Patent Application
20160009340
1. Field of the Invention
The present invention relates to a module for estimation of pedalling torque and consumed energy of bicycler, particularly to a non-mechanical module for estimation of pedalling torque and consumed energy of bicycler, which utilizes the measured bicycle speed, slope, and motor output torque to estimate the pedalling torque of bicycle and the consumed energy of bicycler.
2. Description of Related Art
To meet the demand for diversified functions of the bicycle, the electrically-assisted bicycle has become a major study subject of the bicycle manufacturer, and the pedaling torque sensor, which receives the sensed pedalling torque such that an on board intelligent module can determine the amount of motor torque output to assist the bicycler, is one of the key components of the electrical bicycle.
The conventional technology of the pedalling torque sensor, such as Japan Patent No. 5-246377, No. 5-310177, and Taiwan Patent No. 453317, No. 288427, No. 325034, is primarily of linkage mechanism, which converts the pedalling torque generated by the human into a linear or angular displacement proportionally, which is then further converted into a proportional voltage signal by a displacement sensor.
The prior arts mentioned above are all mechanical mechanisms, and assembly of such a mechanical mechanism takes extra time for the bicycle production. Besides, adding a torque sensor on to a bicycle raises the bicycle cost. Therefore, the present invention provides a non-mechanical module for estimation of pedalling torque in order to solve the aforementioned problems.
The objective of the present invention is to provide a non-mechanical module for estimation of pedalling torque and consumed energy of bicycler for a man-powered bicycle, wherein the measured bicycle speed and slope is utilized to estimate the pedalling torque and the consumed energy of the bicycler.
Another objective of the present invention is to provide a non-mechanical module for estimation of pedalling torque and consumed energy of bicycler for a electrical bicycle, wherein the measured bicycle speed, slope, and motor output torque is utilized to estimate the pedalling torque and the consumed energy of the bicycler.
To achieve the aforementioned objectives, the non-mechanical module for estimation of pedaling torque and consumed energy of bicycler of the present invention comprises an estimation program package embedded inside a single-chip microprocessor, a bicycle speed sensor, a slope sensor and a motor torque sensor, and the estimation program package further comprises: a feed-forward control program, a feed-back control program, a bicycle dynamics calculation program, a pedal torque calculation program, and a bicycler consumed energy calculation program, wherein with the preset parameters, such as rear wheel radius, mass of the bicycle and bicycler, gear ratio of the transmission, effective moment of inertia at the rear wheel, aero drag coefficient, and rolling resistance coefficient, and with the input variables, such as slope of the real bicycle position, forward speed of the real bicycle, and motor torque on the real bicycle, the feed-forward control program and the feed-back control program can provide a tracking control of the bicycle speed and output the results thereof to the bicycle dynamics calculation program, and the bicycle dynamics calculation program receives the outputs of the feed-forward control program and the feed-back control program and simulates the bicycle speed change under the action of the external forces and feeds the result back to the feed-back control program, and when the simulated speed worked out by the bicycle dynamics calculation program is the same as the object speed, i.e. the measured speed of the real bicycle, the results worked out by the feed-forward control program and the feed-back control program can represent the external forces acting on the bicycle and can be utilized by the pedal torque calculation program to calculate the estimated pedaling torque of the bicycler, and with the calculation result of the pedal torque calculation program, the bicycler consumed energy calculation program can work out the power output by the bicycler and the energy consumed by the bicycler, and further, the estimated pedaling torque of the bicycler can be utilized to determine the corresponding torque the motor needs to output. Furthermore, the estimation program package can be specifically designed to be a dedicated integrated circuit.
Via the detailed description of the preferred embodiments in cooperation with the attached drawings, the objectives, technical contents, characteristics and accomplishments of the present invention is to be more easily understood.
Refer to
The feed-forward and feed-back control algorithms are to generate a control effort so that the simulated bicycle speed can track the measured real bicycle speed. Then, the control effort is transformed algebraically to estimate the bicycler pedaling torque.
The algorithms of those programs mentioned above are stated below:
If there is no sliding motion between the rear wheel and the ground, deduced from the Newton's principle, the dynamics of the bicycle can be described by: (Tmotor+Trider)g.sub.r-T.sub.eff=J.sub.eff{dot over (.omega.)}.sub.wu=r.sub.w.omega..sub.w (1) wherein T.sub.motor is motor output torque;
T.sub.rider pedalling torque generated by the bicycler;
g.sub.r gear ratio of the transmission device;
T.sub.eff effective road loading on the rear wheel;
J.sub.eff effective moment of inertia at the rear wheel;
.omega..sub.w rear wheel speed;
u simulated bicycle speed in the estimation module;
r.sub.w rear wheel radius.
