A commercial vehicle eps bidirectional isomerism boost compensation method and system based on direction state recognition

Abstract

The application discloses a kind of based on direction state identification's commercial vehicle EPS two-way isomerism boost compensation method and system.The method includes: collecting steering wheel torque, output end angular velocity and motor speed signal and filtering;Equivalent mapping from motor side to output end is established;Based on filtering signal, through hysteresis comparison and duration confirmation, identify forward boost, reverse return or commutation transition state;According to state, call different physical compensation model-adopt first model containing efficiency coefficient Cf and friction term Ff in forward direction, adopt second model containing efficiency coefficient Cr and damping term Dr in reverse direction;When commutation transition, according to steering wheel torque, weighted smooth transition is carried out and pre-filling gap compensation is superimposed;Based on position and temperature, update parameters on line, forward parameter is only updated in forward direction, and reverse parameter is only updated in reverse direction.The application solves the problem that steering feeling is heavy when turning in, return feeling is light, commutation impact and center area following is poor caused by asymmetric transmission of commercial vehicle EPS.

Description

Technical Field

[0001] This invention relates to the field of electric power steering control technology for commercial vehicles, and in particular to a bidirectional heterogeneous power steering compensation method and system for commercial vehicle EPS based on directional state recognition. This invention is particularly applicable to power steering control under conditions such as low-speed high-load steering, turning the wheel in place, returning to center, changing direction, and small-angle correction. Background Technology

[0002] Electric power steering (EPS) systems use an electric motor to provide auxiliary torque to assist the driver in steering. In the commercial vehicle sector, the front axle load typically exceeds 3 tons, and in some models it exceeds 6 tons, placing much higher technical demands on the assist capability, response speed, and control precision of EPS compared to passenger vehicles.

[0003] In commercial vehicle EPS with an actuator chain structure of "motor + reduction mechanism + roller screw + nut gear sector", due to the inherent efficiency differences of mechanical transmission components such as roller screw and nut gear sector, there is a significant asymmetry in the transmission efficiency, friction characteristics and damping characteristics of the actuator chain when driving forward (motor → wheel) and returning to center (wheel → steering wheel). This asymmetry is particularly prominent under heavy load and low speed conditions, specifically manifested as: (1) heavy steering feel and light return feel - insufficient forward assist, too fast return to center and lack of damping; (2) reversing impact and free travel feel - when the steering wheel passes zero, the sudden change in actuator chain clearance and transmission efficiency causes jerking and knocking; (3) poor central area following performance - the nonlinearity of tight meshing in the middle of the nut gear sector and loose at both ends leads to uneven assist when correcting at small angles; (4) temperature and wear effects - low temperature viscosity increases dramatically in harsh environments of -40℃ to 80℃, and the clearance increases after long-term wear, and the unified compensation model completely fails.

[0004] Existing technologies include solutions for commutation lag identification, mid-position deviation compensation, low-temperature differential compensation, friction model compensation, or clearance compensation. However, none of these solutions have been able to resolve the series of problems caused by the inherent asymmetry in the forward and reverse transmission characteristics of the "motor + roller screw + nut gear sector" execution chain. This is because their compensation models do not change their physical composition with changes in the steering direction and cannot perform targeted smooth connection in the commutation transition zone.

[0005] Therefore, there is an urgent need for a control method that can identify the steering direction state in real time, call compensation models with different physical structures according to the direction state, and perform smooth weighted switching and pre-filling compensation in the reversing transition zone to eliminate the difference in forward and reverse feel and the reversing shock. Summary of the Invention

[0006] The purpose of this invention is to address the shortcomings of commercial vehicle EPS systems, which include an actuator chain structure of "motor + reduction mechanism + roller screw + nut gear sector". These shortcomings stem from the inherent asymmetry in transmission efficiency, friction, and damping characteristics during forward drive and reverse return. This leads to inconsistent steering feel, a sense of free travel and mechanical knocking during reversing, and poor power assist tracking during small-angle corrections in the center zone. The invention provides a control method and system based on steering state recognition that can identify the steering direction state in real time, call compensation models with different physical structures according to the direction state, and perform smooth weighted switching and pre-filling compensation in the reversing transition zone.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: a bidirectional heterogeneous power assist compensation method for commercial vehicle EPS based on orientation state recognition, comprising the following steps: S1: Acquire the running signal with a sampling period of Ts and perform filtering using a second-order low-pass Butterworth filter; Furthermore, the operating signals include the steering wheel torque signal Tsw, the output angular velocity signal ωout, and the motor speed signal ωm; the cutoff frequency of the second-order low-pass Butterworth filter is preferably 15Hz, and the steering wheel torque Tsw is obtained after filtering. f The filtered output angular velocity ωout f and the filtered motor speed ωm f .

