Dynamic torque distribution system based on road condition energy recovery

Abstract

The invention provides a dynamic torque distribution system based on road condition energy recovery. The dynamic torque distribution system comprises a road condition sensing modeling module, an energy recovery evaluation module, a dynamic torque decision module and a torque distribution execution module. The road condition sensing modeling module acquires data through multiple sensors such as an inertia measurement unit and a wireless tire pressure monitoring unit, and constructs a short-term road resistance model containing gravity, rolling and air resistance components through fusion processing; the energy recovery evaluation module predicts future travel theoretical recoverable energy based on a discrete time state space model; the dynamic torque decision-making module is combined with the motor safety boundary and the battery charging power to solve an optimization function with maximization of the energy recovery efficiency as a target, and an optimal torque instruction sequence is obtained; the torque distribution execution module converts the instruction into a control signal and tracks and controls the electromagnetic torque; and a fault diagnosis and self-adaptive control module is additionally arranged to deal with abnormal resistance, so that the robustness of the system is improved, and the endurance and driving safety of the electric bicycle can be greatly improved.

Description

Technical Field

[0001] This invention relates to the field of vehicle power control technology, and more specifically to a dynamic torque distribution system based on road condition energy recovery. Background Technology

[0002] With the continuous development of electric bicycle technology, energy recovery systems are becoming increasingly important in improving range and energy efficiency. However, existing energy recovery technologies still have shortcomings in dynamic torque distribution and road condition adaptability, limiting their effectiveness in complex operating conditions.

[0003] A search revealed a patent, CN111016873B, which discloses an energy management method and system based on a steady-state engine strategy. This patent optimizes energy management and improves fuel economy by acquiring the vehicle's current driving conditions and combining a dynamic SOC balancing strategy, an adaptive start-stop strategy, a torque distribution strategy, and an economical shift line strategy. However, this technical solution is mainly aimed at hybrid vehicles and does not fully consider the energy recovery needs of electric bicycles under different road conditions. Its torque distribution strategy relies on fixed driving condition classifications and cannot dynamically adjust torque distribution in real time according to changes in road conditions, which may lead to low energy recovery efficiency, especially in complex road conditions where it is difficult to achieve optimal energy recovery results.

[0004] A search revealed a patent, CN111016922B, which discloses an energy management system and method for optimizing torque distribution in a single-motor hybrid power system. This patent establishes a torque distribution strategy model through experimental data acquisition and simulation modeling to achieve optimal torque control. While this technical solution can adapt to various operating conditions, its core focus remains on powertrain efficiency optimization in hybrid electric vehicles, failing to address the specific needs of electric bicycles. Furthermore, the construction of its torque distribution strategy model relies on multi-dimensional operating condition simulation, resulting in high computational complexity. This may make it difficult to operate efficiently in the embedded control system of an electric bicycle, thus affecting real-time performance and practicality.

[0005] In summary, existing energy management and torque distribution technologies still have certain shortcomings in adapting to complex road conditions for electric bicycles, real-time dynamic adjustments, and computational efficiency. Summary of the Invention

[0006] To address the shortcomings of existing technologies, the present invention aims to provide a dynamic torque distribution system based on road condition energy recovery. This system can construct a physical model of the road ahead in real time and accurately through multi-sensor fusion technology. Based on model predictive control theory, it solves and executes the optimal energy recovery torque sequence under multiple constraints of vehicle driving safety and ride comfort, thereby maximizing the system's energy recovery efficiency under complex and variable actual working conditions.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A dynamic torque distribution system based on road condition energy recovery includes: The road condition perception modeling module collects and processes information on the slope of the road ahead of the vehicle, the road surface adhesion coefficient, and the distribution of obstacles in real time through various sensors, and constructs a short-term road resistance model. The energy recovery assessment module calculates the theoretical recoverable energy value for a future journey based on the short-term road resistance model and the current vehicle operating status parameters. The dynamic torque decision module, based on the theoretical recoverable energy value, preset motor safe operation boundary conditions, and the current acceptable charging power of the battery, solves an optimization function with the goal of maximizing the total energy recovery efficiency of the system, and calculates the optimal energy recovery torque command sequence in real time. The torque distribution execution module converts the first instruction value of the optimal energy recovery torque instruction sequence into a physical control signal for the drive motor and performs tracking control on the actual generated electromagnetic torque.

