratio, heart rate, blood lactate concentration, a bar chart of the five energy supply substrates, and wake turbulence following prompts. The trainee adjusts their operation based on the visual feedback, returning to step S1, forming a complete interactive simulation loop of "operation input → multi-system state calculation → multimodal feedback output".

[0054] Using the scheme in Embodiment 1 above, a discrete operation event flow is used as the sole driving input. Within each bounded simulation substep, cycling mechanics calculations and parallel updates of seven metabolic state variables are simultaneously advanced. The outputs of each subsystem maintain numerical consistency between physics and physiology under any operation sequence. Hysteresis locking is triggered when phosphagen reserves reach zero. The duration of locking is endogenously determined by the physiological recovery rate of phosphagen reserves, unlike fixed-time penalties.

[0055] Example 2 Based on Example 1, this embodiment further implements the externally supplied delayed absorption pool mechanism and the normalized calculation of the substrate for five-way power supply.

[0056] The external supply trigger conditions are: the trainee clicks the supply button and the remaining usage count is greater than 0, and the time since the last use is greater than the set minimum interval of 6 seconds. The number of times external supply can be used in each simulation cycle does not exceed the set limit (no more than 2 times per cycle), and the time interval between two adjacent uses is not less than 6 seconds.

[0057] Exogenous supplementation does not instantaneously boost reserves; instead, it enters the variable absorption pool in the gut. At each step, it is transferred from the absorption pool to anaerobic glycolysis reserves at a set absorption rate k_abs (approximately 2.4% / s), while simultaneously increasing the proportion of blood glucose in the energy substrate composition. This delayed absorption pool mechanism simulates the delayed absorption process in the gut during real-world nutrient supplementation, ensuring that the increase in the proportion of exogenous blood glucose for energy and the recovery of glycolytic reserves both exhibit physiologically delayed characteristics.

[0058] The normalized calculation method for the composition of the five energy-supplying substrates is as follows: the baseline proportions of the five substrates—phosphagen, muscle glycogen, blood glucose, free fatty acids, and mitochondrial lipids—are obtained by interpolation based on exercise intensity. Lactate threshold pressure correction terms, sprint state correction terms, and exogenous supply correction terms are then added to the baseline proportions, normalized to 100%. During sprinting, the proportion of phosphagen can transiently increase, reaching up to 85% in a short period. It should be noted that the proportion of phosphagen in the energy-supplying substrate composition is a normalized proportion of the transient energy source, which decays exponentially with sprint duration and is modulated by the remaining phosphagen reserve. It and the remaining phosphagen reserve, as an independent metabolic state quantity, are two variables with different dimensions.

[0059] Through this embodiment, the simulation system can present the dynamic changes in the body's energy composition that cannot be directly measured in real time. The energy gel takes effect approximately 40 to 60 seconds after ingestion, increasing the proportion of exogenous blood glucose. The restoration of glycolytic reserves slows the depletion of phosphagen, providing an intuitive mechanism demonstration for exercise physiology teaching.

[0060] Example 3 This embodiment replaces the single-machine, single-user architecture with a server-authoritative, multi-user architecture, based on Embodiment 1. This embodiment supports 2 to 4 students sharing a simulation scenario, using a 4-digit room code to organize multiple users into the same circular track.

[0061] The server-side authoritative scheduling module 201 is communicatively connected to the operation acquisition module 101, the resistance solution module 104, and the metabolic coupling module 105. It is used to uniformly advance the simulation substeps of all users at fixed time intervals (dt=0.1 s). The server-side authoritative scheduling module 201 calculates the wake relation matrix based on the real position coordinates of each user, uniformly calculates the wake geometry parameters between each pair of riders, and then sends the status snapshot to each user's client.

[0062] Non-player riders use a pace strip mechanism to anchor their speed to the last user's current speed: within the pace offset range of the last user's current speed, non-player riders adjust their speed in an accordion-like manner, creating a realistic chasing and retreating effect, rather than a rubber band-like forced following. The server's authoritative scheduling module 201 ensures the time synchronization of each user's simulation substeps in multi-user scenarios, and enables multiple users to share aerodynamic constraints in the same physical scene by uniformly calculating the wake relation matrix.

[0063] The behavior telemetry module 202 is signal-connected to the operation acquisition module 101. It records the operation events and metabolic state data of each user at a fixed sampling interval of 1 second, and folds them into statistics within a 3-second non-overlapping time window, storing them in the telemetry buffer. The output of the behavior telemetry module 202 is only used for teaching analysis and does not participate in simulation calculations. It is isolated from the simulation calculation link to ensure that the telemetry data acquisition does not affect the real-time performance of the simulation engine.

