System for adapting the control of an electric bicycle motor to the resistance to forward travel without torque sensor

Abstract

The invention relates to a method for controlling an electric motor (12) of a crankset vehicle, comprising steps consisting in: measuring the instantaneous speed of the vehicle; measuring the rotation of the crankset; determining a torque at the reference crankset (m_T100) as a function of the speed of the vehicle, assuming that the power at the crankset is equal to a constant mean value (100 W); and using the torque at the reference crankset as the motor torque setpoint (Tm*).

Description

System for adapting the control of an electric bicycle motor to forward resistance without a torque sensor

[0001] The invention relates to improving the riding comfort of an electric-assisted bicycle without a pedal torque sensor. Background

[0002] In the case of electric-assist bicycles equipped with a torque sensor, a conventional control loop is configured to provide the motor with a torque (or current) command proportional to the torque supplied by the cyclist. The proportionality coefficient can often be selected by the cyclist from several pre-programmed levels, frequently referred to as "assistance levels." Thus, the "assistance level" is defined as the ratio between the motor torque and the torque at the pedals.

[0003] For a given level of assistance, the user must adapt their effort to changing riding conditions (variations in gradient, wind, terrain). For example, when climbing a hill, they must increase their pedaling effort if they don't want to slow down. If the perceived effort becomes too great, the user can select a higher level of assistance or a lower gear, or both.

[0004] To enhance riding comfort in varying conditions, eBikeLabs' patent application WO2023 / 135373 proposes an automatic and continuous adjustment of the assistance level based on a parameter representing actual variations in rolling resistance. Such a system requires pedaling torque information, which is provided in practice by a torque sensor mounted on the bottom bracket. This, in principle, precludes the use of this technology on more economical electric bikes that lack a torque sensor. Summary

[0005] A method for controlling an electric motor of a pedal-driven vehicle is generally envisaged, comprising steps of measuring the instantaneous speed of the vehicle; measuring the rotation of the pedals; calculating the variance of the vehicle speed over a sliding window corresponding to half a pedal revolution; calculating an estimated pedal torque as a function of the variance; and using the estimated pedal torque in the control of the motor.

[0006] The method may further include the following steps: determining a reference pedal torque as a function of vehicle speed, assuming that pedal power is equal to a constant average value; and using the reference pedal torque in the control when it is greater than the estimated pedal torque.

[0007] The method may include the step of using the reference crank torque in the control when the variance of the average speed over a sliding window is greater than a threshold.

[0008] The method may further include the following steps: determining a vehicle acceleration by differentiating the measurement of the vehicle speed; using the reference pedal torque in the control when the average acceleration over a sliding window is less than a negative threshold.

[0009] The estimated torque at the pedals may be proportional to the variance of the vehicle's speed.

[0010] The method may further include the following steps: determining an indicative parameter of a deviation in the vehicle's resistance to forward motion from nominal conditions, such as a linear combination of a measurement of the engine torque, indicative of an instantaneous force supplied by the engine, the estimated pedal torque, indicative of an instantaneous force resulting from pedaling, and a nominal force of resistance to forward motion, a function of the vehicle speed and constant friction coefficients determinable for nominal rolling conditions; and continuously modulating a level of vehicle assistance as a function of the indicative parameter of the deviation.

[0011] The process may further include the following steps: determining a vehicle acceleration by differentiating the measurement of the vehicle's speed; and including the acceleration in the linear combination.

[0012] The method may further include the following steps: determining a reference pedal torque as a function of vehicle speed, assuming that the pedal power is equal to a constant average value; and using the reference torque as the motor torque setpoint.

[0013] An alternative method for controlling an electric motor of a pedal-driven vehicle includes the following steps: measuring the instantaneous speed of the vehicle; measuring the rotation of the pedals; determining a reference pedal torque as a function of the vehicle speed, assuming that the pedal power is equal to a constant average value; and using the reference pedal torque as the motor torque setpoint.