[0008] Furthermore, the road condition perception modeling module includes a sensor array, which comprises: An inertial measurement unit used to collect triaxial acceleration and triaxial angular velocity data of electric bicycles; Wireless tire pressure monitoring unit used to collect real-time tire pressure and temperature of the front and rear tires of electric bicycles; Wheel speed sensor used to collect real-time speed data of electric bicycles; and An airflow sensor used to collect the relative velocity of an electric bicycle with the surrounding air.

[0009] Furthermore, the road condition perception modeling module also includes a data fusion processing unit. The data fusion unit calculates the total vehicle driving resistance for constructing a short-term road resistance model by collecting multiple data from the sensor array. The total vehicle driving resistance is composed of the linear superposition of gravity resistance component, rolling resistance component, and air resistance component.

[0010] Furthermore, the data fusion unit includes a gravity drag component calculation step, a rolling drag component calculation step, and an air drag component calculation step. The gravity drag component calculation step involves fusing the triaxial linear acceleration and triaxial angular velocity using a Kalman filter to calculate the vehicle pitch angle, which serves as the current road slope angle, in real time, and then calculating the gravity drag component based on the vehicle pitch angle. The rolling resistance component calculation step involves calculating the road surface bump index based on the vertical acceleration in the triaxial acceleration and the instantaneous tire pressure value, then determining the dynamic rolling resistance coefficient by querying a preset lookup table based on the road surface bump index and the real-time tire pressure value, and finally calculating the rolling resistance component. The air resistance component calculation step calculates the air resistance component based on the real-time relative velocity between the electric bicycle and the surrounding air.

[0011] Furthermore, the rolling resistance component calculation step includes a bump index calculation strategy. The bump index calculation strategy includes applying a high-pass filter to the signal corresponding to the vertical acceleration to extract the vehicle vibration component caused by road surface unevenness, performing a first-order difference operation based on the instantaneous tire pressure to obtain the tire pressure change rate, and then weighting and fusing the root mean square of the vehicle vibration component within the pre-approval time window with the absolute value of the tire pressure change rate to generate the road surface bump index.

[0012] Furthermore, at the beginning of each control cycle, the energy recovery assessment module predicts the vehicle speed state over the next N time steps using a discrete-time state-space model. The discrete-time state-space model is configured as follows: ; in, Let the velocity be at the k-th time step. To apply the recovery torque to the wheels, For the wheel radius, For the predicted driving resistance at step k, This refers to the total mass of the vehicle.

[0013] Furthermore, the dynamic torque decision module obtains the optimal energy recovery torque command sequence by minimizing a cost function. This cost function includes a recovery power term, a speed tracking error term, and a control quantity change rate penalty term. The cost function is configured as follows: ; in, Let the recovery power be at step k. To predict speed, This is a dynamically set reference safe speed based on current road conditions. These represent the recovery torque at adjacent time steps. These are the weighting coefficients.

[0014] Furthermore, the dynamic torque decision module includes a constraint strategy, which includes torque constraint conditions, speed constraint conditions, and state equation constraints. The torque constraint conditions include, for each time step in the prediction time domain, the value of the recovered torque is between 0 and the maximum generating torque determined by a specific curve of the motor at the current speed, and the torque corresponding to the maximum allowable charging power calculated by the power battery management system based on the current state of charge and temperature. The speed constraint conditions include, for each time step in the prediction time domain, the value of the predicted speed is between a preset absolute safe speed lower limit and a preset absolute safe speed upper limit. The state equation constraints include, the discrete-time state-space model is valid throughout the entire prediction time domain.

[0015] Furthermore, it also includes a fault diagnosis and adaptive control module. The fault diagnosis and adaptive control module continuously compares the actual vehicle acceleration with the theoretical acceleration calculated by the current recovery torque and short-term road resistance model through a state observer. When the absolute value of the residual between the actual vehicle acceleration and the theoretical acceleration exceeds a preset threshold for a preset time after low-pass filtering, it is determined that there is an abnormal resistance that has not been modeled, and the abnormal resistance value is estimated in real time based on the residual.