[0008] S2: Establish the equivalent mapping relationship from the motor side to the output end; Based on the transmission relationship of the actuator chain (reduction ratio ir, roller screw lead p, nut sector conversion coefficient kn), the quantities on the motor side (speed, torque, angle) are equivalently mapped to the output end (sector end) and the actual equivalent torque T at the output end is calculated. driver =Kc×Kt×Iq; where Kc is the global equivalent conversion coefficient (including reduction ratio, lead screw, and sector gear conversion), Kt is the motor torque constant, and Iq is the current motor current; T driver This represents the equivalent torque (feedback amount) actually transmitted to the output terminal under the current motor current Iq, and it is explicitly stated in S2 that this amount is used for error calculation in S4; S3: Based on the filtered signal obtained from S1, the current directional state is identified through hysteresis comparison and duration confirmation; The filtered steering wheel torque Tsw f and the output angular velocity ωout fCombining the preset torque dead zone threshold Tdz, speed dead zone threshold ωdz, and commutation duration threshold Ttrn, the current directional state is identified as one of the following three states through hysteresis comparison logic and state duration confirmation: Positive assist state: when |Tsw f |>Tdz and|ωout f |>ωdz and sign(T) target )==sign(ωout f And this state continues for more than the first preset time; Reverse return to positive state: when |Tsw f |>Tdz and|ωout f |>ωdz and sign(T) target )≠sign(ωout f ), or |Tsw f | The rate of decrease exceeds the preset rate threshold, but ωout f The direction remains unchanged; Commutation transition state: when |Tsw f |≤Tdz or|ωout f |≤ωdz, or within the Ttrn time after the direction status flag changes; Among them, T target Desired output torque: based on the current vehicle speed V and the filtered steering wheel torque Tsw f The target torque of the motor is obtained by looking up the basic assist characteristic curve in a table, and then obtained by equivalent mapping to the output end; the sign of this value represents the desired assist direction, which is used as a feedforward term for direction state identification and compensation calculation.

[0009] S4: Based on the current direction state obtained from S3, call the corresponding equivalent compensation model. The first compensation model is used in the positive assist state, and the second compensation model is used in the reverse return state. The physical composition of the first compensation model and the second compensation model are different, forming a bidirectional heterogeneous compensation.

[0010] Specifically: Under positive assist conditions, the first compensation model, Tcomp, is invoked. f =Cf×T target +Ff×tanh(ωout f / ω0)+kf×e; where Cf is the positive efficiency compensation coefficient, Ff is the positive friction compensation term, kf is the positive error correction coefficient, ω0 is the speed normalization parameter, and the assist error e=T target -T driver ; In the reverse recovery state, the second compensation model, Tcomp, is invoked. r =Cr×Ttarget -Dr×tanh(ωout f / ω0)+kr×e, where Cr is the reverse efficiency compensation coefficient, Dr is the reverse damping compensation term, and kr is the reverse error correction coefficient; The first compensation model includes efficiency compensation and friction compensation, while the second compensation model includes efficiency compensation and damping compensation. The two models have different physical compositions (Ff≠Dr), and Cf≠Cr and kf≠kr, respectively, to adapt to the different physical requirements of positive assist and reverse return.

[0011] S5: Based on the current direction state identified in step S3, select the corresponding compensation model to calculate the basic compensation amount Tcomp. base ; If the current state is a positive assist state, then substitute it into the formula of the first compensation model to calculate Tcomp. base =Tcomp f .

[0012] If the current state is a reverse-to-positive state, then substitute it into the formula of the second compensation model to calculate Tcomp. base =Tcomp r .

[0013] If the current state is a commutation transition state, then calculate Tcomp first. f and Tcomp r Two intermediate values ​​(used for the weighted transition in step S6), at which point Tcomp is not output. base .

[0014] S6: During the steering transition state, the outputs of the first compensation model and the second compensation model are weighted and transitioned according to the smooth weighting function of the steering wheel torque, and the pre-fill gap compensation term Tgap is superimposed.

[0015] When a commutation transition state is identified, perform the following operations: Based on the filtered absolute value of the steering wheel torque |Tsw f |Calculate the smoothing weight λ: λ = sat(|Tsw) f | / Ttr,0,1), where Ttr is the preset commutation transition torque threshold, and sat() is the saturation function that limits λ to the range [0,1]. The total compensation amount under the commutation transition state is calculated as: Tcomp total =λ×Tcomp f +(1-λ)×Tcomp r +Tgap; Tgap is the pre-fill gap compensation term, which is activated only in the commutation transition state; in the other two states, Tgap = 0. The formula for calculating Tgap is Tgap = Kg × dωout. f / dt, where Kg is the pre-filling gap compensation coefficient, the value of which is proportional to the equivalent inertia of the total backlash of the execution chain; If the current state is either a forward or reverse state: Tcomp total =Tcomp base .

[0016] S7: The total compensation amount is superimposed on the basic assist command to form the final assist output, which drives the EPS motor to work; First, based on the current vehicle speed V and the filtered steering wheel torque Tsw f Locate the basic assist characteristic curve (MAP) to obtain the basic assist command T. base (The unit is Nm, which represents the basic assist torque that the motor should output).