[0014] The method may further include the following steps: determining an indicative parameter of a deviation in the vehicle's resistance to forward motion from nominal conditions, such as a linear combination of a measurement of the engine torque, indicative of an instantaneous force supplied by the engine, the estimated pedal torque, indicative of an instantaneous force resulting from pedaling, and a nominal force of resistance to forward motion, a function of the vehicle speed and constant friction coefficients determinable for nominal rolling conditions; and continuously modulating a level of vehicle assistance as a function of the indicative parameter of the deviation.

[0015] The process may further include the following steps: determining a vehicle acceleration by differentiating the measurement of the vehicle's speed; and including the acceleration in the linear combination.

[0016] The method may further include the following steps: calculating the variance of the vehicle speed over a sliding window corresponding to half a pedal revolution; calculating an estimated pedal torque as a function of the variance; determining an indicative parameter of a deviation of the vehicle's resistance to forward motion from nominal conditions, such as a linear combination of a measurement of the motor torque, indicative of an instantaneous force supplied by the motor, the estimated pedal torque, indicative of an instantaneous force resulting from pedaling, and a nominal resistance force, a function of the vehicle speed and constant friction coefficients determinable for nominal rolling conditions; and continuously modulating a level of vehicle assistance as a function of the indicative parameter of the deviation. Brief description of the drawings

[0017] Some embodiments will be described below, which is not exhaustive, in relation to the attached figures, among which:

[0018] This is a block diagram of a system for adapting the level of assistance according to the resistance to forward motion, such as that described in patent application WO2023 / 135373;

[0019] This is a graph representing an example of the evolution of a theoretical pedaling torque used for a simulation;

[0020] This is a graph representing an example of the simulated evolution of the instantaneous speed of the bicycle in response to the theoretical pedaling torque of the;

[0021] This is a graph representing the evolution of an average of the pedaling torque of the and an average of a reference torque;

[0022] Laet and lasont are graphs representing variances relative to the speed of las on two different scales;

[0023] Laest is a graph representing an estimated pedaling torque as a function of variance compared with an average of the theoretical pedaling torque and the average of the reference torque;

[0024] This is a graph representing an example of the evolution of the estimate of the gap in resistance forces provided by a system according to the estimated pedaling torque;

[0025] This is a block diagram of the assistance level adaptation system, according to a variant of its control configuration; and

[0026] This is a block diagram of the assistance level adaptation system, according to another configuration variant of its control. Detailed description

[0027] Several methods exist for estimating crank torque in an electric bike without a torque sensor, using other available sensors such as a speed sensor. For example, eBikeLabs' patent application EP4182215 proposes an estimation of the rider's effort based on a Fourier analysis of the crank rotation speed.

[0028] It turns out that these methods, given the discontinuities in the evolution of the estimate provided in the field and the lack of reproducibility, are not suitable for replacing the measurement of torque at the crankset in an automatic adaptation system of the level of assistance, in particular according to the aforementioned patent application WO2023 / 135373.

[0029] A method for estimating the instantaneous speed of a bicycle is described below. This method provides an indication of the pedal torque operating satisfactorily within an automatic assistance level adjustment system. Generally, a variance is calculated for the bicycle speed over a rolling time interval corresponding to half a pedal revolution. The variance thus calculated proves to be roughly proportional to the pedal torque at cruising speed. Recall that the variance is the square of the standard deviation, the standard deviation being the root mean square of the deviations from the average speed calculated over the period considered, namely half a pedal revolution.

[0030] The variance represents the amplitude of the velocity undulations due to pedaling. This amplitude is found to be correlated, particularly in steady-state conditions, with the effort exerted by the cyclist with each pedal stroke.

[0031] This is a block diagram of a system for adapting the level of assistance based on resistance to forward motion, such as that described in patent application WO2023 / 135373. A multiplier 10 receives information about human torque Th, normally the torque measured at the pedals. This torque is multiplied by an assistance level γ to provide a torque command Tm* to the motor 12.