[0064] Example 4 This embodiment provides simulation data verification of a complete teaching scenario to demonstrate the technical effects brought about by the distinguishing technical features described in Embodiments 1 to 3. The simulation uses a fixed random seed 20260715, a time step dt = 0.1s, and the mechanical parameters are the same as in Embodiment 1.

[0065] The sequence of operations is as follows: Warm-up phase (0 to 25 s, 1 time / s) → Endurance phase (25 to 85 s, 2 times / s) → Threshold phase (85 to 140 s, 3.2 times / s) → All-out sprint (140 to 146 s, 6.8 times / s) → Glide recovery (146 to 210 s, external replenishment at t=168 s) → Wake cruise (210 to 255 s, 2.6 times / s, 5 riders ahead) → Finish sprint (255 to 268 s, 5.2 times / s) → Cool-down phase.

[0066] Key Event 1 – Threshold Period (t≈118 s): Work-to-body ratio 3.88 W / kg, heart rate 172 bpm, aerobic energy contribution 66.2%, blood lactate concentration 2.54 mmol / L; both LT1 and LT2 are triggered. The interaction between various metabolic state quantities at this moment is a coupled result that cannot be simultaneously output by a single power-rate formula scheme or a preset parameter driving scheme.

[0067] Key Event 2 – Explosion Lock-off (t≈142.4 s): A full-speed sprint causes the phosphate reserve to reach zero at t≈142.4 s, triggering a lock-off state. The minimum phosphate reserve is 0%, and the lock-off lasts for approximately 6.5 s until the phosphate reserve recovers to 80%. Comparative schemes A (single power-speed formula) and B (CN121641471A preset parameter driven scheme) lack a hysteresis lock-off mechanism and cannot express this dynamic characteristic.

[0068] Key Event 3 – Wake Cruise (t≈235 s): 5 riders ahead, wake drag reduction rate 55.3% (closest wheel distance approximately 0.50 m), balanced speed 43.2 km / h, power-to-body ratio 3.56 W / kg, heart rate 155 bpm, blood lactate concentration 3.28 mmol / L. At the same 300 W power, the single-rider speed is 40.5 km / h, while the wake cruise speed increases to 52.67 km / h (an increase of 12.18 km / h), and the model only changes the air resistance term, not the rolling resistance. Comparison schemes A and B lack a wake cruise determination mechanism based on queue geometry parameters and therefore cannot output this result.

[0069] Key Event 4 – Extreme Values ​​Throughout the Journey: The maximum blood lactate concentration throughout the journey was 5.84 mmol / L; after exogenous supplementation was ingested at t=168s, the proportion of energy supplied by blood glucose increased by about 8%, and the recovery of glycolysis reserves delayed the subsequent depletion of phosphagen.

[0070] Table 1 lists the simulation results of this embodiment with those of Comparative Scheme A and Comparative Scheme B under key teaching events.

[0071] Table 1. Comparison of simulation results between this embodiment and the comparative scheme under key teaching events.

[0072] As can be seen from Table 1, in Embodiment 4 of the present invention, through the synergistic effect of the above steps S1 to S6, the seven types of metabolic state quantities output under the complete operation sequence are consistent with the cycling mechanical parameters, thus reproducing the cross-system coupling dynamics and hysteresis fatigue mechanism that cannot be expressed by a single formula scheme or a preset parameter driven scheme.

[0073] It is understood that the operation input method in the above embodiments is not limited to clicking, touching the screen or pressing the button. It can also be continuous pressing of the directional keys (counted by the equivalent duration of pressing) or voice commands (converted into equivalent operation events by voice recognition). As long as it can generate a discrete event timestamp sequence, it is applicable.

[0074] It is understandable that the above numerical iterative solution method is not limited to the bisection method. Newton's iteration method or table lookup interpolation method can also be used, as long as it can converge to the power-velocity balance point within a finite number of iterations.

[0075] It is understandable that the simulation substep time step is not limited to 0.1 s, but can be configured in the range of 0.001 s to 0.2 s. The smaller the time step, the higher the numerical accuracy, but the greater the computational load.

[0076] Obviously, the specific implementation of the above multimodal rendering layer is not limited to Canvas 2D animation, but can also use WebGL 3D rendering or other graphics interfaces. The specific implementation of the rendering layer does not affect the core simulation logic of steps S1 to S5.