[0017] Then, the basic assist command is added to the final compensation amount obtained in step S6 to obtain the total assist command: T output =T base +Tcomp total ; Finally, regarding T output The system performs amplitude limiting (limiting the output torque within the maximum / minimum range of the motor) and slope limiting (preventing abrupt changes), and sends the final value to the motor driver to drive the EPS motor to output the corresponding assist torque.

[0018] S8: Based on the output position and / or temperature, the parameters of the compensation model are updated online adaptively.

[0019] Under operating conditions that meet preset enabling conditions (including but not limited to: no impact disturbance, no system fault, no rapid reversing, and stable and effective driver input), based on the current output angle θout and temperature Temp, parameters such as Cf, Cr, Ff, Dr, kf, and kr are updated online by looking up a pre-calibrated position-temperature two-dimensional parameter table to adapt to mechanical wear and temperature changes. Among them, under the premise of meeting the enabling conditions, the positive parameters (Cf, Ff, kf) are updated only in the positive assist state, and the negative parameters (Cr, Dr, kr) are updated only in the negative return state. The update step size is preferably 0.001 to 0.01 to prevent parameter abrupt changes.

[0020] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) The commutation shock is significantly reduced. During the commutation transition state, the positive compensation amount and the reverse compensation amount are weighted by a smoothing weight function λ, and a pre-filling gap compensation term Tgap=Kg×dωout is superimposed. f / dt eliminates the idle travel caused by the execution chain gap. Bench tests show that when the steering wheel crosses zero at an angular velocity of 100° / s, the peak value of the output angular acceleration fluctuation is reduced by 42% compared to the existing unified compensation method.

[0021] (2) The tracking performance of small-angle corrections in the central region is significantly improved. This is achieved through hysteresis comparison and duration confirmation mechanisms in directional state recognition (Tdz=0.5Nm, Tf). min =10ms), effectively avoiding frequent state jitter caused by signal noise when fine-tuning the steering wheel in the center zone. Combined with the linear characteristics of the tanh speed normalization function in the low-speed zone, it provides fine proportional friction compensation. In the small-angle (±5°) sinusoidal frequency sweep input (0.5Hz) test in the center zone, the root mean square error (RMSE) of the power steering tracking is reduced by 30% compared with existing methods.

[0022] (3) The consistency of steering feel is significantly improved. Innovatively, heterogeneous compensation models with different physical compositions are established for the positive assist state and the reverse return-to-center state—the positive model focuses on efficiency compensation and friction compensation to solve the problem of heavy steering input, while the reverse model focuses on efficiency compensation and damping compensation to solve the problem of excessively fast return-to-center without damping feel. In a blind test conducted by 10 professional driving evaluators (out of 10), the proposed solution scored 8.9, 8.7, and 8.5 respectively in terms of steering smoothness, return-to-center consistency, and clarity of road feel in the center area, all of which are significant improvements compared to existing solutions (7.1, 6.6, 7.0).

[0023] (4) Parameters adapt to temperature changes and mechanical wear. A two-dimensional position-temperature MAP table with Cf=f1(θout,Temp) and Cr=f2(θout,Temp) was established, covering the entire temperature range from -40℃ to 80℃ and the entire angle range from the median to the limit angle. The forward parameter group and the reverse parameter group are updated independently, avoiding the problem of mutual contamination of forward and reverse data in the existing unified update method, so that the compensation model can maintain accuracy over a long period of time. Attached Figure Description

[0024] Figure 1 This is a flowchart illustrating the implementation steps of the present invention; Figure 2 This is a schematic diagram of the state machine transition for orientation state recognition. Detailed Implementation

[0025] To provide a further understanding of the purpose, structure, features, and functions of the present invention, detailed descriptions are provided below with reference to specific embodiments.

[0026] This invention provides a bidirectional heterogeneous power steering compensation system for commercial vehicles based on directional state recognition. The system includes: a steering wheel 1, a steering column 2, a torque sensor 3 (for detecting steering wheel torque Tsw), an angle sensor 4 (for detecting steering wheel angle), a motor 5 (power steering motor), a reduction mechanism 6, a ball screw 7, a nut gear sector 8, an output position sensor 9 (for detecting output angular velocity ωout and angle θout), a temperature sensor 10 (for detecting ambient temperature or grease temperature Temp), and a controller 11 (ECU). Arrows indicate signal flow to controller 11. Controller 11 executes the bidirectional heterogeneous power steering compensation method described in this invention.

[0027] Please refer to the reference. Figure 1 This invention provides a method for bidirectional heterogeneous power assist compensation in commercial vehicles based on orientation state recognition, comprising the following steps: S1: The controller 11 acquires the steering wheel torque signal Tsw through the torque sensor 3 with a sampling period Ts (e.g., Ts=1ms), acquires the output angular velocity signal ωout through the output position sensor 9 (obtained by differentiating the angle signal), and acquires the motor speed signal ωm at the same time.