The effective road loading mentioned above can be expressed as: T.sub.eff=T.sub.r+r.sub.wF.sub.g+r.sub.wF.sub.a, (2) wherein F.sub.g is slope resistance,
F.sub.a aero drag,
T.sub.r rolling resistance, and the slope resistance, the aero drag, and the rolling resistance can be further respectively expressed as: F.sub.g=m.sub.sg sin .theta..sub.slope, (3) F.sub.a=C.sub.au.sup.2, (4) T.sub.r=r.sub.w.mu.m.sub.sg cos. theta..sub.slope, (5) wherein m.sub.s is mass of the bicycle and bicycler,
g gravity coefficient,
.theta..sub.slope slope of the real bicycle position,
C.sub.a aero drag coefficient,
.mu. rolling resistance coefficient; therefore, the effective road loading is obtained by inserting the above three equations into equation (2), and the result is T.sub.eff=r.sub.w.mu.m.sub.sg cos .theta..sub.slope+r.sub.wm.sub.sg sin .theta..sub. slope+r.sub.wc.sub.a(r.sub.w.omega..sub.w).sup.2. (6)
The effective moment of inertia at the rear wheel can be expressed as: J.sub.eff=J.sub.w+r.sub.w.sup.2 m.sub.s, (7) wherein J.sub.w is rear wheel moment of inertia.
If equation (6) and (7) are plugged into equation (1), the results is (T motor+T rider ).times.gr-(rw.times..mu..times..times.ms.times. g.times..times.cos.times..times..theta.slope+rw.times.ms.times. g.times..times.sin.times..times..theta.slope )=Jeff rw.times.u.+rw.times.c a.times.u2, (8) ##EQU1## wherein (T.sub.motor+T.sub.rider)g.sub.r can be looked as the dynamic input of the bicycle and (r.sub.w.mu.m.sub.sg cos .theta..sub.slope+r.sub.wm.sub.sg sin .theta..sub.slope) can be looked as the dynamic disturbance.
Subsequently, if a variable I is established such that I=(T.sub.motor+T.sub.rider)g.sub.r-(r.sub.w.mu.m.sub.sg cos .theta..sub.slope+r.sub.wm.sub.sg sin .theta..sub.slope); (9) then, equation (8) is simplified into=.times. Jeff rw.times.u.+rw.times.ca.times.u2=.times. J eff.times..omega..w+rw3.times.ca.times..omega.w2 (10) ##EQU2## which is the differential equation used in the bicycle dynamics calculation program 123 for dynamic simulation of the bicycle.
A feed-forward control program 121 and a feed-back control program 122 are then developed to generate I such that u=r.sub.w.omega..sub.w can track the measured real bicycle speed, u.sub.real. In other words, the measured bicycle speed is the tracking object of the control program u.sub.d (control object in the program 121, 122), i.e. u.sub.d=u.sub.real. In this situation, I is used in the following equation to calculate bicycler pedaling torque T̂rider=1gr.times.(−T motor.times.gr+r w.times..mu..times..times.ms.times.g.times..times.cos.times..times..theta. slope+rw.times.ms.times.g.times..times.sin.times..times..theta.slope), (11) ##EQU3## which is derived from equation (9), wherein {circumflex over (T)}.sub.rider is the estimated pedaling torque; the pedal torque calculation program 124 is designed according to equation (11).
As shown in
As shown in
When the estimated speed U worked out by the bicycle dynamics calculation program 123 is the same as the object speed, i.e. u=u.sub.real, I worked out by the feed-forward control program 121 and the feed-back control program 122 can be utilized to calculate the estimated pedaling torque {circumflex over (T)}.sub.rider according to equation (11). When the estimated pedaling torque {circumflex over (T)}.sub.rider is worked out, the power output by the bicycler {circumflex over (P)}.sub.rider can be calculated as {circumflex over (P)}.sub.rider(t)={circumflex over (T)}.sub.rider(t).omega..sub.w(t)g.sub.r, (19) and further the energy consumed by the bicycler can be calculated as Ŵrider.function. (t)=.intg.0t.times.P̂rider.function. (.lamda.) .times..times.d.lamda.. (20) ##EQU8##
To validate the estimation algorithm and study the performance of the estimation, a Simulink simulation code is developed and several different simulations are conducted. The results are discussed in below.
