Energy recovery and assistance cooperative control method for electric two-wheeled vehicle

Abstract

The invention discloses an electric two-wheeled vehicle energy recovery and assistance cooperative control method, which belongs to the technical field of electric vehicle control, and comprises the following steps of: generating load prediction characteristics and intention confidence vectors by acquiring vehicle operation data, fusing and constructing master-slave constraint data with phase difference, and calculating the energy recovery and assistance cooperative control of the electric two-wheeled vehicle. Dynamic cutting and compensation distribution are carried out on energy recovery and electric power assistance demands, and a dynamic torque distribution diagram is generated; and a motor smooth control instruction is generated by combining a smooth mapping function simulating human body operation physiological characteristics, and feedback is executed to correct parameters. According to the method, master-slave constraint data with phase difference is constructed to perform dynamic cutting and compensation distribution on energy recovery and electric power assistance demands, and a control strategy of smooth mapping and closed-loop feedback correction for simulating human physiological characteristics is combined, so that the control conflict between power assistance and recovery can be solved, the control smoothness of a vehicle under a complex working condition is improved, and the control reliability of the vehicle is improved. And safety and energy efficiency are improved.

Description

Technical Field

[0001] This invention relates to the field of electric vehicle control technology, and in particular to a method for coordinated control of energy recovery and power assist in electric two-wheeled vehicles. Background Technology

[0002] Electric two-wheelers, as an environmentally friendly and convenient personal transportation tool, rely heavily on the control strategy of their electric drive system. This system not only includes electric assist to power the ride but also increasingly integrates energy recovery, converting the vehicle's kinetic or potential energy into electrical energy during braking or coasting and storing it back in the battery to extend the driving range. Energy recovery and electric assist are two core functions of the electric drive system, and their effective coordinated control directly impacts the vehicle's energy efficiency, driving safety, and riding comfort.

[0003] In existing technologies, the control of energy recovery and electric assist in electric two-wheelers typically employs relatively independent control logic. The activation and assist level of the electric assist system are usually determined based on signals from cadence or torque sensors, outputting through a preset assist level curve. The energy recovery system, on the other hand, is mostly triggered by independent braking signals, such as gripping the brake lever or stopping pedaling in a specific mode. Its recovery intensity is often limited to a few fixed levels or simply correlated with vehicle speed. These two functions are treated as separate modules at the control level, operating independently based on different triggering conditions.

[0004] However, in actual riding, the rider's operational intentions and the vehicle's driving status are complex and constantly changing, often resulting in overlapping trigger conditions for two functions. For example, a rider might lightly brake while going downhill, while simultaneously maintaining pedaling to sustain speed. Under such complex conditions, the aforementioned independent control strategy becomes flawed. The system's control logic may conflict, causing the motor to switch rapidly between assist and braking states, producing a noticeable jerkiness and affecting riding comfort. Furthermore, this simple control strategy struggles to adapt to complex road conditions such as slippery surfaces, failing to adjust the recovery intensity based on real-time traction conditions, thus posing certain safety hazards. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides a method for coordinated control of energy recovery and electric assist in electric two-wheeled vehicles. This method employs master-slave constraint data with phase differences to dynamically tailor and compensate for the energy recovery and electric assist demands. Combined with a control strategy that simulates smooth mapping of human physiological characteristics and uses closed-loop feedback correction, it resolves control conflicts between assist and recovery, improving vehicle handling smoothness, safety, and energy efficiency under complex operating conditions.

[0006] The above objectives can be achieved through the following approach: A method for coordinated control of energy recovery and electric assist in an electric two-wheeler includes: acquiring and processing vehicle operating data to generate load prediction features and intent confidence vectors; fusing the load prediction features and intent confidence vectors and constructing constraints to generate master-slave constraint data with phase difference; dynamically trimming and compensating the energy recovery demand and electric assist demand based on the master-slave constraint data to generate a dynamic torque distribution map; performing a smooth and safe mapping based on the dynamic torque distribution map and a preset mapping function simulating the physiological characteristics of human operating parts to generate a smooth motor control command; executing the smooth motor control command and collecting vehicle state data and energy data after executing the smooth motor control command, and correcting the parameters of the master-slave constraint data and the mapping function.

[0007] Optionally, the step of acquiring and processing vehicle operation data to generate load prediction features and intent confidence vector includes: acquiring the vehicle's cadence and pedal force parameters, and observing load disturbances based on the rate of change of the cadence and pedal force parameters to generate short-term load disturbance features; acquiring the vehicle's braking operation parameters and vehicle tilt angle parameters, and performing intent trend analysis on the gradient and angular velocity of the changes in the braking operation parameters and vehicle tilt angle parameters to generate intent intensity and clarity features; correcting the intent intensity and clarity features with the short-term load disturbance features, and performing load disturbance analysis with the corrected intent intensity and clarity features to generate load prediction features and intent confidence vector.

[0008] Optionally, the step of fusing the load prediction features and the intention confidence vector and constructing constraints to generate master-slave constraint data with phase difference includes: using the intention confidence vector as the dominant factor and fusing the load prediction features to generate an active phase constraint subset defining the power output response speed and torque smoothness; using the load prediction features as the dominant factor and fusing the intention confidence vector to generate a recovery phase constraint subset defining the energy recovery intensity and disturbance rejection stability; calculating the master control phase based on the intention confidence vector and the load prediction features, and determining the phase difference relationship between the active phase constraint subset and the recovery phase constraint subset in terms of time sequence and weight based on the master control phase to generate master-slave constraint data.

[0009] Optionally, generating the recovery phase constraint subset defining energy recovery intensity and disturbance rejection stability includes: obtaining the adhesion risk prediction level from the load prediction features; generating the influence range and constraint strength based on the intention confidence vector and the adhesion risk prediction level; generating an initial constraint subset using the influence range and constraint strength; obtaining real-time vehicle load estimation parameters and calculating the basic recovery boundary; and fusing the initial constraint subset with the basic recovery boundary to obtain the recovery phase constraint subset defining energy recovery intensity and disturbance rejection stability.

[0010] Optionally, generating the dynamic torque distribution map includes: identifying overlapping and conflicting regions between energy recovery demand and electric assist demand in the time domain or torque domain at the current moment; determining the demand side and trimming it according to the master control phase in the master-slave constraint data and the overlapping and conflicting regions to generate trimmed demand features; generating compensation torque commands based on the trimmed demand features and fusing them with the untrimmed demand features to generate the dynamic torque distribution map.