[0028] To avoid misjudgment of direction due to sensor noise and high-frequency road vibration, this invention filters the aforementioned signals. A second-order low-pass Butterworth filter is preferably used, with a cutoff frequency set to 15Hz. The reason for choosing 15Hz is that this frequency can effectively suppress high-frequency noise (such as road vibration and gear meshing impact) while fully preserving the driver's operating frequency band (typically ≤5Hz).

[0029] The discrete recursive formula for the filter is: y(k) = b0×x(k) + b1×x(k-1) + b2×x(k-2) - a1×y(k-1) - a2×y(k-2); The filter coefficients b0, b1, b2, a1, and a2 are calculated using the standard Butterworth filter design method based on a cutoff frequency of 15Hz and a sampling period Ts.

[0030] The filtered signals are denoted as: Tsw f (Filtered steering wheel torque), ωout f (Output angular velocity after filtering), ωm f (Motor speed after filtering).

[0031] As an alternative, a Kalman filter can be used for integrated state estimation and filtering, which can simultaneously obtain the filtered signal and its rate of change.

[0032] S2: To compensate for the forward and reverse transmission characteristics of the actuator chain, the kinematic and dynamic parameters on the motor side need to be equivalently mapped to the output end (gear sector end). The equivalent mapping relationship is as follows: (1) Determination of the global equivalent conversion factor Kc: Kc takes into account the following factors: Lead of a ball screw (the linear displacement of the nut per revolution of the ball screw). The linear displacement to angular displacement conversion coefficient kn of the nut gear sector, i.e. the relationship between the linear displacement δ of the nut and the output angle θout: θout=kn×δ.

[0033] In practical implementation, kn is not a constant value, but a function of the output angle θout. Specifically, near the midpoint of the output (-30° to +30°), kn is close to the theoretical value due to the tight meshing of the nut teeth; as it approaches the limit angle (±600°), kn decreases by 5% to 10% due to the increased meshing backlash. This relationship is obtained through offline bench calibration and stored as a two-dimensional table: kn = MAP kn (θout).

[0034] Therefore, Kc is also a function of θout: Kc(θout)=(i×kn(θout)) / (2π).

[0035] (2) Calculation of the actual equivalent torque at the output end: T driver =Kc×Kt×Iq; (3) Calculation of the expected assist torque at the output end: T target =Tm ref ×Kc(θout) Among them, Tm ref The reference torque that the motor should output in basic power assist mode is determined by the basic power assist characteristic curve based on vehicle speed V and steering wheel torque Tsw. f Determine by referring to the table.

[0036] S3: Based on the filtered signal obtained from S1, the current directional state is identified through hysteresis comparison and duration confirmation; The current direction state includes three direction states, which are defined as follows: Positive assist mode: The direction of the motor output torque is the same as the direction of wheel rotation, which is the situation where the driver actively steers and the motor assists in driving.

[0037] Reverse return-to-center state: The steering wheel torque decreases (the driver returns to center or releases the steering wheel), but the wheels continue to turn under the action of the return-to-center torque. At this time, the actuator chain is in the reverse transmission state (wheel → motor).

[0038] Transitional state: The brief transition period after the steering wheel torque is close to zero or the steering status indicator has just changed.

[0039] The preset parameters used for the determination are: Torque dead zone threshold Tdz = 0.5 Nm: Steering wheel torque below this value is considered zero torque; Velocity dead zone threshold ωdz = 0.1 rad / s: the output angular velocity below this value is considered stationary; The commutation duration threshold Ttrn = 50ms: After the direction status flag changes, the commutation transition state is considered to be within this time. Shortest duration of positive state Tf min =10ms: This time must be maintained after the positive judgment condition is met before confirmation; Torque decrease rate threshold dT thr =-100Nm / s: When the steering wheel torque drops too quickly, a reverse judgment is triggered.

[0040] To avoid frequent state jumps ("jitter") caused by signal noise near zero, this invention employs hysteresis comparison logic: When the steering wheel torque rises from less than Tdz to more than Tdz (rising edge), the system does not immediately confirm entering the forward / reverse state, but waits for the signal to further strengthen and stabilize. When the value drops from greater than Tdz to below Tdz (falling edge), the transition to the switching state is not immediately confirmed. Instead, a hysteresis bandwidth (e.g., 0.1 Nm) is set, meaning the state transition is only triggered when the value drops below Tdz-0.1 Nm.

[0041] The complete state determination logic is as follows: (a) Positive assist state determination: all of the following conditions are met simultaneously and the positive assist state is maintained for Tf. min time: |Tsw f |>Tdz;|ωout f |>ωdz;sign(T target )==sign(ωout f ) (b) Reverse return to center state determination: One of the following conditions must be met: Compared to existing technologies that only identify torque lag at the moment of commutation without establishing a complete three-state directional state machine, this invention effectively avoids frequent misjudgments caused by signal noise near the zero point through hysteresis comparison and duration confirmation, providing a reliable state basis for the accurate invocation of subsequent heterogeneous models.