[0077] It is understandable that the network communication method of the above-mentioned authoritative server architecture is not limited to WebSocket real-time push. HTTP long polling or other network communication methods can also be used, as long as the server can keep up with the pace and send state snapshots to each client.

[0078] Example 5 Based on Example 1, this embodiment upgrades the multimodal rendering layer to three-dimensional rendering, expands the wake drag reduction model to a wake cone model with lateral offset, and introduces a non-player rider driven by the tactical decision module. Figure 7 This is a schematic diagram of the transverse cross-section of the wake cone and the entry and exit of the wake shielding zone during lane changes. Figure 8 A flowchart illustrating the tactical decision-making process for non-player riders, categorized by difficulty level.

[0079] In this embodiment, the lateral position of each rider is represented by a continuous scalar "lane," with 1 pixel corresponding to 0.04 m. The lateral movement range is ±64 pixels, approximately ±2.6 m. The trainee outputs the lane change command "laneDir" via the left / right movement buttons, with values ​​of -1, 0, or +1. The rider responds to the lane change command at a set lateral movement speed of 30 pixels / second (approximately 1.2 m / s). When the longitudinal overlap between two riders is less than 1.7 m, their lateral distance is only allowed to increase, not decrease to less than 13 pixels (approximately 0.52 m). When the lateral overlap between two riders is less than 10 pixels (approximately 0.4 m), the longitudinal position of the rear rider is clamped 2.0 m behind the rear wheel of the front rider, forming a wheel blocking constraint, reproducing the blocking mechanism in real riding where the rear wheel cannot pass through the front wheel.

[0080] The drag reduction calculation method of the wake cone model is as follows: For each rider in front with a longitudinal spacing gap ranging from 0.3 m to 14 m, the longitudinal weight gapW = clamp(1-(gap-0.3) / 13.7,0,1), and the lateral weight laneW = clamp(1-max(0,laneM-0.25) / 0.85,0,1), where laneM is the lateral offset between the rider in front and the current rider; the rider-by-rider weight w = gapW × laneW, riders in front with w greater than 0.05 are included in the number of riders to be shielded, and the effective shielding intensity eff is obtained by summing the w values; the wake drag reduction rate dragReduction = clamp(0.42×eff^0.7,0,0.65), which only reduces the effective frontal area CdA_eff = CdA×(1-dragReduction). Changing lanes to the left or right alters the rider's lateral offset, allowing the rider to actively enter another rider's wake cone for shielding or exit the wake cone to expose themselves to clean airflow. This transforms lateral aerodynamic exposure into tactical degrees of freedom that can be manipulated by the rider in real time. The wake signal, after first-order inertial smoothing (rise rate 4.6, fall rate 2.2), is applied to the aerodynamic power term, which is calculated as 0.5 × 1.226 × CdA_eff × v³.

[0081] When a rider is confined to a position 2.0 m to 2.6 m directly behind the rider in front due to wheel blockade, the confinement amount (boxAmount) is calculated using linear interpolation, ranging from 0 to 1. Exemptions are granted by multiplying the sprint intention value, the consumption of phosphagen reserves by acceleration stress, and the glycolytic click load by 1 minus 0.9 times the confinement amount, respectively. The actual cadence is limited by a soft top of 162-76 × √boxAmount rpm. This confinement exemption mechanism reproduces the physiological-tactical coupling that prevents riders from wasting phosphagen reserves when pedaling hard while pinned behind the front wheel.

[0082] In this embodiment, the speed calculation uses a bisection method to perform 28 iterations within the search range of 0 to 150 km / h. No additional speed cap is set; the balanced speed is naturally self-limited by aerodynamic drag increasing cubically with speed. The gear ratio constraint maps the actual cadence to a soft speed limit of 0.46 km / h / rpm, allowing for an additional 8.5 km / h relaxation of the sprint intention value. It should be noted that the implementation method described in Embodiment 1, which categorizes the basic drag reduction table based on the number of riders ahead and includes three correction coefficients, still applies to the free exploration mode. The wake cone model in this embodiment is applicable to both racing and multi-user modes; these are two parallel implementation methods.

[0083] Non-player riders are driven by the tactical decision-making module, reusing the same mechanical calculations and metabolic state updates within the same simulated substep as trainees, without receiving any power boost. The tactical decision-making module outputs a motion intensity value u at re-decision intervals, which decrease from 1.2 s to 2.5 s to 0.4 s to 0.8 s depending on the difficulty level. The motion intensity value is converted into a virtual click rate using an inverse function of the target cadence segmentation mapping, where the target cadence is equal to 55 plus 120 multiplied by u. After generating a virtual operation event stream that is completely identical to the fields reported by the trainee's client, it is sent to the same simulation kernel.