[0011] Optionally, generating the compensation torque command based on the trimmed demand characteristics includes: identifying the demand type corresponding to the trimmed demand characteristics; if it is an energy recovery demand, generating an enhanced recovery pulse as a compensation torque command; if it is an electric assist demand, generating an additional smooth assist torque as a compensation torque command; wherein the amplitude and duration of the compensation torque command are limited within the dynamic safety boundary of the master-slave constraint data.

[0012] Optionally, generating the motor smooth control command includes: mapping the original torque command in the dynamic torque distribution diagram through a preset mapping function that simulates the physiological characteristics of human operating parts to generate a comfortable torque curve; obtaining the vehicle's wheel speed difference parameters and road surface roughness parameters, and converting them into stiffness adjustment factors; dynamically adjusting the response stiffness of the mapping function using the stiffness adjustment factors to correct the comfortable torque curve; and generating the motor smooth control command using the corrected comfortable torque curve.

[0013] Optionally, the step of obtaining the vehicle's wheel speed difference parameters and road surface roughness parameters and converting them into a stiffness adjustment factor includes: obtaining the vehicle's wheel speed difference parameters and road surface roughness parameters; calculating a real-time slip trend index based on the wheel speed difference parameters; calculating the current road surface peak adhesion coefficient based on the mapping relationship between the road surface roughness parameters and a preset adhesion coefficient; and fusing the real-time slip trend index and the current road surface peak adhesion coefficient to calculate the stiffness adjustment factor.

[0014] Optionally, the step of correcting the parameters of the master-slave constraint data and the mapping function includes: collecting the actual energy recovery and consumption after executing the motor smoothing control command and comparing them with a preset set of theoretical values ​​to generate energy efficiency deviation characteristics; collecting the cadence and pedal force parameters after executing the motor smoothing control command and analyzing the convergence stability to generate experience evaluation characteristics; correcting the parameter configuration related to the energy boundary in the master-slave constraint data according to the energy efficiency deviation characteristics, and correcting the parameter configuration related to smoothness in the mapping function according to the experience evaluation characteristics.

[0015] Based on the same inventive concept, this invention also provides an energy recovery and electric assist coordinated control system for an electric two-wheeler. The system includes: a sensing module for acquiring and processing vehicle operating data to generate load prediction features and intention confidence vectors; a constraint generation module for fusing the load prediction features and the intention confidence vectors and constructing constraints to generate master-slave constraint data with phase difference; a dual-ring sensing module for dynamically trimming and compensating the energy recovery demand and electric assist demand based on the master-slave constraint data to generate a dynamic torque distribution map; a safety control module for performing a smooth safety mapping based on the dynamic torque distribution map and a preset mapping function simulating the physiological characteristics of human operating parts, generating smooth motor control commands; and an execution and feedback module for executing the smooth motor control commands and collecting vehicle status data and energy data after executing the smooth motor control commands, and correcting the parameters of the master-slave constraint data and the mapping function.

[0016] Compared with the prior art, the present invention has the following advantages: This invention constructs a master-slave constraint decision-making mechanism based on intent and load prediction, which can determine the priority of energy recovery and electric power assist demand and perform dynamic trimming and compensation. This resolves the conflict between the two in the control logic, ensuring smooth and continuous torque output under complex conditions such as acceleration, deceleration, and cornering. It avoids the jerking and power interruption phenomena that may occur under traditional control methods, thereby improving the vehicle's handling stability and driving quality.

[0017] This invention simulates the physiological characteristics of human body operating parts through a mapping function and dynamically adjusts its response stiffness by combining real-time road condition data. This makes the motor's power response no longer a rigid execution of mechanical commands, but rather more in line with the rider's subjective feelings and physiological habits, improving the naturalness and comfort of human-vehicle interaction and making the riding experience smoother and more comfortable.

[0018] This invention can track the actual energy efficiency of the control strategy and rider feedback over a long period, and accordingly optimize the energy boundary of the master-slave constraint and the smoothness parameters of the mapping function. This learning ability enables the vehicle control system to continuously evolve and adapt to changes in user habits and vehicle conditions, thereby maintaining optimal energy efficiency and the most comfortable riding experience in the long run.

[0019] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart illustrating a method for coordinated control of energy recovery and power assist in an electric two-wheeled vehicle according to an embodiment of the present invention.

[0022] Figure 2 This is a schematic diagram of demand conflict identification and dynamic trimming in an embodiment of the present invention.

[0023] Figure 3 This is a schematic diagram of the comfort torque curve according to an embodiment of the present invention.

[0024] Figure 4 This is a schematic diagram of the structure of an energy recovery and power assist coordinated control system for an electric two-wheeled vehicle according to an embodiment of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] Reference Figure 1 One embodiment of the present invention proposes a method for coordinated control of energy recovery and electric assist in electric two-wheeled vehicles. It adopts a master-slave constraint data with phase difference to dynamically trim and compensate the energy recovery and electric assist requirements, and combines a control strategy that simulates the smooth mapping of human physiological characteristics and closed-loop feedback correction. This method can resolve the control conflict between assist and recovery, and improve the smoothness of vehicle handling, safety and energy efficiency under complex working conditions.

[0027] The method described in this embodiment specifically includes: S1. Acquire and process vehicle operation data to generate load prediction features and intent confidence vectors; Optionally, the step of acquiring and processing vehicle operation data to generate load prediction features and intent confidence vectors includes: The vehicle's cadence and pedal force parameters are acquired, and load disturbance is observed based on the rate of change of cadence and pedal force parameters to generate short-term load disturbance characteristics. The vehicle's braking operation parameters and body tilt angle parameters are obtained, and the intention trend analysis is performed on the change gradient and angular velocity of the braking operation parameters and body tilt angle parameters to generate intention intensity and clarity features. The intent strength and the clarity features are corrected using the short-term load disturbance features, and load disturbance analysis is performed using the corrected intent strength and clarity features to generate load prediction features and intent confidence vectors.

[0028] Specifically, the system quantifies sudden changes in the rider's output power in real time to identify external load variations such as headwinds and uphill climbs. This is achieved by using torque sensors mounted on the bike's bottom bracket or rear hub and cadence sensors on the crank, at a sampling rate of, for example, 50 Hz to 100 Hz, to acquire real-time pedaling force and cadence parameters. The system generates short-term load disturbance characteristics by calculating the rate of change of these two parameters within a short time window. This feature aims to capture load fluctuations not caused by the rider's subjective will, and its quantification process can be represented by the following formula: , in, The final output short-term load disturbance characteristic is a dimensionless evaluation value. and These represent the changes in pedal force and cadence parameters within a very short time period, such as 100 milliseconds. and Preset normalized benchmark values, such as the maximum output torque of the vehicle design and the normal cruising cadence, are used to eliminate the influence of different physical dimensions, so that changes from different sources can be measured uniformly. and This is an adjustable weighting coefficient, whose value is calibrated based on the vehicle model and test data, used to adjust the contribution of pedal force and cadence changes to the final characteristics.