[0042] Condition 1: |Tsw f |>Tdz and|ωout f|>ωdz and sign(T) target )≠sign(ωout f ) Condition 2: d|Tsw f | / dt <dT thr (Torque drops rapidly) and sign(ωout) f Unchanged (c) Judgment of commutation transition state: One of the following conditions must be met: Condition 1: |Tsw f |≤Tdz; Condition 2: |ωout f |≤ωdz; Condition 3: Within the time Ttrn after the direction status flag changes; Figure 2 The diagram illustrates the transition relationships between the three states. State transition conditions include torque dead zone, speed dead zone, direction signal consistency, and duration threshold. Hysteresis regions are marked in the diagram to prevent state jitter caused by noise.

[0043] Its goal is to classify the operating status of the EPS system into three directional states in real time, providing a basis for selecting different compensation models in the future.

[0044] S4: Based on the current direction state obtained from S3, call the corresponding equivalent compensation model. The first compensation model is used in the positive assist state, and the second compensation model is used in the reverse return state. The physical composition of the first compensation model and the second compensation model are different, forming a bidirectional heterogeneous compensation.

[0045] In the positive assist state, the actuator chain is in positive transmission (motor → wheel). This invention adopts a first compensation model that includes efficiency compensation and friction compensation: Tcomp f =Cf×T target +Ff×tanh(ωout f / ω0)+kf×e; The meanings of each term are as follows: The Cf positive efficiency compensation coefficient is a function of the output angle θout and the temperature Temp, and is stored as a first two-dimensional lookup table MAP. f1 At room temperature (25℃), Cf is 1.0 at θout=0° (median position) and rises to 1.3 at θout=±600° (extreme position) to compensate for insufficient assistance due to reduced mechanical efficiency at the extreme position. Regarding temperature effects: at -20℃, all Cf values ​​increase by 15% overall (e.g., median 1.15); at 80℃, they decrease by 5% overall.

[0046] Compared to existing technologies, the bidirectional heterogeneous compensation model of this invention distinguishes for the first time the physical quantities that need to be compensated in the forward direction (efficiency compensation + friction compensation) and the physical quantities that need to be compensated in the reverse direction (efficiency compensation + damping compensation) from the directional dimension, rather than using a single model to cover all directions.

[0047] Ff is the positive friction compensation term, with units of Nm. Ff considers velocity dependence. When |ωout f When |ωout| < 0.2 rad / s, the friction compensation coefficient increases rapidly with increasing velocity (simulating the breakthrough process of static friction); when |ωout| < 0.2 rad / s, the friction compensation coefficient increases rapidly with increasing velocity (simulating the breakthrough process of static friction); f The friction approaches saturation at 0.5 rad / s (the constant stage of simulated Coulomb friction).

[0048] tanh(ωout f / ω0): Velocity normalization function, preferably ω0 = 0.5 rad / s. The advantage of choosing the tanh function is that when |ωout f When | < 0.5 rad / s, it is approximately linear (providing proportional friction compensation), and when it exceeds this value, it saturates (providing constant friction compensation), avoiding torque jumps that may be introduced by the sign function.

[0049] kf: Positive error correction coefficient, used to adjust the convergence speed of the tracking error, preferably kf=0.2.

[0050] In the reverse return state, the execution chain is in reverse transmission (wheel → motor). The physical structure of the second compensation model of this invention is different from that of the first compensation model—it includes efficiency compensation and damping compensation, rather than friction compensation. Tcomp r =Cr×T target -Dr×tanh(ωout f The meanings of each term are as follows: (ω0)+kr×e; The Cr reverse efficiency compensation coefficient is stored as a second two-dimensional lookup table MAP. r1 Since reverse transmission efficiency is usually higher than forward transmission efficiency, Cr is typically set between 0.85 and 1.0. It also varies with θout and Temp.

[0051] The Dr (Reverse Damping Compensation) term, measured in Nm, is the core physical difference between this model and the first compensation model. In the reverse-to-center state, the reverse transmission efficiency of the actuator chain is high, resulting in a large return force. Without damping control, the steering wheel returns to center too quickly, leading to overshoot and oscillation. The role of Dr is to provide appropriate damping torque during the return process, ensuring a smooth and controllable return.

[0052] kr: reverse error correction coefficient, preferably kr=0.15. Note that kf>kr (0.2>0.15), this is because the forward state emphasizes rapid establishment of assist response, while the reverse state focuses more on stability and compliance.

[0053] Different from the unified compensation model in the prior art, the present invention establishes compensation models with different physical structures for the forward assist state and the reverse centering state respectively, so as to adapt to the difference in transmission characteristics between the two directions.