[0084] The difficulty levels are divided into four tiers: Beginner, Intermediate, Advanced, and Devil. The parameters are based on perceived noise (0.34 to 0.02), reaction delay (4.0 s to 0.6 s), following skills, blocking skills, chasing skills, attacking skills, supply skills, sprinting skills (0 to 1.00), self-discipline coefficient (0.22 to 0.96), minimum protection limit of phosphate reserves (10% to 42%), and team role configuration. The Beginner tier includes two beginner profiles: Casual Riding and Repeated Exhaustion Riding. The Advanced and Devil tiers involve tail-following, chasing, guarding, probabilistic attacks, 8-second leader rotation, sprinting train 520 m before the finish line, predictive supply, active blocking and counter-blocking lane-changing tactics. The planned sprint distance is reduced from 170 m to 280 m in the Beginner tier to 28 m to 48 m in the Devil tier. All tactical output is constrained by physiological gates: when phosphagen reserves fall below the protective lower limit plus 14 times the self-discipline coefficient, the exercise is forcibly halted for recovery; when blood lactate concentration exceeds the limit or glycolytic reserves are depleted, the exercise intensity is proportionally reduced. Differences in difficulty levels only reflect differences in the quality of tactical decision-making and do not include power bonuses.

[0085] The 3D rendering layer presents the track, rider, wake occlusion area, and lane change maneuvers from a pseudo-3D perspective; the specific implementation of the rendering layer does not affect the core simulation logic of steps S1 to S5.

Claims

1. A real-time coupled simulation method for multiple subsystems, characterized in that, Within each time-limited simulation substep, the following steps are performed sequentially on the rider: Collect the discrete operation event stream generated by the trainees, count the discrete operation events in the first time window, the second time window and the third time window respectively, obtain the count value of each time window, and record the time since the last operation event; The target cadence is calculated using a piecewise mapping function based on the count values ​​of each time window; the sprint intention value is obtained by gating and smoothing based on the rate jumps of the count values ​​of each time window. Based on the target cadence, the actual cadence is calculated using a first-order inertial relaxation method; the steady-state power is calculated based on the actual cadence, and the steady-state power is combined with the starting cost term and the sprint surge term to obtain the composite power, wherein the sprint surge term is limited by anaerobic availability, which is synthesized from phosphagen reserves and anaerobic glycolysis reserves; Obtain the geometric parameters of the preceding continuous rider queue, including the number of riders in the preceding continuous queue, wheel spacing, lateral alignment, and pace difference; The wake drag reduction ratio is determined based on the geometric parameters, and the wake drag reduction ratio only applies to the air resistance term; the balance speed is obtained by numerically iterating through the power and drag balance equations based on the combined power and the effective frontal area corrected by the wake drag reduction ratio. Within the same simulation substep, multiple metabolic state variables, including phosphagen reserve, anaerobic glycolysis reserve, blood lactate concentration, aerobic energy supply ratio, heart rate, recovery oxygen debt, and energy substrate composition, are updated in parallel. The metabolic state variables are mutually constrained according to a preset cross-coupling relationship. When the phosphagen reserve drops to zero, it enters a locked state and operation input is prohibited until the phosphagen reserve recovers to a set proportional threshold and the lock is released. The multiple metabolic state quantities and the equilibrium vehicle speed are output in a multimodal manner, and the output results are used as feedback for the trainee to adjust the operation, thus returning to the acquisition step to form a closed loop.

2. The multi-subsystem real-time coupled simulation method according to claim 1, characterized in that, The first time window has a window length of 0.8 s to 1.5 s, the second time window has a window length of 2 s to 3 s, and the third time window has a window length of 8 s to 12 s; the time step of the simulation substep does not exceed 0.2 s, and the time elapsed since the last operation event is limited before it can be used for state updates; the sprint intention value is detected by: when the rate of operation events within the first time window exceeds the set sprint judgment threshold, sprint intention gating is triggered, and continuous sprint intention values ​​are obtained through first-order low-pass filtering; the segmented mapping of the target cadence includes three intervals: low frequency, mid frequency, and high frequency, with different mapping slopes for each interval, and the high frequency interval corresponding to the sprint interval.