[0029] In parallel, the system initiates trend analysis of the rider's control intentions, interpreting the rider's explicit intention to actively decelerate or steer, which is crucial for determining the timing and intensity of energy recovery. The system obtains braking operation parameters from the brake controller, such as the brake lever opening or the pressure value of the hydraulic braking system, and acquires vehicle lean angle parameters via the onboard inertial measurement unit (IMU). By analyzing the gradient of changes in braking operation parameters, i.e., their derivative with respect to time, the system can determine the urgency of the rider's braking intention. By analyzing the angular velocity of the vehicle lean angle parameters, the system can determine whether the vehicle is entering or exiting a curve. These analytical results are integrated into intention intensity and clarity characteristics, collectively describing the decisiveness and predictability of a control action.

[0030] Finally, to generate the final load prediction features and intent confidence vector for downstream control modules, the system performs a fusion correction process, fusing disturbance observations and intent analysis results to form a unified, high-confidence understanding of the riding environment and the rider's intent. The system first uses the aforementioned short-term load disturbance features to correct the intent strength and clarity features. For example, when a strong uphill load disturbance is detected, even if the rider slightly slows down, the system will reduce the clarity of their "slowing down" intent, because this behavior may not be an active slowdown but a natural response to the slope. The corrected intent strength and clarity features are constructed into an intent confidence vector, which is a data pair containing intensity and clarity values; for example, [0.8, 0.9] represents a braking intent with 80% intensity and 90% clarity. Simultaneously, the system uses this high-confidence intent vector to re-analyze the load disturbance, thereby generating the load prediction features. If the intention confidence vector indicates that the rider is about to brake hard, the system predicts that the subsequent load will be dominated by braking torque and packages the prediction result into a load prediction feature, such as predicting an equivalent braking load of -15 Nm within the next 200 milliseconds for use in subsequent dynamic torque distribution.

[0031] S2. The load prediction features and the intention confidence vector are fused together and constraints are constructed to generate master-slave constraint data with phase difference; Optionally, the step of fusing the load prediction features with the intent confidence vector and constructing constraints to generate master-slave constraint data with phase difference includes: Using the intent confidence vector as the primary driver and incorporating the load prediction features, an active phase constraint subset is generated that defines the power output response speed and torque smoothness. Based on the load prediction features and fused with the intention confidence vector, a subset of recovery phase constraints defining energy recovery intensity and disturbance rejection stability is generated; The master control phase is calculated based on the intent confidence vector and the load prediction features. The phase difference relationship between the active phase constraint subset and the recovery phase constraint subset in terms of time and weight is determined based on the master control phase, and master-slave constraint data is generated.

[0032] Specifically, the system first generates an active phase constraint subset to define the behavioral boundaries of the power assist system based on the rider's subjective intent. The system uses the intent confidence vector as the primary input and load prediction features as secondary input. When the intent confidence vector clearly indicates a high-intensity acceleration intent, for example, a vector value of [0.9, 0.95], the system generates a constraint subset that allows for higher power output response speed and a larger torque change rate; for example, allowing torque to climb from 0 to 80% of its peak value within 100 milliseconds. Simultaneously, the load prediction features, such as predicting a flat road ahead, further confirm the rationality of this constraint. Conversely, if the intent is unclear, this subset will limit the response speed, enhance torque smoothness, and ensure a comfortable riding experience.

[0033] In parallel, the system generates a subset of recovery phase constraints to define the behavioral boundaries of the energy recovery system based on objective vehicle states and environmental loads, prioritizing safety and stability. The system uses load prediction features as the primary input and intention confidence vectors as an auxiliary input. For example, when load prediction features indicate the vehicle is on a long downhill slope with good traction, the system generates a subset of constraints that allows for higher energy recovery intensity. In this case, a slight braking intention reflected in the intention confidence vector will trigger the recovery strategy. However, if load prediction features indicate a slippery road surface, even if the intention confidence vector indicates a strong braking intention, this subset of constraints will strictly limit the maximum recovery torque to ensure disturbance rejection stability and prevent wheel lock-up and sideslip.

[0034] Finally, the master phase is calculated using a weighted fusion function that combines the contributions of the intent confidence vector and the load prediction features: , in, Representing the dominant phase, it is a continuous value between negative one and positive one, used to characterize the dominant trend of the current vehicle state. It is a scalar value of the intent confidence vector after quantization, such as by multiplying intensity and clarity. Positive values ​​represent assist intent, and negative values ​​represent deceleration or braking intent. It is a quantified scalar value of load prediction characteristics. Positive values ​​represent loads that require assistance, such as uphill loads, while negative values ​​represent loads that can recover energy, such as downhill loads. and These are preset weighting coefficients used to adjust the influence of subjective intent and objective load in decision-making. The calculated... If the value is greater than 0.5, the system determines that the current phase is the master control's assist phase. Under this phase, the active phase constraint subset is set as the "master constraint," and its weight is set to a high value, such as 0.9, while the reclaimed phase constraint subset is set as the "slave constraint," and its weight is set to a low value, such as 0.1. Furthermore, its response is subject to a time delay of, for example, 50 to 200 milliseconds. This represents the phase difference relationship in terms of timing and weight. Finally, a complete data packet containing the two constraint subsets and their current weights and timing relationships is generated, forming the master-slave constraint data, which is then output to downstream modules.

[0035] Optionally, the generation of the recovery phase constraint subset defining the energy recovery intensity and disturbance rejection stability includes: Obtain the attachment risk prediction level from the load prediction characteristics; Based on the intent confidence vector and the attachment risk prediction level, the scope of influence and constraint strength are generated. The initial constraint subset is generated using the influence range and constraint strength. Obtain real-time load estimation parameters for the vehicle and calculate the basic recovery boundary. The initial constraint subset is fused with the basic recovery boundary to obtain a recovery phase constraint subset that defines the energy recovery intensity and disturbance rejection stability.