[0054] S5: Based on the model of step S4, in step S5, the compensation amount is calculated according to the current direction state: If the current state is the forward assist state, the first compensation model is used to calculate Tcomp f , and keep Tcomp r as the value of the previous cycle or zero; If the current state is the reverse centering state, the second compensation model is used to calculate Tcomp r , and keep Tcomp f as the value of the previous cycle.

[0055] When the state is stabilized in one of the states (non- commutation transition state), only one compensation model is activated to generate the compensation amount.

[0056] S6: Step S6 is a key step for the present invention to solve the problem of "commutation impact". When the direction state enters the "commutation transition state", two tasks need to be completed: (1) smoothly switch from forward compensation to reverse compensation (or vice versa); (2) perform pre-gap-filling compensation for the actuator chain clearance.

[0057] Compared with the prior art, the pre-gap-filling compensation Tgap of the present invention is only activated in the commutation transition state, and is forced to zero in other direction states, which avoids additional torque disturbance that may be caused by full-time compensation, and achieves accurate gap filling at the commutation moment.

[0058] The calculation formula of weight λ is: λ=sat(|Tsw f | / Ttr,0,1); where Ttr is a preset commutation transition torque threshold (recommended value Ttr=2Nm). This function has the following properties: When |Tsw f |≥Ttr, λ=1, and the output of the forward compensation model is completely used; When |Tsw f |=0, λ=0, and the output of the reverse compensation model is completely used; When 0<|Tsw f |<Ttr, λ changes linearly between [0,1], realizing a gradual transition from forward direction to reverse direction.

[0059] As an alternative, an S-curve (such as a cubic spline interpolation function) can be used instead of a linear function to achieve a smoother transition. In this case, λ = 3 × (|Tsw) f | / Ttr)²-2×(|Tsw f | / Ttr)³.

[0060] When chain reversing is performed, the transmission direction changes, and the meshing surfaces of the gears / lead screws also switch accordingly. Mechanical backlash (total backlash) causes free travel in the initial stage of reversing, which manifests as a "free travel feel" and "knocking feel" when the steering wheel crosses zero. To eliminate this phenomenon, this invention introduces a pre-filling backlash compensation term, Tgap.

[0061] The formula for calculating Tgap is: Tgap = Kg × dωout f / dt Where: dωout f / dt is the output angular acceleration, reflecting the speed of commutation; Kg is the pre-filling compensation coefficient, with units of Nm·s / rad, and its value is determined by bench calibration based on the total backlash equivalent inertia and transmission stiffness of the actuator chain.

[0062] The activation logic of Tgap is as follows: it is only activated in the commutation transition state: when the direction state is in the commutation transition state, Tgap is calculated according to the above formula; in other states, it is forced to be zero: when the direction state is in the positive assist state or the reverse return state, Tgap=0.

[0063] S7: Forming the final boost output First, based on the current vehicle speed V and the filtered steering wheel torque Tsw f Locate the basic assist characteristic curve (MAP) to obtain the basic assist command T. base (The unit is Nm, which represents the basic assist torque that the motor should output).

[0064] Then, the basic assist command is added to the final compensation amount obtained in step S6 to obtain the total assist command: T output =T base +Tcomp total； Where T base Based on the base assist torque (determined from the base assist characteristic curve based on vehicle speed and steering wheel torque), Tcomp total This is the total compensation amount calculated through steps S3 to S6. The controller will... output The target current command is converted into a motor current command, which drives the EPS motor 5 through the motor driver to output the corresponding assist torque.

[0065] S8: To avoid erroneous parameter updates under unstable operating conditions, online updates are only allowed when all of the following conditions are met simultaneously: No shock disturbance: |dωout f / dt|<50rad / s², meaning the output angular acceleration is stable; The system is fault-free: the sensors are normal (no open circuit / short circuit), the motor temperature is normal (not overheated), and the bus voltage is normal (within the operating range). The direction-bound parameter update mechanism is another feature of this invention—the positive parameters (Cf, Ff, kf) are updated only in the positive assist state, and the negative parameters (Cr, Dr, kr) are updated only in the negative return state. This avoids the problem of data interference between the two directions in the existing unified update method, which leads to the decay of compensation accuracy. This allows the positive and negative models to independently adapt to the changes in transmission characteristics in their respective directions.

[0066] No fast reversal: The current direction state remains stable for a duration > 500ms; Driver input is stable and valid: |dTsw f / dt|<20Nm / s and|Tsw f |>1Nm.