[0016] Furthermore, the fault diagnosis and adaptive control module includes a control strategy. When abnormal resistance exists, the abnormal resistance is used as a known disturbance to update the discrete-time state-space model to obtain a prediction model. Based on the abnormal resistance value, the fault is classified into multiple levels, and the weight coefficients in the cost function are dynamically adjusted for different levels. The prediction model is configured as follows: ; in, This is an abnormal resistance value.

[0017] The beneficial effects of this invention are as follows: It enables accurate and comprehensive road condition perception, laying a solid foundation for energy recovery. Specifically, it collects multi-dimensional data through a multi-sensor array including an inertial measurement unit and a wireless tire pressure monitoring unit. Combined with data fusion algorithms such as Kalman filtering and high-pass filtering, it accurately calculates key parameters such as road slope and bumpiness. The constructed short-term road resistance model can reflect the three types of resistance components—gravity, rolling, and air—in real time, solving the problem of ambiguous road condition perception in traditional technologies and making energy recovery decisions more targeted. In addition, this invention can maximize energy recovery efficiency while taking into account safety and ride comfort. Specifically, based on model predictive control theory, it predicts vehicle speed state through a discrete-time state-space model, with the total... The optimal cost function is constructed with the goal of maximizing energy recovery efficiency. It integrates the recovered power, speed tracking error, and torque change rate penalty terms, while strictly adhering to the constraints of the motor, battery, and speed. The calculated optimal torque command sequence can achieve a dynamic balance between recovery potential and driving safety, avoiding the problems of low recovery efficiency or driving jerking caused by traditional fixed strategies. Finally, the system of this invention has strong robustness and adaptability to complex and variable working conditions. Specifically, the added fault diagnosis and adaptive control module monitors abnormal resistance in real time through a state observer, dynamically updates the prediction model, and adjusts the weight coefficients of the cost function. It can effectively deal with unmodeled interferences such as braking drag and special road surfaces, solving the defect of poor adaptability of traditional systems under complex working conditions. Attached Figure Description

[0018] Figure 1 This is the overall flowchart of the system in this invention; Figure 2 This is a flowchart of the data fusion process of the road condition perception and modeling module in this invention; Figure 3 This is a flowchart of the dynamic torque decision-making process in this invention. Detailed Implementation

[0019] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Identical components are denoted by the same reference numerals. It should be noted that the terms "front," "rear," "left," "right," "upper," and "lower" used in the following description refer to directions in the accompanying drawings, and the terms "bottom surface," "top surface," "inner," and "outer" refer to directions toward or away from the geometric center of a specific component, respectively.

[0020] Current electric bicycles suffer from a lack of real-time perception of dynamic road conditions in their energy recovery control strategies, leading to a mismatch between the recovered torque and the actual energy recovery potential, thus limiting the overall energy efficiency of the system. Therefore, this invention designs a dynamic torque distribution system based on road condition-based energy recovery. This system can construct a physical model of the road ahead in real time and accurately using multi-sensor fusion technology. Based on model predictive control theory, it solves for and executes the optimal energy recovery torque sequence under multiple constraints of vehicle driving safety and ride comfort, thereby maximizing the system's energy recovery efficiency under complex and changing real-world conditions. Specifically, for example... Figure 1 As shown, it includes a road condition perception modeling module, an energy recovery assessment module, a dynamic torque decision-making module, and a torque distribution execution module.

[0021] The system comprises several modules: a road condition perception modeling module, a road resistance model, and a torque distribution execution module. The road condition perception modeling module collects and processes multiple physical parameters affecting the vehicle's energy state to construct a short-term road resistance model that characterizes the vehicle's driving resistance. The energy recovery potential assessment module uses the short-term road physical model and the vehicle's real-time operating parameters to predict and calculate the theoretical maximum recoverable energy value of the vehicle over a future journey within a preset time domain. The dynamic torque decision module calculates the optimal energy recovery torque command sequence in real time by solving a constrained optimization problem online, which aims to maximize the total energy recovery efficiency of the system, based on the theoretical maximum recoverable energy value and considering preset boundary conditions for safe operation of the drive motor, the current acceptable charging power of the power battery, and the vehicle's ride comfort requirements. The torque distribution execution module accurately converts the first command value of the optimal energy recovery torque command sequence into a physical control signal applied to the drive motor and ensures that the electromagnetic braking torque actually generated by the motor accurately tracks the command value through high-frequency closed-loop feedback control.