[0036] Specifically, the first step is to parse the adhesion risk prediction level from the load prediction features, quantifying the level of grip that the current road surface may provide. This level is a numerical value, ranging from Level 1 (Excellent) to Level 5 (Poor). It is generated by a pre-trained classification model or fuzzy logic rule base through comprehensive analysis of multi-dimensional information such as historical wheel slip rate, ABS system activation frequency, current ambient temperature, and rain sensor signals. A higher level represents low adhesion risk, and vice versa.

[0037] Subsequently, based on this risk level and the intent confidence vector obtained from the previous level, the system generates the influence range and constraint strength, combining the risk assessment with the rider's intent to initially define the permissible range of the recovery action. The influence range defines the speed range or braking depth at which the constraint takes effect, while the constraint strength is directly related to the maximum permissible recovery torque. For example, even if the intent confidence vector indicates that the rider has a strong braking intent, but the attached risk prediction level is the highest level of 5, the system will still force the constraint strength to be set at a very low level, such as corresponding to a recovery torque of 1 to 3 Newton-meters, and expand the influence range to the entire speed range to minimize wheel slippage.

[0038] Next, the system uses the generated influence range and constraint strength to construct an initial constraint subset. This subset is a data structure containing specific parameters, such as {maximum recovered torque, upper limit of torque growth rate}, where the values ​​are directly mapped from the constraint strength.

[0039] Simultaneously, to establish an inviolable physical safety baseline, the system acquires real-time vehicle load estimation parameters and calculates the basic recapture boundary to determine the maximum braking torque the rear wheels can withstand without instability under the current vehicle dynamic load distribution. The real-time vehicle load estimation parameters primarily refer to the total mass of the vehicle including the rider. This mass can be initially estimated through the motor's current response at startup or set as a standard value. Basic recapture boundary. Calculated using the following formula: , in, It is the basic recovery boundary torque. It is the vehicle's real-time load estimation parameter, i.e., total mass. It is gravitational acceleration. It is the horizontal distance from the vehicle's center of gravity to the center of the front wheel axle. These are the vehicle's wheelbase and other design parameters. It is the radius of the drive wheel. This is the current peak adhesion coefficient of the road surface, mapped from the adhesion risk prediction level. This formula calculates the recovery torque corresponding to the maximum ground braking force generated by the gravity distributed to the rear wheels, without considering the simplification of dynamic load transfer.

[0040] Finally, the system fuses the initial constraint subset with the basic recovery boundary to obtain the final recovery phase constraint subset, ensuring that the final recovery strategy simultaneously satisfies both intention-risk constraints and physical limit constraints. The fusion process typically adopts the principle of taking the more stringent value of the two; that is, the final maximum allowable recovery torque is taken as the maximum recovery torque value in the initial constraint subset and the basic recovery boundary. The smaller of the two. This generates a subset of recovery phase constraints that both responds to the rider's intentions and fully considers road risks and vehicle physical stability, defining the energy recovery intensity and disturbance rejection stability.

[0041] S3. Based on the master-slave constraint data, dynamically trim and compensate the energy recovery demand and electric assist demand to generate a dynamic torque distribution map. Optionally, generating the dynamic torque distribution map includes: Identify the overlapping and conflicting areas between the current energy recovery demand and the electric assist demand in the time domain or torque domain; Based on the master control phase in the master-slave constraint data and the overlapping conflict area, the demand side is determined and trimmed to generate the trimmed demand features; Compensation torque commands are generated based on the trimmed demand features and then fused with the untrimmed demand features to generate a dynamic torque distribution map.

[0042] Specifically, the system first identifies the overlapping and conflicting areas of the two types of demands, capturing in real time the instant when the opposing torque commands of power assist and regenerative braking occur simultaneously. Within an extremely short control cycle, for example every 10 milliseconds, the system simultaneously checks the electric power assist demand calculated by the cadence and pedal force sensors. and energy recovery requirements determined by the braking system or downhill conditions. .when For positive value When the values ​​are negative and both absolute values ​​are greater than a preset noise threshold, the system determines that there is an overlapping conflict area in the current time domain and torque domain.

[0043] Subsequently, the system performs demand arbitration and dynamic pruning, determining which side's demand should be prioritized in a conflict based on a higher-level strategy, namely, master-slave constraint data. The system reads the master-slave constraint data passed in from the previous level and extracts the master phase. If the master phase The indication is currently in a dominant position, for example... If the value is greater than a positive threshold, such as 0.2, the system determines that electric assist is the primary demand source. At this point, energy recovery demand... The value will be clipped, typically by forcing it to 0, or by clipping it to a minimum value that doesn't produce noticeable drag, according to the constraints. Conversely, if the master phase... Indicate recycling as the dominant factor, for example If the value is less than a negative threshold, such as -0.2, then the electric assist demand is low. The requirements are then tailored accordingly. The output of this process is the tailored requirements and their numerical values, i.e., the tailored requirements characteristics.

[0044] Finally, the system compensates and merges the results based on the trimming, generating a final dynamic torque distribution map. This not only resolves conflicts but also compensates for performance losses due to trimming, optimizing the overall experience. The system first generates a compensation torque command based on the trimmed demand characteristics. Then, the system will process the unpruned demand features, the pruned demand features that may become zero, and the compensation torque command. By performing algebraic superposition, the final target torque at the current moment is generated. : , in, The original required torque retained after arbitration. This represents the new value of the required torque after the trimming. This is the compensation torque generated to smooth the transition or enhance the effect. It is calculated for each control cycle. The values ​​are arranged in chronological order to form a dynamic torque distribution map within a specified time window. This map serves as the input target for the motor controller, guiding the motor to output smooth and logical torque.

[0045] Optionally, generating the compensation torque command based on the trimmed demand characteristics includes: Identify the demand type corresponding to the trimmed demand characteristics. If it is an energy recovery demand, generate an enhanced recovery pulse as a compensation torque command. If electric power steering is required, an additional smooth power steering torque is generated as a compensation torque command. The magnitude and duration of the compensation torque command are limited within the dynamic safety boundary of the master-slave constraint data.

[0046] Specifically, such as Figure 2 As shown, the process begins with the precise identification of the type of demand to be cut. The system checks the cut demand characteristics output from the previous step, which include the type of demand to be cut, namely energy recovery demand or electric assist demand, as well as its original value before being cut.

[0047] If the system identifies the suppressed demand as an electric assist demand, which typically occurs during braking or downhill regenerative braking, the system will generate additional smooth assist torque as compensation. This allows the vehicle to smoothly and seamlessly transition back to assisted mode after the regenerative braking-dominated situation ends. For example, when a rider resumes pedaling after braking downhill, the system will add a moderately large, short-duration additional smooth assist torque (e.g., 100 to 300 milliseconds) to the regular assist torque. This counteracts the slight drag that may occur when the energy recovery system disengages and compensates for the power response delay caused by the temporary suppression of assist demand, ensuring a smooth power transition.