[0067] For the positive parameter set {Cf, Ff, kf} (stored in MAP) f (In the middle), when all enabling conditions are met and the current state is a positive assist state, the following update rules are used: Cf(i+1)=Cf(i)+μ1×(T target -T driver ) Ff(i+1)=Ff(i)+μ2×(T target -T driver )×tanh(ωout_f / ω0) kf(i+1)=kf(i)+μ3×(T target -T driver ) For the inverse parameter set {Cr,Dr,kr} (stored in MAP) r (In the middle) When all enabling conditions are met and the current state is a reverse-to-positive state, the following update rules are used: Cr(i+1)=Cr(i)+μ4×(T target -T driver ) Dr(i+1)=Dr(i)+μ5×(T target -T driver )×tanh(ωout_f / ω0) kr(i+1)=kr(i)+μ6×(T target-T driver The recommended update step size μ1 to μ6 is 0.001 to 0.01. A smaller step size can prevent parameter abrupt changes and maintain the stability of the control system. As an alternative, the update algorithm can also use adaptive filtering methods such as recursive least squares (RLS) to achieve faster convergence speed and better noise suppression performance.

[0068] The online parameter updates of this invention are closely linked to the directional state: forward parameters are updated only in the forward assist state, and reverse parameters are updated only in the reverse return state. This "direction-bound parameter update" ensures that the two models independently adapt to the changes in transmission characteristics in their respective directions, avoiding the problem of mutual "contamination" of forward and reverse direction data in the unified update method, which leads to a decrease in compensation accuracy.

[0069] Example 1: Low-speed commutation operation under normal temperature conditions Scenario: A vehicle is making a U-turn at a city intersection at a speed of approximately 5 km / h, turning the steering wheel from the far left (approximately -540°) to the far right (approximately +540°), passing the center position (0°) in the middle.

[0070] Controller operation procedure: Step S1: Acquire signals from each sensor and process them using a 15Hz Butterworth filter.

[0071] Throughout the entire steering process, the directional state in step S3 undergoes a complete transition cycle of "positive assist → steering transition → reverse return to center": When the steering wheel moves from -540° to 0°: it is in positive assist mode; When the steering wheel passes near 0° (|Tsw) f |<0.5Nm or within 50ms after the reversing indicator changes): This is the reversing transition state; When the steering wheel continues to move from 0° to +540°: it changes to the reverse return-to-center state.

[0072] Steps S4 to S6: Use the first compensation model (including Cf and Ff) in the positive assist state, switch to the second compensation model (including Cr and Dr) in the reverse return state, and smooth the transition through λ weights and superimpose Tgap in the commutation transition state.

[0073] There is no jerking or free play when the steering wheel crosses the zero point, and the torque transition is smooth. Compared with existing unified compensation methods, the peak value of angular acceleration fluctuation is reduced by approximately 42%.

[0074] Example 2: Small-Angle Lane Keeping Correction in the Center Zone Scenario: The vehicle is traveling at 100 km / h on a highway, and the driver makes minor lane-keeping corrections, with the steering wheel angle changing within ±5°.

[0075] Controller operation procedure: Because of |ωout f |<0.2rad / s, tanh(ωout) f / ω0) is in the linear region, and friction compensation provides fine proportional compensation.

[0076] During minor corrections, the driver frequently switches between forward and reverse directions, but due to the hysteresis comparison and duration confirmation mechanism in step S3 (Tdz=0.5Nm, Tf) min =10ms), so there will be no frequent state jitter.

[0077] Results: The center area correction is lag-free, which helps to make the following smooth and natural, and the root mean square error is reduced by about 30% compared with existing methods.

[0078] Example 3: Small-angle correction in the central region under low-temperature conditions Scenario: On a winter morning with an ambient temperature of -25°C, after a cold start, the driver makes minor lane-keeping corrections on the highway (steering wheel angle change within ±10°).

[0079] Controller operation procedure: Step S1: Collect temperature Temp = -25℃.

[0080] Step S4: Read the positive parameters from the two-dimensional calibration table, Cf(θout≈0°,Temp=-25℃)=1.15 (15% higher than at room temperature), and increase Ff accordingly to compensate for the high viscosity resistance of the grease at low temperature.

[0081] Step S5: In the calculation of the positive compensation amount, since |ωout f | is very small (<0.1 rad / s), the tanh term is in the linear region, and provides fine friction compensation.

[0082] When the driver slightly releases the handle to return to center, step S3 identifies it as a reverse return to center state, switches to Cr=0.9 (higher reverse efficiency), and Dr is appropriately increased to provide sufficient return damping to prevent return overshoot.

[0083] Effect: Even at extremely low temperatures, the driver can still get a steering feel close to that at room temperature, with no lag or jerking in the center zone correction and smooth return to center.

[0084] The present invention has been described in the above-described embodiments; however, these embodiments are merely examples for implementing the present invention. It must be noted that the disclosed embodiments do not limit the scope of the present invention. Conversely, any modifications and refinements made without departing from the spirit and scope of the present invention are within the scope of patent protection of the present invention.