3. The real-time coupled simulation method for multi-subsystems according to claim 1, characterized in that, An effective wake chain is determined when there are at least three riders ahead, the wheel spacing of each successive rider is less than a set wheel spacing threshold, the lateral offset is less than a set lateral threshold, and the pace difference is less than a set pace threshold. The wake drag reduction rate increases with the number of riders ahead and tends to saturate, and is capped after being corrected by three factors: wheel spacing coefficient, alignment coefficient, and pace coefficient. The wheel spacing threshold is 1.5 m. When any wheel spacing is not less than the wheel spacing threshold or when there are fewer than three riders ahead, the wake drag reduction rate is zero. The capped value of the wake drag reduction rate does not exceed 60%. The numerical iterative solution adopts the bisection method, performing a finite number of iterations within a preset speed search interval. The convergence condition is that the difference between the drag power and the effective power is less than a preset error threshold. When the rider is in a coasting state, the gliding deceleration is calculated based on the sum of rolling resistance and aerodynamic resistance.

4. The multi-subsystem real-time coupled simulation method according to claim 1, characterized in that, The preset cross-coupling relationships include: when the anaerobic glycolysis reserve is below a set depletion threshold, the phosphagen reserve is accelerated; when the blood lactate concentration increases, the upper limit of the aerobic energy supply ratio is lowered and the lactate clearance rate is reduced; when the oxygen debt accumulates during recovery, the heart rate and the aerobic energy supply ratio are maintained during the low-power phase; the sprint surge term is limited by the anaerobic availability; the blood lactate concentration is updated as follows: the blood lactate generation rate increases with the increase of anaerobic glycolysis flux and the sprint intention value; the blood lactate clearance rate is maximum in the active recovery power range and is inhibited during high-intensity exercise; the blood lactate concentration is limited to a preset physiological range; the range of the set ratio threshold is 60% to 90%; during the locked state, the operation input in the acquisition step is cleared, and the stepping operation is prohibited.

5. The real-time coupled simulation method for multi-subsystems according to claim 1, characterized in that, Also includes: Receive external supply trigger command; store the equivalent amount of the external supply in the absorption pool variable, replenish the anaerobic glycolysis reserve at a set absorption rate and increase the proportion of blood glucose in the energy supply substrate composition; the number of times the external supply is used in each simulation cycle does not exceed the set upper limit, and the time interval between two adjacent uses is not less than the set minimum interval. The energy-supplying substrates consist of five substrates: phosphagen, muscle glycogen, blood glucose, free fatty acids, and mitochondrial lipids. The baseline proportion of each substrate is obtained by segmented interpolation based on exercise intensity, and then normalized to 100% after adding lactate threshold pressure correction, sprint state correction, and exogenous supply correction. The proportion of phosphagen can increase transiently during sprint.

6. The real-time coupled simulation method for multi-subsystems according to claim 1, characterized in that, The geometric parameters also include the lateral offset between the rider in front and the rider in question; the method for determining the wake drag reduction rate based on the geometric parameters is as follows: for each rider in front with a longitudinal distance between 0.3 m and 14 m, calculate the longitudinal weight and the lateral weight. The longitudinal weight is 1 when the longitudinal distance is 0.3 m and decreases linearly with the increase of the longitudinal distance, and is 0 when the longitudinal distance reaches 14 m. The lateral weight is 1 when the lateral offset does not exceed 0.25 m and decreases linearly with the increase of the lateral offset, and is 0 when the lateral offset reaches 1.10 m. The product and sum of the longitudinal weight and the lateral weight of each rider in front are used to obtain the effective shielding strength; the wake drag reduction rate is 0.42 multiplied by the effective shielding strength raised to the power of 0.7, and is capped at no more than 65%; wherein, in response to the trainee's left or right lane change command, the rider's lateral offset is changed, causing the rider to enter or exit the wake shielding zone of the rider in front.

7. The real-time coupled simulation method for multi-subsystems according to claim 6, characterized in that, The lateral position of each rider is represented by a continuous scalar and responds to the lane change command at a set lateral movement speed; when the longitudinal overlap between two riders is less than 1.7 m, the lateral distance between the two riders is prohibited from being reduced to less than 0.52 m. When the lateral overlap between the two riders is less than 0.4 m, the longitudinal position of the rider behind is clamped to 2.0 m behind the rear wheel of the rider in front, forming a wheel blocking constraint.