[0048] Conversely, if the system identifies the suppressed demand as an energy recovery demand, this typically occurs when the rider pedals heavily, with assist intention dominant, and a brief braking signal appears. In this case, the system generates an enhanced recovery pulse as compensation. The moment the assist demand weakens, it quickly makes up for the previously suppressed energy recovery opportunity, maximizing energy utilization efficiency. For example, when a rider accelerates hard to overtake and then returns to normal pedaling force, the system immediately generates a short, for example, 50 to 150 milliseconds, negative torque pulse with an amplitude higher than the normal recovery intensity. This enhanced recovery pulse efficiently utilizes the window of higher speed to capture more regenerative energy, but because its duration is extremely short, it does not cause significant drag on the riding experience.

[0049] Crucially, all generated compensation torque commands, whether adding smooth assist torque or enhancing regenerative braking pulses, must be subject to strict boundary constraints in their final output. Upon generating the command, the system immediately retrieves the currently effective dynamic safety boundary from the master-slave constraint data. This boundary defines the maximum positive and negative torque that the motor is allowed to output under the current vehicle conditions, as well as the maximum rate of torque change. The integral effect of the amplitude and duration of the compensation torque command—the torque impulse—must be strictly limited within this dynamic safety boundary to ensure that the compensation action itself does not induce new vehicle instability or exceed the rider's safety expectations.

[0050] S4. Perform a smooth and safe mapping based on the dynamic torque distribution diagram and the preset mapping function of the simulated human body operating component motion physiological characteristics to generate a smooth motor control command. Optionally, the generated motor smooth control command includes: The original torque command in the dynamic torque distribution diagram is mapped through a preset mapping function that simulates the physiological characteristics of human body operating parts to generate a comfort torque curve. Obtain the vehicle's wheel speed difference parameters and road surface roughness parameters, and convert them into stiffness adjustment factors; The response stiffness of the mapping function is dynamically adjusted using the stiffness adjustment factor to correct the comfort torque curve; The modified comfort torque curve is used to generate smooth motor control commands.

[0051] Specifically, such as Figure 3 As shown, the process first performs a comfort mapping on the original torque command, mimicking the physiological curves of the human body when exerting or releasing force, thus avoiding sudden changes in motor torque. The system maps each original torque command point in the dynamic torque distribution diagram... The torque is processed through a pre-defined mapping function that simulates the physiological characteristics of the human body's operating components. This mapping function is essentially a nonlinear filter, such as an sigmoid function or a higher-order polynomial, which smooths the rising and falling edges of the torque command, resulting in a smaller torque change rate at the beginning and end, and a larger rate in the middle.

[0052] , in, It is a point on the comfort torque curve generated after mapping. This represents the mapping function. It is a parameter related to smoothness, which determines the steepness of the curve; These are parameters related to time delay. Both parameters are pre-calibrated based on a large amount of riding data and subjective evaluations to achieve the best comfort experience.

[0053] In parallel, the system initiates calculations of a dynamic adjustment factor for the response stiffness, allowing the motor's "stiffness" to match road conditions in real time. It provides a firm response on smooth surfaces and a softer response on bumpy or slippery surfaces to improve stability and traction. The system acquires wheel speed difference parameters via wheel speed sensors between the non-drive and drive wheels, and obtains road roughness parameters through spectral analysis of acceleration sensor signals or a dedicated road surface recognition algorithm. The wheel speed difference parameter is used to identify minute wheel slippage, while the road roughness parameter directly reflects the degree of road unevenness. These two parameters are fused into a single stiffness adjustment factor. .

[0054] Next, the system uses this stiffness adjustment factor to dynamically adjust the behavior of the mapping function, feeding back road surface information to the torque output characteristics in real time. The system uses the stiffness adjustment factor. To correct mapping functions online The response stiffness parameter in the model. For example, when a rough road surface or a tendency to slip is detected, The value will decrease, and the system will reduce the response stiffness of the mapping function, that is, decrease the value in the aforementioned formula. This value makes the comfort torque curve smoother, which is equivalent to reducing the "hardness" of the motor's response to torque commands, thereby absorbing road impacts and suppressing slippage.

[0055] , in, These are the adjusted response stiffness parameters. It is the reference stiffness parameter.

[0056] Finally, the system uses the corrected comfort torque curve to generate the final smooth motor control command. After stiffness adjustment factor correction, the comfort torque curve can better adapt to the current road conditions. The system discretizes this real-time updated curve to generate a series of specific current or voltage commands, which are sent to the motor controller as the final smooth motor control commands. These commands not only include the magnitude of the torque but also embed the torque change rate after physiological characteristic simulation and road adaptability adjustment, thereby achieving the highest level of coordinated safety and comfort control.

[0057] Optionally, obtaining the vehicle's wheel speed difference parameters and road surface roughness parameters, and converting them into stiffness adjustment factors, includes: Obtain the vehicle's wheel speed difference parameters and road surface roughness parameters; Calculate the real-time slip trend index based on the wheel speed difference parameter; Based on the mapping relationship between the road surface roughness parameter and the preset adhesion coefficient, the current peak adhesion coefficient of the road surface is calculated; The stiffness adjustment factor is calculated by fusing the real-time slip trend index with the current peak adhesion coefficient of the road surface.

[0058] Specifically, the system first acquires two key raw data inputs in parallel. First, it samples the vehicle's wheel speed difference parameters in real time using Hall effect sensors or encoders mounted on the front and rear wheels at a high frequency, such as 200 Hz. Simultaneously, it acquires road roughness parameters reflecting the physical unevenness of the road surface by analyzing the vertical axis acceleration signal from the onboard IMU (Inertial Measurement Unit) or through specialized vision and vibration sensors.

[0059] Next, the system calculates the real-time slip trend index based on the wheel speed difference parameter, quantifying the tendency of the driving wheel to slip relative to the driven wheel. This is a key indicator for determining whether the grip is approaching its limit. The formula for calculating the slip ratio S is: , in It is the angular velocity of the drive wheel. This refers to the angular velocity of the driven wheel. The system not only calculates the instantaneous slip ratio but also focuses on its changing trend within a short time window, i.e., the derivative of the slip ratio. The instantaneous slip ratio... By weighting and combining these factors with their rate of change, and then normalizing the results, we obtain the real-time slip trend index. This index can warn of the risk of getting out of control earlier than the slip rate alone.