Claims

1. A method for bidirectional heterogeneous power assist compensation in commercial vehicles based on orientation state recognition, characterized in that, Includes the following steps: S1: Acquire the operating signal with a sampling period Ts and filter it using a second-order low-pass Butterworth filter to obtain the filtered steering wheel torque Tsw. f Output angular velocity ωout f and motor speed ωm f ; S2: Establish the equivalent mapping relationship from the motor side to the output end, and calculate the actual equivalent torque T at the output end. driver ; Among them, the equivalent conversion coefficient Kc is based on the reduction ratio i, the lead of the roller screw and the linear displacement-angular displacement conversion coefficient kn of the nut tooth sector, and is related to the output angle θout to reflect the position nonlinearity; S3: Based on the filtered steering wheel torque Tsw f and the output angular velocity ωout f By using hysteresis comparison logic and state duration confirmation, the current directional state is identified as one of the following: positive assist state, reverse return state, or reversal transition state. S4: Based on the current directional state obtained in S3, call the corresponding equivalent compensation model. The positive assist state adopts the first compensation model containing the efficiency compensation coefficient Cf and the friction compensation term Ff, and the reverse return state adopts the second compensation model containing the efficiency compensation coefficient Cr and the damping compensation term Dr. The physical composition of the first compensation model and the second compensation model is different, forming a bidirectional heterogeneous compensation. S5: Based on the current direction state identified in S3, select the corresponding compensation model to calculate the basic compensation amount Tcomp. base ; If the current state is a positive assist state, then substitute it into the formula of the first compensation model to calculate Tcomp. base =Tcomp f； If the current state is a reverse-to-positive state, then substitute it into the formula of the second compensation model to calculate Tcomp. base =Tcomp r； If the current state is a commutation transition state, then calculate Tcomp first. f and Tcomp r Two intermediate values, in this case, Tcomp is not output. base； S6: When the current direction state is a reversal transition state, the smoothing weight λ is calculated based on the filtered absolute value of the steering wheel torque, according to Tcomp. total =λ×Tcomp f +(1-λ)×Tcomp r +Tgap forms the total compensation amount Tcomp total , where Tgap is the pre-filling compensation term, whose value is proportional to the angular acceleration at the output end and is forced to be zero in the non-commutation transition state; S7: The total compensation amount is superimposed on the basic assist command to form the final assist output, which drives the EPS motor to work; S8: When the preset enabling conditions are met, the Cf, Cr, Ff, and Dr are updated online adaptively based on the output angle θout and temperature Temp. The positive parameters are updated only in the positive assist state, and the negative parameters are updated only in the negative return state.

2. The method as described in claim 1, characterized in that, In S3: the hysteresis comparison logic includes not immediately confirming entry into the forward or reverse state when the absolute value of the steering wheel torque rises from less than the torque dead zone threshold Tdz to more than Tdz, and only triggering the state switch when it falls from greater than Tdz to less than Tdz; the determination condition for the forward assist state is that |Tsw| is satisfied simultaneously. f |>Tdz、|ωout f |>ωdz, the desired direction of the assist is consistent with the direction of the output end movement, and continues for more than the first preset time; The determination condition for the reverse return-to-center state is that the desired assist direction is inconsistent with the output end movement direction, or the steering wheel torque decrease rate exceeds a preset rate threshold and the output end movement direction remains unchanged; the determination condition for the reversing transition state is |Tsw f |≤Tdz、or|ωout f |≤ωdz, or within the second preset time after the direction status flag changes.

3. The method as described in claim 1, characterized in that, In S4: the first compensation model is Tcomp f =Cf×T target +Ff×tanh(ωout f / ω0)+kf×e, where Ff is the positive friction compensation coefficient, in Nm; the second compensation model is Tcomp r =Cr×T target -Dr×tanh(ωout f / ω0)+kr×e, where Dr is the reverse damping compensation coefficient, in Nm; where T target For the desired output torque, e = T target - T driver To help track errors, ω0 is the velocity normalization parameter, and kf and kr are the positive and negative error correction coefficients, respectively, with kf > kr.

4. The method as described in claim 3, characterized in that, Cf and Cr are functions of the output angle θout and temperature Temp, respectively, and are obtained by looking up a pre-calibrated two-dimensional MAP table. Cf is 1.0 near the midpoint of the output and rises to 1.3 near the extreme position. Cr is less than Cf to reflect the characteristic that the reverse transmission efficiency is higher than the forward transmission efficiency.

5. The method as described in claim 1, characterized in that, In S6: the smoothing weight λ = sat(|Tsw) f | / Ttr,0,1), where Ttr is the preset commutation transition torque threshold, and sat() is the saturation function that limits λ between 0 and 1; the pre-filling gap compensation term Tgap=Kg×dωout f / dt, where dωout f / dt is the output angular acceleration, and Kg is the pre-filling compensation coefficient. Its physical meaning is the amount of torque required to pre-accelerate the total backlash of the execution chain to the equivalent inertia.

6. The method as described in claim 1, characterized in that, The preset enable conditions mentioned in S8 include: stable output angular acceleration (|dωout) f / dt|<preset threshold), system fault-free, current direction state stable for more than the preset time threshold, and driver input stable and valid; the online adaptive update is performed by recursive formula or recursive least squares method, and the update step size ranges from 0.001 to 0.

01.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 6.

Images