[0032] The level of assistance γ is modulated according to a parameter ∆Fr representing the deviation of the resistance forces to forward motion from a nominal resistance force Fr ref The modulation is typically proportional.

[0033] The deviation parameter is determined by the force relationship:

[0034] ΔFr = Fm + Fh - Mt • Av - Fr ref

[0035] The parameter Fm is the motor force, determinable from measurements of the currents in the motor, directly indicative of the instantaneous torque supplied by the motor.

[0036] The parameter Mtest represents the mobilized mass and can be set to an average value for the bicycle, the cyclist, and their load.

[0037] The parameter Avest is the acceleration of the bicycle and can be determined by differentiating samples of bicycle speed measurement.

[0038] The parameterFr ref This corresponds to the friction opposing forward motion under nominal conditions, for example on a horizontal paved road. It depends on the mass Mt and the speed of the bicycle.

[0039] Finally, human force (Fh) is normally determined from a measurement of the torque at the crankset (Th), by multiplying it by the gear ratio and the radius of the drive wheel. Torque (Th) is normally measured when a torque sensor is available. Here, we use a human torque estimated from the variance of the instantaneous speed of the bicycle, calculated over a sliding window corresponding to half a pedal revolution. In practice, we use a motor speed sensor that provides a relatively large number of samples per pedal revolution, typically 100 samples per second.

[0040] In steady state, the estimated pedaling torque (Th) is determined proportionally to the variance (V), namely:

[0041] ~Th=a•V+ b, where a and b are constants.

[0042] During transient phases, particularly at startup, the variance proves to be less representative of pedaling torque. During these phases, an alternative torque estimation can be implemented, as will be explained later.

[0043] An example of calculating and implementing the estimated pedaling torque in a simulation of a bicycle equipped with a system that adapts the assistance level according to the control system is described below. This example uses the estimated torque for the setpoint Th. The simulation considers the case of starting on an 8% incline under standardized conditions for a city bike (average weight, paved road, no wind, etc.). For the purposes of the simulation, the cyclist is assumed to exert a theoretical pedaling torque generating an average pedaling power of 200 W. The simulated behavior of the bicycle in response to this theoretical pedaling torque is observed when the estimated torque corresponding to the control system is applied.

[0044] Laillume illustrates an example of the evolution of the theoretical pedaling torque Ths over 30 seconds, as parameterized for the simulation. It is assumed that the torque is sinusoidal and oscillates between 0 and a peak value based on the average power of 200 W.

[0045] At the very beginning, at near-zero speed, the torque corresponding to 200 W would be far too high for a cyclist to apply. The peak torque is therefore limited to what the cyclist can provide, for example, 100 Nm in this simulation. A 75 kg cyclist standing on 175 mm crank arms can exert a peak torque of approximately 130 Nm. The bicycle accelerates and the pedaling cadence increases.

[0046] From about 9 seconds, the average power of 200 W is reached and the peak torque decreases inversely proportionally to the speed.

[0047] From 20 seconds onwards, a substantially stable regime is reached under the conditions parameterized for the simulation.

[0048] This illustrates the evolution of the bicycle's instantaneous speed under simulated conditions. We observe that the speed oscillates with the pedaling rate, and that the oscillations decrease as the speed stabilizes, without, however, reaching zero. The speed tends asymptotically towards 25 km / h.

[0049] This illustrates an average theoretical pedaling torque. This average is calculated as the ratio between a constant power output of 200 W and the instantaneous rotational speed of the crankset. The crankset speed is measured by a rotation sensor or deduced from the instantaneous speed of the bicycle. During the starting phase, where the theoretical peak torque is limited to 100 Nm, the average is 50 Nm.

[0050] We seek to produce an estimated pedaling torque ~Thse approaching this theoretical average m_Th.