[0060] Meanwhile, based on the mapping relationship between road surface roughness parameters and preset adhesion coefficients, the system calculates the current peak adhesion coefficient of the road surface, providing a macroscopic assessment of the maximum grip capability that the current road surface type can offer. Internally, the system stores either a lookup table or a function model that describes the relationship between different road surface roughness parameter values ​​and the peak adhesion coefficient of typical road surfaces. The correspondence between them. For example, a lower roughness value may correspond to a higher roughness value for dry asphalt pavement. The value ranges from approximately 0.8 to 1.0, while the roughness value corresponding to a high-frequency vibration may map to a low roughness value on a gravel surface. The value is approximately 0.4 to 0.6.

[0061] Finally, the system integrates the real-time slip trend index with the current peak adhesion coefficient of the road surface to calculate the final stiffness adjustment factor. By combining micro-level instability trends with macro-level road surface capabilities, a comprehensive and dynamic adjustment command is generated. The fusion process can be represented by the following equation: , in, It is the stiffness adjustment factor of the final output, and its value range is usually between 0 and 1. It is a fusion function, which can be a simple weighted average or a more complex fuzzy logic system. and These are the normalized current peak adhesion coefficient and the real-time slip trend index, respectively. This fusion strategy ensures that when macroscopic road conditions deteriorate or microscopic slip trends intensify, The value of will decrease accordingly, thereby instructing the motor controller to reduce the response stiffness, and conversely, to maintain or increase the stiffness, so as to achieve precise adaptive adjustment for different road conditions.

[0062] S5. Execute the motor smooth control command and collect vehicle status data and energy data after executing the motor smooth control command, and correct the master-slave constraint data and the parameters of the mapping function.

[0063] Optionally, the modification of the master-slave constraint data and the parameters of the mapping function includes: The actual energy recovery and consumption after executing the motor smooth control command are collected and compared with a preset set of theoretical values ​​to generate energy efficiency deviation characteristics; Collect and analyze the cadence and pedal force parameters after executing the motor smoothing control command, and generate experience evaluation features; The parameter configuration related to the energy boundary in the master-slave constraint data is corrected according to the energy efficiency deviation characteristics, and the parameter configuration related to smoothness in the mapping function is corrected according to the experience evaluation characteristics.

[0064] Specifically, the system first initiates an energy efficiency deviation analysis to quantify the gap between actual energy performance and theoretical optimal values, thereby driving adjustments to energy-related parameters. After executing the motor smoothing control command, the system uses the Battery Management System (BMS) to accurately collect the actual energy recovery and total consumption during a complete ride. Simultaneously, based on road conditions, speed, load, and other data from the ride, the system calls upon an offline vehicle energy consumption model to calculate the theoretically optimal energy recovery and minimum consumption under those conditions; these theoretical values ​​constitute a set of theoretical values. By comparing the actual collected values ​​with the theoretical values, the system generates an energy efficiency deviation characteristic ΔE. , in, Representing actual energy efficiency indicators, such as the ratio of recovered energy to consumed energy. This represents the corresponding theoretical optimal value. A negative value indicates that the actual energy efficiency is lower than the theoretical expectation.

[0065] In parallel, the system initiates an evaluation and analysis of the riding experience, quantifying the rider's comfort and control at the human-computer interaction level to guide the optimization of smoothness-related parameters. After command execution, the system continuously collects cadence and pedal force parameters, focusing on analyzing the convergence smoothness of these two parameters during the intervention and withdrawal phases of assist or regenerative torque. For example, the system calculates the time required for cadence and pedal force to recover to a stable state before and after torque changes, as well as the fluctuation amplitude during this process. If the recovery time is too long or the fluctuation is too drastic, it indicates that the torque change is too abrupt, affecting the riding experience. These quantitative analysis results are integrated into an experience evaluation feature. .

[0066] Finally, based on these two newly generated features, the system performs online or offline corrections to the core parameters within the control system, transforming the evaluation results into specific control strategy improvements. This is based on the energy efficiency deviation characteristics. The system will automatically correct the parameter configurations related to the energy boundary in the master-slave constraint data. For example, if A consistently negative value indicates that the energy recovery potential is not being fully utilized. The system may appropriately relax the maximum recovery torque limit in the recovery phase constraint subset, or adjust the weight of the load prediction characteristics in the master phase calculation formula. This makes it easier for it to enter a recycling-dominated state. Based on experience evaluation characteristics... The system will accordingly adjust the smoothness-related parameter configuration in the mapping function that simulates the motion physiological characteristics of human operating parts. For example, if If the cyclist's pedaling force fluctuates significantly, the system will increase the smoothness parameter in the mapping function. This makes the torque curve smoother, thereby improving the smoothness of human-machine interaction. These adjustments are ongoing, enabling the control system to have self-learning and self-optimization capabilities.

[0067] Based on the same inventive concept, such as Figure 4 As shown, the present invention also provides an energy recovery and power assist coordinated control system for electric two-wheeled vehicles, the system comprising: The sensing module is used to acquire and process vehicle operation data to generate load prediction features and intent confidence vectors. The constraint generation module is used to fuse the load prediction features with the intent confidence vector and construct constraints to generate master-slave constraint data with phase difference. The dual-ring sensing module is used to dynamically trim and compensate the energy recovery demand and electric assist demand based on the master-slave constraint data, and generate a dynamic torque distribution map. The safety control module is used to perform smooth safety mapping based on the dynamic torque distribution diagram and the preset mapping function of the motion physiological characteristics of the simulated human body operating parts, and generate smooth motor control commands. The execution and feedback module is used to execute the motor smooth control command and collect vehicle status data and energy data after the execution of the motor smooth control command, and to correct the master-slave constraint data and the parameters of the mapping function.

[0068] To verify the feasibility of this invention in practice, it was applied to the control system of an electric-assist bicycle. The test vehicle was equipped with a bottom bracket torque sensor, crank cadence sensor, electronic brake sensor, six-axis inertial measurement unit (IMU), and front and rear wheel speed sensors. The microcontroller (MCU) of the control system operated at a frequency of 200MHz, and the control algorithm had a cycle time of 10 milliseconds. The test scenario was set to simulate a typical "flat road-downhill-flat road" continuous road condition in urban cycling.