In the Simulink simulation code, a bicycle dynamics block is developed to simulate the forward speed of a MERIDA PC 400 electrical bicycle under the actuation of bicycler pedaling torque, motor torque and road loads. The specification of MERIDA PC 400 is listed in Table. 1. TABLE-US-00001 TABLE 1 Specification of MERIDA PC 400 Weight 40 kgw Gear ratio 3.0 Rear wheel radius 0.33 m Rear wheel weight 0.0118 kgw Aero drag coefficient 0.328 Rolling resistance 0.01 coefficient
The first simulation with the Simulink code is to validate the proposed estimation algorithm. In the simulation, the bicycle is driven on a flat surface and then meets a slope at 100 second. The bicycler then raises the pedaling torque to maintain the same speed. In the simulation, the slope, the motor torque, and bicycle speed are assumed perfectly measured. The bicycler pedaling torque features an amplitude of 20 N-m initially and 34 N-m after 120 second and a frequency of 0.5 rad/sec. The pedaling torque is a half-wave function, which mimics the real human pedaling; the torque is zero in between the positive wave. The desired estimation convergence rate (i.e. the desired close loop pole) is designed as 0.1 second. The speed of the bicycle is shown in
Beside the above simulation, several other similar simulations with differences in pedaling torque frequency and designed observation convergence rate are also conducted. It is noticed that appropriate convergence rate must be chosen with respect to the variation of pedaling torque frequency; a fast convergence rate tends to increase the error bias and a slow convergence rate tends to increase the error peak. Thus, for the real application, adaptive law must be developed to adjust the feed-back loop gain in real time. It is also noticed that the nominal speed for the calculation of close loop gain in equation (20) has little effect on the estimation performance. Thus, a fixed nominal speed can be used for the close loop gain design.
Next, the sensitivity of the estimation error with respect to the parameter deviation of the dynamics model in the estimation module from the real bicycle values is also studied. In each simulation, simulation conditions are the same as that in the previous simulations with the exception that one parameter value deviates from the real value for 10%. The maximum torque tracking error, average value of the torque tracking error, and standard deviation of the torque tracking error are recorded for each simulation. The results are shown in Table.2. Table.2 reveals that the ratio between peak values of the torque estimation error and the real torque are similar to the previous result. Furthermore, the average value and standard deviation of the ratios do not change significantly. Thus, a 10% deviation of the system parameter identification error is allowable for this purpose for the average performance. TABLE-US-00002 TABLE 2 Torque Tracking Errors With Respect To Parameter Value Deviations Parameters.rho..sub.p.rho..sub.avg.rho..sub.std No parameter 0.07 −0.0012 0.0526 deviation Bicycle and bicycler 0.07 −0.0124 0.0476 mass deviation Aero drag 0.07 −0.0139 0.0750 coefficient Rolling resistance 0.07 −0.0057 0.0684 coefficient
Finally, the effects of the measurement errors of the motor torque, slope, and bicycle speed on the estimation performance are studied. This issue is studied by adding a white noise to the measurements. The standard deviations of the white noises are set to be 5% the peak value of each variable measurement. The results are included in Table.3. The results show that the motor torque measurement noise and slope measurement noise do not introduce significant values on the torque tracking error. However, bicycle speed measurement error has significant effect on the result. Therefore, an appropriate filter is required to eliminate the relative measurement noise. For the real application, the filter design can be accomplished via collecting the measurement and identifying the spectrum of the measurement noise. Then, a band-limited filter can be designed. TABLE-US-00003 TABLE 3 Torque Tracking Errors With Respect To Measurement Noise Parameters .rho..sub.p. rho..sub.avg. rho..sub.std No 0.07 −0.0012 0.0526 measurement noise Motor torque 0.07 0.0039 0.0597 Slope 0.07 0.0035 0.0604 Bicycle speed 1.00 −1.0152 3.6298
Simulation results show that the torque estimation can track the real torque satisfactorily. Under the case of no measurement noise and no parameter value deviation, the peak value of the tracking error is about 7% the peak value of real torque except at the point of dramatic dynamics variation. The average value of the ratio between the torque track error and the real torque value is −0.0012 and the relative standard deviation is 0.0526. Simulation results also reveal that the estimated bicycler consumed energy follows the real consumed energy closely. The maximum estimation error is 1.25% the real consumed energy.
It is also noticed that appropriate convergence rate must be chosen with respect to the variation of pedaling torque frequency. Thus, for the real application, adaptive law is suggested to adjust the feed-back loop gain in real time. It is also noticed that the nominal speed for the calculation of estimation close loop gain has little effect on the estimation performance. Thus, a fixed nominal speed can be used for the close loop gain design.
The sensitivity of the estimation error with respect to the parameter deviation of the dynamics model in the estimation module from the real bicycle values is also studied. For a 10% deviation in the parameter values, the average value and standard deviation of the ratios do not change significantly. Thus, a 10% deviation of the system parameter identification error is allowable for this purpose.
Finally, the effects of the measurement errors of the motor torque, slope, and bicycle speed on the estimation performance are studied. The results show that the motor torque measurement noise and slope measurement noise do not introduce extra values on the torque tracking error. However, bicycle speed measurement error has significant effect on the result. Therefore, an appropriate filter is required to eliminate the relative measurement noise.
It is to be emphasized that those described above are only the preferred embodiments of the present invention and not intended to limit the scope of the present invention, and any equivalent modification or variation according to the spirit of the present invention is to be included within the scope of the present invention.