8. The real-time coupled simulation method for multi-subsystems according to claim 7, characterized in that, When the rider is restrained by the wheel blockade and is positioned directly behind the rider in front, the blocking amount is calculated linearly within the range of 2.0 m to 2.6 m directly in front, with the value of the blocking amount ranging from 0 to 1. The sprint intention value, acceleration stress consumption, and glycolytic click load are each multiplied by 1 minus 0.9 times the blocking amount for exemption, and the actual cadence is limited by the value obtained by multiplying 162 minus 76 by the arithmetic square root of the blocking amount.

9. The real-time coupled simulation method for multi-subsystems according to claim 1, characterized in that, It also includes non-player riders; the non-player riders are driven by the tactical decision module: the tactical decision module outputs motion intensity values ​​at re-decision intervals, the motion intensity values ​​are converted into virtual operation event streams by the inverse function of the piecewise mapping function, the virtual operation event streams reuse the mechanical calculations and metabolic state updates within the same simulation substep as the trainee's discrete operation event streams; the tactical decision module configures perceived noise, reaction delay, re-decision intervals, following skills, blocking skills, chasing skills, attacking skills, supply skills, sprinting skills, self-discipline coefficient, phosphate reserve protection lower limit, and team role parameters according to difficulty level; The difference between different difficulty levels is only reflected in the quality of tactical decision-making, and does not include power bonuses.

10. A real-time coupled simulation system for multiple subsystems, comprising a memory and a processor, characterized in that, Also includes: The operation acquisition module is used to acquire discrete operation event streams generated by trainees, count the discrete operation events in the first time window, the second time window and the third time window respectively, obtain the count value of each time window, and record the time since the last operation event. The intent parsing module is signal-connected to the operation acquisition module. It is used to calculate the target cadence based on the count values ​​of each time window through a piecewise mapping function, and to obtain the sprint intent value through gating and smoothing based on the rate jump of the count values ​​of each time window. The mechanical calculation module is signal-connected to the intent parsing module and is used to calculate the actual pedal frequency based on the target pedal frequency using a first-order inertial relaxation method. The module combines the steady-state power with the starting cost term and the sprint surge term to obtain the composite power. The sprint surge term is limited by anaerobic availability, which is synthesized from phosphagen reserves and anaerobic glycolysis reserves. The drag calculation module is connected to the mechanical calculation module by signal. It is used to obtain the geometric parameters of the continuous rider queue ahead and determine the wake drag reduction rate that only affects the air resistance term. Based on the combined power and the effective frontal area corrected by the wake drag reduction rate, the balance equation of power and drag is solved numerically and iteratively to obtain the balance speed. The metabolic coupling module is signal-connected to the resistance solving module and is used to update multiple metabolic state variables in parallel within the same simulation substep, including phosphagen reserve, anaerobic glycolysis reserve, blood lactate concentration, aerobic energy supply ratio, heart rate, recovery oxygen debt and energy substrate composition. The metabolic state variables are mutually constrained according to a preset cross-coupling relationship. When the phosphagen reserve drops to zero, it enters a locked state to prohibit operation input until the phosphagen reserve recovers to a set proportional threshold and then the lock is released. The rendering feedback module is signal-connected to the metabolic coupling module and is used to output the multiple metabolic state quantities and the balance vehicle speed in a multimodal manner, and transmit the output results as feedback signals to the operation acquisition module to form a closed loop. The server authoritative scheduling module is communicatively connected to the operation acquisition module, the resistance solution module and the metabolic coupling module, and is used to uniformly advance the simulation substeps of multiple users at fixed time intervals; The server's authoritative scheduling module calculates the wake relationship matrix based on the real location coordinates of each user and sends the status snapshot to each user's client; non-player riders use a pace strip mechanism to anchor the current speed of the last user. The behavior telemetry module is connected to the operation acquisition module and is used to record the operation events and metabolic state data of each user at a fixed sampling interval, and store the data in the telemetry buffer as statistics in a non-overlapping time window. The output of the behavior telemetry module is only used for teaching analysis and does not participate in simulation calculations. The tactical decision-making module, signal-connected to the mechanics calculation module, is used to generate virtual operation event streams for non-player riders according to difficulty levels. These virtual operation event streams reuse the mechanics calculations and metabolic state updates within the same simulation substep as the trainees' discrete operation event streams. The tactical decision-making module configures perceptual noise, reaction delay, re-decision interval, following skills, blocking skills, pursuit skills, attack skills, supply skills, sprint skills, self-discipline coefficient, phosphate reserve protection lower limit, and team role parameters according to difficulty levels. The differences between different difficulty levels only reflect differences in the quality of tactical decisions and do not include power bonuses.

Images