[0069] At time T0, the cyclist maintains a constant speed of 20 km / h on a flat road with a cadence of 80 RPM and a pedal force of 60 N·m. At time T1, the bicycle enters a downhill section with a gradient of -5°. The cyclist stops pedaling and lightly squeezes the brake lever, resulting in 15% travel. Due to the downhill slope, the pedal force drops from 60 N·m to 0, and the cadence also stops. The controller calculates the short-term load disturbance characteristics within a 100-m window. Let the normalized baseline pedal force Tn be 100 N·m, the baseline cadence Cn be 90 RPM, and the weighting coefficients be... , Change in pedaling force cadence change Short-term load disturbance characteristics This negative value indicates that the vehicle encountered a significant external assist load (downhill).

[0070] The controller acquires braking operation parameters such as 15% travel and vehicle tilt angle parameters measured by the IMU (-5°). The system analyzes this as a clear, moderately strong deceleration intention. The initial intention strength and clarity characteristics are quantified. The system uses... The strong load disturbance characteristics were used to correct and confirm the rationality of the deceleration intention. The final generated intention confidence vector is [-0.5, 0.9], representing a deceleration / braking intention with a strength of 50% and a clarity of 90%. Simultaneously, based on this high-confidence intention and the downhill load, the system generates load prediction features, predicting that an equivalent -8 N·m energy recovery opportunity will continue to exist within the next 500 milliseconds.

[0071] Based on the features generated in the previous step, the system begins to construct master-slave constraint data. The active phase constraint subset is dominated by the intention confidence vector [-0.5, 0.9], defining a need for a slower power output response and high torque smoothness. The recovery phase constraint subset is dominated by the load prediction feature, i.e., the -8 N·m recovery opportunity. Combining IMU and wheel speed sensor data, it determines that the current adhesion risk prediction level is level 1, which is excellent, generating constraints that allow a maximum recovery torque of -15 N·m and a relatively high torque growth rate.

[0072] Intended weights Load weight Intended to be confident in vector quantization values Quantitative values ​​of load forecast characteristics This is obtained based on a prediction mapping of -8 N·m. Main control phase. .because If the value is much less than 0, the system determines that the current energy recovery phase is the master control phase. The subset of recovery phase constraints is set as the "master constraint" with a weight of 0.9; the subset of active phase constraints is set as the "slave constraint" with a weight of 0.1, and a response delay of 100 milliseconds is applied.

[0073] At time T2, the bike is still going downhill, but the gradient is decreasing. The rider begins to pedal lightly, preparing to regain speed at the bottom of the hill while maintaining light braking. At this point, a control conflict occurs. The system recognizes the electric assist demand generated by pedaling. and the energy recovery requirements generated by braking / downhill. According to the master phase (Recycling-led), the system will meet the demand for electric power assistance. The demand is reduced to 0 N·m. At this point, the demand type being reduced is identified as electric power assist demand. However, since the recovery intention is still strong, the system generates an additional smooth assist torque as a compensation torque command, but its amplitude is 0, to be activated when the recovery intention weakens.

[0074] The final target torque at the current moment Calculated. This -6 N·m torque command is written into the dynamic torque distribution map.

[0075] At time T2, a smooth and safe mapping and execution is performed. The original command of -6 N·m is processed by a preset S-shaped mapping function to generate a comfortable torque curve that smoothly decreases to -6 N·m within 120 milliseconds, avoiding abrupt drag.

[0076] At this point, the vehicle traverses a slightly bumpy road surface, causing the road surface roughness parameter to increase. The system calculates a real-time slip trend index of 0.1 based on the wheel speed difference and this parameter, and the current peak road surface adhesion coefficient is 0.8. These factors are then combined to obtain the stiffness adjustment factor. The reference response stiffness parameters of the mapping function Adjusted stiffness The reduced stiffness flattens the comfort torque curve, resulting in a smoother motor response and effective absorption of road impacts. Ultimately, the corrected comfort torque curve is translated into smooth control commands sent to the motor controller.

[0077] After several similar cycling cycles, the system collected data and made adaptive corrections. The system found that on several downhill sections, the actual energy recovery was 12% lower than the theoretical value, indicating an energy efficiency deviation. The system also analyzed that during the transition from recovery to assist at the bottom of the slope, the rider's pedaling force exhibited an average overshoot fluctuation of 15%, resulting in a poor experience. Based on the energy efficiency deviation characteristics, the system adjusted the load weight in the main control phase calculation formula. The value was increased from 0.3 to 0.35, making the system more inclined to enter a recovery-dominant state. Based on experience evaluation characteristics, the system increased the configuration of smoothness-related parameters in the mapping function and lowered the trigger threshold for additional smoothness assist torque, so that it intervenes as soon as the recovery intention begins to weaken, for a smooth transition.

[0078] Through the above process, this invention achieves seamless coordinated control of power assist and regenerative braking on a test vehicle, and continuously optimizes performance through closed-loop feedback.

[0079] Table 1 Data table of key parameter generation process

[0080] Table 2 Dynamic Torque Distribution and Smoothing Control Data Table

[0081] Table 3. Data on the Adaptive Correction Effect of the Control System

[0082] As can be seen from the data in Tables 1 to 3 above, the method of the present invention can transform the rider's behavior and the vehicle's environment into control characteristics. As shown in Table 1, under different operating conditions, T0 to T2 accurately generated control characteristics from assist-dominant to... Towards recycling as the mainstay The system achieves a smooth transition. Table 2 demonstrates the effectiveness of this method in resolving control conflicts. At time T2, facing a +3 N·m assist demand and a -6 N·m recovery demand, the system decisively selects to execute recovery based on the master control phase and smooths and softens the output torque in conjunction with road conditions, ensuring the stability and predictability of vehicle dynamics. The data in Table 3 proves the long-term value of the closed-loop feedback correction mechanism. Through learning from energy efficiency and experience data, the system autonomously improves both range performance and riding smoothness to a higher level, fully verifying the advanced nature and practicality of this invention.

[0083] It should be noted that the above descriptions are merely exemplary embodiments of the present invention and should not be construed as limiting the scope of the invention. All equivalent changes and modifications made in accordance with the teachings of this invention are still within the scope of this invention. Those skilled in the art will readily conceive of other embodiments of the invention upon considering the disclosure of the specification and practical truths. This application is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or conventional techniques in the art not described herein.