[0051] The graph also illustrates, in dotted lines, an average pedaling torque m_T100 calculated in the same way for a constant power output of 100 W, with a limit of 70 Nm instead of 50 Nm. This corresponds to comfortable nominal riding conditions, but a more energetic start. This curve can serve as an alternative to estimation based on variance in transient conditions, as will be seen later. Such a curve m_T100 can also be used in bicycles without a torque sensor to produce a normalized torque estimate, offering a smoother ride than using the motor in an on / off fashion when a pedaling speed threshold is exceeded.

[0052] The fact that the T100 starts with a relatively high torque means that, even if the rider doesn't reach that torque, the control system still receives that torque as a command and produces a corresponding, higher level of assistance than the rider actually "requests." This provides a temporary boost of extra assistance for starting.

[0053] Figures 5A and 5B represent, in solid lines, the variance V of the speed at two different scales. In the figure, the variance scale is in (km / h). 2 , is 2.5. In the, the lowest values ​​are visible using a scale of 0.05 (km / h) 2 . The varianceV tends to around 0.003 (km / h) 2 .

[0054] The dotted line represents the variance Vm of the average bicycle speed. It represents the degree of deviation from a steady state and is calculated over a sliding window of fixed width on the order of the average pedaling period, for example 1 second.

[0055] The varianceV and mean speed are calculated over a sliding window corresponding to half a pedal revolution. Since the pedaling cadence is variable, the calculation windows are also variable. To extrapolate the duration of a current calculation window, the synchronization pulses provided by the crankset's rotation speed sensor can be used. The crankset of an e-bike without a torque sensor can typically be equipped with a sensor providing 32 pulses per revolution. At each pulse, the current rotation speed can be estimated along with the time interval separating it from the previous pulse, and the duration of the current calculation window can be dynamically readjusted accordingly.

[0056] As previously indicated, the estimated pedaling torque ~Th can be expressed as a function of the variance V according to a linear law. In this example, it is expressed in Newton-meters as follows:

[0057] ~Th = 1250•V + 25(1)

[0058] Laillustre an evolution of the estimated pedaling torque using this formula (solid line) compared with the average of the theoretical pedaling torque m_Th (dotted line).

[0059] From approximately 17 seconds onward, once the uphill start has begun, the estimated torque ~Th closely follows relation (1) above. The estimated torque oscillates around the average of the theoretical torque m_Th, with an amplitude that decreases and tends toward the average of the theoretical torque. These oscillations are in phase with the oscillations of the instantaneous theoretical torque Th, which provides a pedaling torque estimate that is somewhat closer to reality than using a torque average.

[0060] Before 17 seconds, equation (1) produces estimated torque values ​​above what a cyclist could produce. The torque estimated according to equation (1) is then capped, for example to a value of 50 Nm, as shown.

[0061] To take into account transient phases producing a less representative varianceV, we can occasionally switch to an alternative method of estimating the torque, for example using the m_T100 curve mentioned in relation to the and represented in dotted lines on the.

[0062] A transition criterion between the torque curves can be the crossing of a threshold by the variance Vm of the mean speed (Figures 5A, 5B). An example of a threshold for the variance Vm, used in this simulation, is 0.4 (km / h). 2This threshold is briefly crossed (see) around 4 s and around 5 s. Each time the threshold is crossed, the estimated torque~Thest taken from the curvem_T100. This is what we observe on the estimated torque curve around these times.

[0063] Along with this criterion relating to the variance Vm, a second priority criterion can be applied according to which the curve m_T100 is taken for the estimated torque ~Th when this curve is above the value calculated according to relation (1), after clipping at 50 Nm. This is also what is observed at the.

[0064] Yet another criterion can be applied, according to which the curvem_T100 is taken for the estimated torque when the average acceleration of the bicycle over a sliding window, for example half a pedal revolution, is below a negative threshold. The acceleration threshold can be on the order of -0.05 m / s². 2This helps to avoid irregular behavior of the varianceV during phases of strong deceleration.