[0022] The road condition perception modeling module consists of a sensor array and a data fusion processing unit. The data fusion processing unit is an independent software task. The sensor array includes an inertial measurement unit, a wireless tire pressure monitoring unit, a Hall effect wheel speed sensor, and a microelectromechanical system (MEMS) airflow sensor. The inertial measurement unit is a six-axis sensor module integrating a three-axis MEMS gyroscope and a three-axis MEMS accelerometer. The wireless tire pressure monitoring unit consists of two independent sensor units installed at the front and rear tire valves and a central receiver. The sensor units measure tire pressure and temperature respectively through built-in piezoelectric thin-film pressure sensors and NTC thermistors, and send data packets containing unique identifiers, pressure data, and temperature data to the central receiver every 2 seconds. The Hall effect wheel speed sensor is used to accurately calculate the vehicle's real-time speed. The MEMS airflow sensor is a sensor based on the principle of a hot-film anemometer, which determines the relative airspeed by measuring the change in heat transfer between the heating element and the sensing element caused by airflow.

[0023] like Figure 2 As shown, the data fusion processing unit is configured to execute deterministic algorithm steps to establish a short-time road resistance model, which is used to characterize the total driving resistance of vehicles. And the total driving resistance consists of the gravitational resistance component. Rolling resistance component and air resistance component Linear superposition constitutes, that is: To calculate the gravitational drag component, the data fusion processing unit first uses the triaxial acceleration and triaxial angular velocity data output from the inertial measurement unit and performs fusion processing through an extended Kalman filter. The state vector of this filter is defined as a seven-dimensional vector. ,in to A quaternion representing the vehicle's attitude. For three-axis zero-bias estimation using gyroscopes, this filter uses the attitude obtained from gyroscope integration as the state prediction and uses the gravity vector direction measured by the accelerometer under quasi-static conditions as the observation value to correct the attitude, thereby calculating the vehicle pitch angle in real time that accurately reflects the current road slope angle. The filter's process model integrates the attitude quaternions based on angular velocity data and assumes that the zero bias is a random walk process. Its observation model utilizes the physical constraint that the projection of the gravitational acceleration vector onto the vehicle coordinate system in a static or uniform motion state should equal the accelerometer reading. Through EKF prediction and update iteration, the system can output a high-precision vehicle pitch angle in real time, eliminating high-frequency noise such as engine vibration and road impact, as well as the influence of gyroscope temperature drift. This pitch angle, after initial zero-point calibration, is considered as the real-time slope angle of the current road. Subsequently, the gravitational drag component is calculated based on this pitch angle. ,in, The total mass of the vehicle and rider is pre-stored in non-volatile memory. is the gravitational acceleration constant.

[0024] Meanwhile, in order to calculate the rolling resistance component, the data fusion processing unit utilizes the vertical acceleration signal output by the inertial measurement unit. Instantaneous tire pressure value output by the wireless tire pressure monitoring unit The data fusion processing unit first applies a second-order Butterworth high-pass digital filter to the vertical acceleration signal, effectively filtering out low-frequency components caused by slope changes and overall vehicle acceleration and deceleration, retaining only high-frequency body vibration components caused by minor road surface irregularities. Then, it combines this with the instantaneous tire pressure value to calculate a comprehensive road bumpiness index. According to the road surface bump index and real-time tire pressure A dynamic rolling resistance coefficient is determined by querying a pre-stored two-dimensional lookup table. The horizontal axis of this two-dimensional lookup table represents the road surface roughness index quantized in steps of 0.01g, and the vertical axis represents the tire pressure quantized in steps of 10kPa. The values ​​stored in the table are rolling resistance coefficients obtained by calibration under different working conditions using an experimental bench. This lookup table can be a 16x16 matrix, with its row index corresponding to the discretized roughness index. The column index corresponds to the discretized tire pressure. The values ​​in the matrix are the corresponding In actual operation, the system obtains the accurate value from the lookup table using bilinear interpolation. This model accurately reflects the physical fact that rolling resistance increases significantly on bumpy roads or under low tire pressure. Ultimately, the rolling resistance component is calculated using the following formula: In addition, to calculate the air resistance component, the data fusion processing unit uses real-time relative airspeed measured by airflow sensors. and through Perform calculations, where air density; This refers to the vehicle's drag coefficient. The vehicle's frontal area; all are system parameters that have been pre-calibrated and stored in non-volatile memory.