Claims

1. A method for coordinated control of energy recovery and power assist in an electric two-wheeled vehicle, characterized in that, The method includes: Acquire and process vehicle operation data to generate load prediction features and intent confidence vectors; By fusing the load prediction features with the intent confidence vector and constructing constraints, master-slave constraint data with phase difference is generated. Based on the master-slave constraint data, the energy recovery demand and electric assist demand are dynamically trimmed and compensated to generate a dynamic torque distribution map. Based on the dynamic torque distribution diagram and the preset mapping function of the simulated human body operating component motion physiological characteristics, a smooth and safe mapping is performed to generate a smooth motor control command. The motor smooth control command is executed, and vehicle status data and energy data are collected after the execution of the motor smooth control command. The master-slave constraint data and the parameters of the mapping function are then corrected.

2. The method for coordinated control of energy recovery and power assist in an electric two-wheeled vehicle according to claim 1, characterized in that, The process of acquiring and processing vehicle operation data to generate load prediction features and intent confidence vectors includes: The vehicle's cadence and pedal force parameters are acquired, and load disturbance is observed based on the rate of change of cadence and pedal force parameters to generate short-term load disturbance characteristics. The vehicle's braking operation parameters and body tilt angle parameters are obtained, and the intention trend analysis is performed on the change gradient and angular velocity of the braking operation parameters and body tilt angle parameters to generate intention intensity and clarity features. The intent strength and the clarity features are corrected using the short-term load disturbance features, and load disturbance analysis is performed using the corrected intent strength and clarity features to generate load prediction features and intent confidence vectors.

3. The method for coordinated control of energy recovery and power assist in an electric two-wheeled vehicle according to claim 2, characterized in that, The step of fusing the load prediction features and the intention confidence vector and constructing constraints to generate master-slave constraint data with phase difference includes: Using the intent confidence vector as the primary driver and incorporating the load prediction features, an active phase constraint subset is generated that defines the power output response speed and torque smoothness. Based on the load prediction features and fused with the intention confidence vector, a subset of recovery phase constraints defining energy recovery intensity and disturbance rejection stability is generated; The master control phase is calculated based on the intent confidence vector and the load prediction features. The phase difference relationship between the active phase constraint subset and the recovery phase constraint subset in terms of time and weight is determined based on the master control phase, and master-slave constraint data is generated.

4. The method for coordinated control of energy recovery and power assist in an electric two-wheeled vehicle according to claim 3, characterized in that, The subset of recovery phase constraints defining the energy recovery intensity and disturbance rejection stability includes: Obtain the attachment risk prediction level from the load prediction characteristics; Based on the intent confidence vector and the attachment risk prediction level, the scope of influence and constraint strength are generated. The initial constraint subset is generated using the influence range and constraint strength. Obtain real-time load estimation parameters for the vehicle and calculate the basic recovery boundary. The initial constraint subset is fused with the basic recovery boundary to obtain a recovery phase constraint subset that defines the energy recovery intensity and disturbance rejection stability.

5. The method for coordinated control of energy recovery and power assist in an electric two-wheeled vehicle according to claim 3, characterized in that, The generation of the dynamic torque distribution map includes: Identify the overlapping and conflicting areas between the current energy recovery demand and the electric assist demand in the time domain or torque domain; Based on the master control phase in the master-slave constraint data and the overlapping conflict area, the demand side is determined and trimmed to generate the trimmed demand features; Compensation torque commands are generated based on the trimmed demand features and then fused with the untrimmed demand features to generate a dynamic torque distribution map.

6. The method for coordinated control of energy recovery and power assist in an electric two-wheeled vehicle according to claim 5, characterized in that, The generation of compensation torque commands based on the trimmed demand characteristics includes: Identify the demand type corresponding to the trimmed demand characteristics. If it is an energy recovery demand, generate an enhanced recovery pulse as a compensation torque command. If electric power steering is required, an additional smooth power steering torque is generated as a compensation torque command. The magnitude and duration of the compensation torque command are limited within the dynamic safety boundary of the master-slave constraint data.

7. The method for coordinated control of energy recovery and power assist in an electric two-wheeled vehicle according to claim 5, characterized in that, The generated motor smooth control command includes: The original torque command in the dynamic torque distribution diagram is mapped through a preset mapping function that simulates the physiological characteristics of human body operating parts to generate a comfort torque curve. Obtain the vehicle's wheel speed difference parameters and road surface roughness parameters, and convert them into stiffness adjustment factors; The response stiffness of the mapping function is dynamically adjusted using the stiffness adjustment factor to correct the comfort torque curve; The modified comfort torque curve is used to generate smooth motor control commands.

8. The method for coordinated control of energy recovery and power assist in an electric two-wheeled vehicle according to claim 7, characterized in that, The process of acquiring the vehicle's wheel speed difference parameters and road surface roughness parameters, and converting them into stiffness adjustment factors, includes: Obtain the vehicle's wheel speed difference parameters and road surface roughness parameters; Calculate the real-time slip trend index based on the wheel speed difference parameter; Based on the mapping relationship between the road surface roughness parameter and the preset adhesion coefficient, the current peak adhesion coefficient of the road surface is calculated; The stiffness adjustment factor is calculated by fusing the real-time slip trend index with the current peak adhesion coefficient of the road surface.

9. The method for coordinated control of energy recovery and power assist in an electric two-wheeled vehicle according to claim 7, characterized in that, The modification of the master-slave constraint data and the parameters of the mapping function includes: The actual energy recovery and consumption after executing the motor smooth control command are collected and compared with a preset set of theoretical values ​​to generate energy efficiency deviation characteristics; Collect and analyze the cadence and pedal force parameters after executing the motor smoothing control command, and generate experience evaluation features; The parameter configuration related to the energy boundary in the master-slave constraint data is corrected according to the energy efficiency deviation characteristics, and the parameter configuration related to smoothness in the mapping function is corrected according to the experience evaluation characteristics.

10. A coordinated control system for energy recovery and power assist in an electric two-wheeled vehicle, characterized in that, The system includes: The sensing module is used to acquire and process vehicle operation data to generate load prediction features and intent confidence vectors. The constraint generation module is used to fuse the load prediction features with the intent confidence vector and construct constraints to generate master-slave constraint data with phase difference. The dual-ring sensing module is used to dynamically trim and compensate the energy recovery demand and electric assist demand based on the master-slave constraint data, and generate a dynamic torque distribution map. The safety control module is used to perform smooth safety mapping based on the dynamic torque distribution diagram and the preset mapping function of the motion physiological characteristics of the simulated human body operating parts, and generate smooth motor control commands. The execution and feedback module is used to execute the motor smooth control command and collect vehicle status data and energy data after the execution of the motor smooth control command, and to correct the master-slave constraint data and the parameters of the mapping function.

Images