[0065] This illustrates the evolution of the estimated resistance force difference ∆Fr provided by the control system under the simulation conditions, particularly as a function of the estimated pedaling torque ~Th. The curve rises from 0 and stabilizes at approximately 80 N. The growth is relatively irregular due to adjustments made to the curve during the initial start-up phase, but its trend is correct, and what matters is that the value of 80 N reached at steady speed corresponds well to the value the system would have produced with an actual measurement of the pedaling torque. In fact, during the start-up phase, the curve is above the curve obtained with an actual measurement of the pedaling torque, meaning that the cyclist receives slightly more assistance during the start-up phase, which does not negatively impact the riding experience.

[0066] The corresponding assistance level of the simulated bicycle varies between 1 and 3 for the curve of the, knowing that the assistance level is limited to 3 and that this value is reached for ∆Fr ≈ 75 N.

[0067] This illustrates a variant configuration of the assistance level adaptation system. Instead of applying the estimated torque ~Th as the system setpoint, the curve m_T100 is applied. The estimated torque ~Th is then applied to the adder as a measure of the pedaling torque to calculate the difference in resistance forces ∆Fr. This variant allows for particularly stable operation that remains quite satisfactory in terms of riding comfort under more challenging conditions.

[0068] This presents a second variant of the assistance level adaptation system configuration illustrated in Figure 1. In this variant, instead of using the estimated torque ~Th as the setpoint for the system, the average value m_T100 is used. Furthermore, the value m_T100 is also applied to the adder as the pedaling torque value to calculate the resistance force difference ∆Fr. This approach allows for estimating resistance forces with a reduced computational load, as it avoids estimating the torque ~Th, although using ~Th provides a more accurate estimate of the resistance forces.

[0069] Although the invention has been described in relation to an electrically assisted bicycle, it applies to any pedal-driven vehicle equipped with an electrically assisted motor.

Claims

Method of controlling an electric motor (12) of a pedal-driven vehicle, comprising the following steps: measuring the instantaneous speed of the vehicle; measuring the rotation of the pedals; determining a reference pedal torque (m_T100) as a function of the vehicle speed, assuming that the power at the pedals is equal to a constant average value (100 W); and using the reference pedal torque as the motor torque setpoint (Tm*). A method according to claim 1, comprising the following steps: determining a parameter (∆Fr) indicative of a deviation in the vehicle's resistance to forward motion from nominal conditions, such as a linear combination of: a measurement of the engine torque, indicative of an instantaneous force supplied by the engine (Fm), the estimated torque at the pedals (m_T100), indicative of an instantaneous force resulting from pedaling, and a nominal force of resistance to forward motion (Fr ref), function of vehicle speed and constant friction coefficients determinable for nominal rolling conditions; and continuously modulate a level of assistance (γ) of the vehicle as a function of the indicative parameter of the deviation (∆Fr). A method according to claim 2, comprising the following steps: determining a vehicle acceleration by differentiating the measurement of the vehicle's speed; and including the acceleration in the linear combination. A method according to claim 1, comprising the following steps: calculating the variance (V) of the vehicle speed over a sliding window corresponding to half a pedal revolution; calculating an estimated pedal torque (~Th) as a function of the variance; determining a parameter (∆Fr) indicative of a deviation in the vehicle's resistance to forward motion from nominal conditions, such as a linear combination of: a measurement of the engine torque, indicative of an instantaneous force supplied by the engine (Fm), the estimated pedal torque (~Th), indicative of an instantaneous force resulting from pedaling, and a nominal resistance force (Fr). ref ), function of vehicle speed and constant friction coefficients determinable for nominal rolling conditions; and continuously modulate a level of assistance (γ) of the vehicle as a function of the indicative parameter of the deviation (∆Fr).

Images