[0025] The data fusion processing unit calculates the road surface bump index. Specifically, the following three sequential steps are executed: First, the acceleration signal output by the inertial measurement unit, which has been transformed to the vehicle's vertical coordinate system, is processed. A fourth-order Butterworth high-pass digital filter is applied. This filter filters out the acceleration components caused by low-frequency motion such as changes in road slope, thereby accurately extracting the high-frequency vehicle vibration components caused by minor road surface irregularities. The second step is to process the tire pressure signal sequence reported by the wireless tire pressure monitoring sensor. Perform a first-order backward difference operation to obtain the tire pressure change rate. ,in The sampling period is defined by this rate of change, which characterizes the instantaneous deformation of the tire due to road impact. The third step involves using a pre-calibrated weighted fusion function. The vibration component of the vehicle body Root mean square value within the past 1 second time window The absolute value of the rate of change of tire pressure A linear combination is performed to generate the road surface bumpiness index. Among them, the weighting coefficient The dimensionless coefficients obtained from regression analysis of experimental data are embedded in the program code to balance the contributions of the two physical quantities in characterizing road surface conditions.

[0026] Among them, the energy recovery assessment module and the dynamic torque decision module together constitute an optimization decision core based on model predictive control, specifically, with a fixed time step. The system operates cyclically within a prediction time domain of length N steps. At the beginning of each control cycle, the energy recovery assessment module uses the vehicle longitudinal dynamics equations based on a discrete-time state-space model to predict the vehicle speed state sequence over the next N time steps. : ,in, Let the velocity be at the k-th time step. To apply the recovery torque to the wheels, For the wheel radius, For the predicted driving resistance at step k, For the total mass of the vehicle, in this prediction process, it is assumed that the road physical properties remain unchanged within a short time domain of 3 seconds, i.e. ,in The total drag value is calculated by the road condition perception and modeling module at the start of the current cycle. To assess the potential, the energy recovery assessment module first assumes that the total drag value will be within the entire prediction time domain. The value is always zero, meaning that the vehicle's natural speed trajectory is calculated without any braking intervention. Based on this trajectory, it is possible to estimate the sum of the total kinetic and potential energy released by the vehicle due to downhill driving or deceleration within the prediction time domain. This is the theoretical maximum recoverable energy.

[0027] Immediately afterwards, such as Figure 3 As shown, the dynamic torque decision module determines the optimal recovery torque sequence by solving a constrained optimization problem online within the same control cycle. The goal of this optimization is to minimize a cost function. The cost function It is precisely defined as: The summation ranges from k=0 to N-1. Cost function It consists of three parts. The first is the recovery power term, which is used to maximize the recovery power. It is the effective recovery power at step k, here, It is the power generation efficiency of the motor at the current speed and torque. The first is the battery's charging efficiency under the current state of charge, temperature, and charging current; both efficiency values ​​are obtained by querying a two-dimensional efficiency characteristic map stored in flash memory. The second is the speed tracking error term used to ensure driving safety and smoothness. To predict speed, The reference safe speed is dynamically set based on current road conditions, with a weighting factor. It is a positive constant, referring to the safe speed. A decision logic module determines the current slope angle. and road surface bumpiness index Dynamically generated on steep and bumpy downhill sections. The first value is set to a lower value to ensure riding safety; on smooth, flat surfaces, this value will be increased accordingly; the third is a penalty term for the rate of change of control quantity used to suppress drastic changes in regenerative torque. It penalizes drastic changes in regenerative torque between adjacent time steps, thereby ensuring smooth application of braking force and avoiding abrupt jerking sensations for the rider. The weighting coefficient represents the recovery torque at adjacent time steps. It is another positive constant, whose function is to ensure the smooth application and removal of braking force.

[0028] When solving the constrained optimization problem, the dynamic torque decision module is subject to the following set of strict boundary conditions: First, torque constraint: for each time step k in the prediction time domain, the recovered torque... The range of values ​​must satisfy ,in, It is determined by the motor's current speed. The maximum generating torque is determined by the torque-speed characteristic curve, which is pre-stored in memory as a one-dimensional array. The maximum allowable charging power is calculated in real time by an independent power battery management system based on the current state of charge (SOC) and internal temperature (Temp) of the battery, and then converted into a corresponding torque upper limit value. The power battery management system dynamically adjusts this upper limit based on the current SOC, internal temperature, and health status of the battery to prevent overcharging or overheating. Secondly, speed constraint: For each time step k in the prediction time domain, the predicted speed... The range of values ​​must satisfy ,in The pre-set absolute safety speed lower and upper limits are used to prevent the vehicle from stalling or exceeding speed during energy recovery. Third, state equation constraints: predictive model. The condition must hold true for all time steps k across the entire prediction time domain. This constrained optimization problem is ultimately solved within each control cycle by an efficient interior-point method or effective set method solver.

[0029] To enhance the robustness and adaptability of the system, the system of the present invention also includes a fault diagnosis and adaptive control module. This module is configured to continuously compare the actual vehicle acceleration measured and calculated by the Hall effect wheel speed sensor with the actual vehicle acceleration measured by the Hall effect wheel speed sensor through a state observer. Compared with the theoretical acceleration calculated based on the currently applied recovery torque and the short-term road physical model. When the residual between these two After passing through a second-order low-pass filter, if the absolute value of the resistance continuously exceeds a preset threshold for a preset time, the system determines that there is an unmodeled abnormal resistance, such as resistance caused by brake drag, bearing damage, or special road surfaces (e.g., mud, water). At this point, the system activates the abnormal resistance estimator, which adjusts an estimated abnormal resistance torque through a proportional-integral controller. This allows the theoretical acceleration based on the updated model to accurately track the actual acceleration, thereby enabling real-time estimation of the magnitude of the abnormal resistance.

[0030] When the fault diagnosis and adaptive control module detects abnormal resistance, the system automatically and seamlessly switches to adaptive control mode. In this mode, the state prediction equation of the model predictive controller used by the dynamic torque decision module is updated online. This involves incorporating the estimated abnormal resistance as a known deterministic disturbance into the future state prediction model, thereby significantly improving the accuracy of control decisions. Furthermore, the system adjusts the control decisions based on the magnitude of the abnormal resistance. The fault is divided into multiple preset levels, for example, when If the value is below a certain threshold, it is considered a minor anomaly, and the system dynamically increases the cost function. Torque change rate weighting coefficient This makes regenerative braking smoother, improving riding comfort; when If the threshold is exceeded, it is considered a serious anomaly, and the system dynamically increases the speed tracking weight coefficient. And at the same time reduce the reference safe speed This makes the control strategy more focused on strict speed control and driving safety, and sends warning messages to the user interface.

[0031] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principle of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A dynamic torque distribution system based on road condition energy recovery, characterized in that: include: The road condition perception modeling module collects and processes information on the slope of the road ahead of the vehicle, the road surface adhesion coefficient, and the distribution of obstacles in real time through various sensors, and constructs a short-term road resistance model. The energy recovery assessment module calculates the theoretical recoverable energy value for a future journey based on the short-term road resistance model and the current vehicle operating status parameters. The dynamic torque decision module, based on the theoretical recoverable energy value, preset motor safe operation boundary conditions, and the current acceptable charging power of the battery, solves an optimization function with the goal of maximizing the total energy recovery efficiency of the system, and calculates the optimal energy recovery torque command sequence in real time. The torque distribution execution module converts the first instruction value of the optimal energy recovery torque instruction sequence into a physical control signal for the drive motor and performs tracking control on the actual generated electromagnetic torque.

2. The dynamic torque distribution system based on road condition energy recovery according to claim 1, characterized in that: The road condition perception and modeling module includes a sensor array, which comprises: An inertial measurement unit used to collect triaxial acceleration and triaxial angular velocity data of electric bicycles; Wireless tire pressure monitoring unit used to collect real-time tire pressure and temperature of the front and rear tires of electric bicycles; Wheel speed sensor used to collect real-time speed data of electric bicycles; and An airflow sensor used to collect the relative velocity of an electric bicycle with the surrounding air.

3. The dynamic torque distribution system based on road condition energy recovery according to claim 2, characterized in that: The road condition perception modeling module also includes a data fusion processing unit. The data fusion unit calculates the total vehicle driving resistance for constructing a short-term road resistance model by collecting multiple data from the sensor array. The total vehicle driving resistance is composed of the linear superposition of gravity resistance component, rolling resistance component and air resistance component.

4. The dynamic torque distribution system based on road condition energy recovery according to claim 3, characterized in that: The data fusion unit includes a gravity drag component calculation step, a rolling drag component calculation step, and an air drag component calculation step. The gravity drag component calculation step involves fusing the triaxial linear acceleration and triaxial angular velocity using a Kalman filter to calculate the vehicle pitch angle, which serves as the current road slope angle, in real time, and then calculating the gravity drag component based on the vehicle pitch angle. The rolling resistance component calculation step involves calculating the road surface roughness index based on the vertical acceleration in the triaxial acceleration and the instantaneous tire pressure value, then determining the dynamic rolling resistance coefficient by querying a preset lookup table based on the road surface roughness index and the real-time tire pressure value, and finally calculating the rolling resistance component. The air resistance component calculation step calculates the air resistance component based on the real-time relative velocity between the electric bicycle and the surrounding air.

5. The dynamic torque distribution system based on road condition energy recovery according to claim 4, characterized in that: The rolling resistance component calculation step includes a bump index calculation strategy. The bump index calculation strategy includes applying a high-pass filter to the signal corresponding to the vertical acceleration to extract the vehicle vibration component caused by road surface unevenness, performing a first-order difference operation based on the instantaneous tire pressure to obtain the tire pressure change rate, and then weighting and fusing the root mean square of the vehicle vibration component within the pre-approval time window with the absolute value of the tire pressure change rate to generate the road surface bump index.

6. A dynamic torque distribution system based on road condition energy recovery according to claim 1 or 5, characterized in that: At the beginning of each control cycle, the energy recovery assessment module predicts the vehicle speed state over the next N time steps using a discrete-time state-space model. The discrete-time state-space model is configured as follows: ； in, Let the velocity be at the k-th time step. The regenerative torque applied to the wheels, For the wheel radius, For the predicted driving resistance at step k, This refers to the total mass of the vehicle.

7. The dynamic torque distribution system based on road condition energy recovery according to claim 6, characterized in that: The dynamic torque decision module obtains the optimal energy recovery torque command sequence by minimizing a cost function. This cost function includes a recovery power term, a speed tracking error term, and a control variable change rate penalty term. The cost function is configured as follows: ； in, Let the recovery power be at step k. To predict speed, This is a dynamically set reference safe speed based on current road conditions. These represent the recovery torque at adjacent time steps. These are the weighting coefficients.

8. The dynamic torque distribution system based on road condition energy recovery according to claim 7, characterized in that: The dynamic torque decision module includes a constraint strategy, which includes torque constraints, speed constraints, and state equation constraints. The torque constraints include, for each time step in the prediction time domain, the value of the recovered torque is between 0 and the maximum generating torque determined by a specific curve of the motor at the current speed, and the torque corresponding to the maximum allowable charging power calculated by the power battery management system based on the current state of charge and temperature. The speed constraints include, for each time step in the prediction time domain, the value of the predicted speed is between a preset absolute safe speed lower limit and a preset absolute safe speed upper limit. The state equation constraints include, the discrete-time state-space model is valid throughout the entire prediction time domain.

9. The dynamic torque distribution system based on road condition energy recovery according to claim 1, characterized in that: It also includes a fault diagnosis and adaptive control module. The fault diagnosis and adaptive control module continuously compares the actual vehicle acceleration with the theoretical acceleration calculated by the current recovery torque and short-term road resistance model through a state observer. When the absolute value of the residual between the actual vehicle acceleration and the theoretical acceleration exceeds a preset threshold for a preset time after low-pass filtering, it is determined that there is an abnormal resistance that has not been modeled, and the abnormal resistance value is estimated in real time based on the residual.

10. The dynamic torque distribution system based on road condition energy recovery according to claim 9, characterized in that: The fault diagnosis and adaptive control module includes a control strategy. When abnormal resistance exists, the abnormal resistance is used as a known disturbance to update the discrete-time state-space model to obtain a prediction model. Based on the abnormal resistance value, the fault is classified into multiple levels, and the weight coefficients in the cost function are dynamically adjusted for different levels. The prediction model is configured as follows: ； in, This is an abnormal resistance value.

Images