A journal of IEEE and CAA , publishes
high-quality papers in English on original
theoretical/experimental research
and development in all areas of automation
Tinghua Li, Qinghua Yang, Xiaowei Tu and Bin Ren, "An Improved Torque Sensorless Speed Control Method for Electric Assisted Bicycle With Consideration of Coordinate Conversion," IEEE/CAA J. Autom. Sinica, vol. 7, no. 6, pp. 1575-1584, Nov. 2020. doi: 10.1109/JAS.2020.1003360
Citation:
Dong Shen, "Iterative Learning Control With Incomplete Information: A Survey," IEEE/CAA J. Autom. Sinica, vol. 5, no. 5, pp. 885-901, Sept. 2018. doi: 10.1109/JAS.2018.7511123
Tinghua Li, Qinghua Yang, Xiaowei Tu and Bin Ren, "An Improved Torque Sensorless Speed Control Method for Electric Assisted Bicycle With Consideration of Coordinate Conversion," IEEE/CAA J. Autom. Sinica, vol. 7, no. 6, pp. 1575-1584, Nov. 2020. doi: 10.1109/JAS.2020.1003360
Citation:
Dong Shen, "Iterative Learning Control With Incomplete Information: A Survey," IEEE/CAA J. Autom. Sinica, vol. 5, no. 5, pp. 885-901, Sept. 2018. doi: 10.1109/JAS.2018.7511123
Dong Shen (M'10-SM'17) received the B. S. degree in mathematics from School of Mathematics, Shandong University, Jinan, China, in 2005. He received the Ph. D. degree in mathematics from the Key Laboratory of Systems and Control, Institute of Systems Science, Academy of Mathematics and System Science, Chinese Academy of Sciences (CAS), Beijing, China, in 2010. From 2010 to 2012, he was a Post-Doctoral Fellow with the State Key Laboratory of Management and Control for Complex Systems, Institute of Automation, CAS. From 2016.02 to 2017.02, he was a visiting scholar at National University of Singapore (NUS), Singapore. Since 2012, he has been with the College of Information Science and Technology, Beijing University of Chemical Technology (BUCT), Beijing, China, where he is a Professor now. His current research interests include iterative learning control, stochastic control and optimization. He has published more than 70 refereed journal and conference papers. He is (co-) author of the monographs Iterative Learning Control under Iteration-Varying Lengths:Synthesis and Analysis(Springer, 2019), Iterative Learning Control with Passive Incomplete Information:Algorithm Design and Convergence Analysis(Springer, 2018), Iterative Learning Control for Multi-Agent Systems Coordination(Wiley, 2017), and Stochastic Iterative Learning Control (Science Press, 2016, in Chinese). Dr. Shen received IEEE CSS Beijing Chapter Young Author Prize in 2014 and Wentsun Wu Artificial Intelligence Science and Technology Progress Award in 2012
This paper conducts a survey on iterative learning control (ILC) with incomplete information and associated control system design, which is a frontier of the ILC field. The incomplete information, including passive and active types, can cause data loss or fragment due to various factors. Passive incomplete information refers to incomplete data and information caused by practical system limitations during data collection, storage, transmission, and processing, such as data dropouts, delays, disordering, and limited transmission bandwidth. Active incomplete information refers to incomplete data and information caused by man-made reduction of data quantity and quality on the premise that the given objective is satisfied, such as sampling and quantization. This survey emphasizes two aspects: the first one is how to guarantee good learning performance and tracking performance with passive incomplete data, and the second is how to balance the control performance index and data demand by active means. The promising research directions along this topic are also addressed, where data robustness is highly emphasized. This survey is expected to improve understanding of the restrictive relationship and trade-off between incomplete data and tracking performance, quantitatively, and promote further developments of ILC theory.
WITH the improvement of modern life, people pay more and more attention to health and personal safety. Almost every country has been calling for green travel and the shared bicycles are very popular in all big cities around the world, like ofo bicycle and mobike [1]. So the development of intelligent bicycles must be an important trend in the future [2].
Traditional bicycles are widely used in our daily life as an individual transportation tool, which has a merit to relieve the traffic pressure. Although bicycle travel is very environmental friendly, there are still some limitations. Since traditional bicycle is completely driven by cyclist, if the road is in a poor condition with various geometric complexity, the torque output by the cyclist may decrease as well as the speed of bicycle. Besides, it will be very laborious for people to ride on an uphill road.
Once lacking of the power of human input, the bicycle will not maintain running. As far as this major drawback is concerned, man-powered bicycles are not as good as electric bicycles.
Electric bicycle equipped with motor can solve this problem. The energy expenditure of riding an electric bicycle was 31% lower than the conventional bicycle on the uphill [3]. It has greatly extended the scope of application for traditional bicycle by providing assisted power with a motor [4]−[6]. Nowadays, electric bicycle has become one of the main vehicles in the world. Compared with the other vehicles powered by gasoline, electric bicycle can reduce the energy consumption, the air pollution and noise [7]. As we can see from Li et al. [8], the emissions of pollutants by electric bicycles per person per kilometer are several times smaller than those by motorcycles and cars, more or less equivalent to those by buses, and higher than those by bicycles.
For electric bicycles, cyclist can control the output of motor by adjusting the power controller to adapt different road conditions or individual preferences. However, electric bicycle also has its limitations. Since the cyclist can control the motor freely, the speed of electric bicycles may be too fast to be safe. It not only affects the service life of the motor, but also increases pollution.
Moreover, electric bicycles are heavy, when the motor runs out of electricity, it will be very hard for people to ride with his own human power.
In order to protect the environment as well as to achieve the assisted riding, the electric bicycle equipped with torque sensor has been produced. The sensor detects the pedal torque and converts it into an electrical signal, which is the input for the motor controller [9]. However, the torque sensor may be easily damaged since it contacts with the cyclist’s feet directly and the damage costs are not negligible [10].
Besides, the emerging development of connected and automated vehicles imposes a significant challenge on current vehicle control systems [11].
To solve the problems mentioned above, the best way is to design a torque estimation method to measure the pedaling torque in real time and provide assistance [12], [13].
Niki et al. [14] proposed a torque separation method based on high pass filter, they used high pass filter (HPF) to separate the human torque input from running friction. In 2015, Sankaranarayanan and Ravichandran [15] designed a torque sensorless controller for bicycles, but they did not take the disturbance force during cycling into account. The Fourier filtering was also applied to decouple the load torque and pedaling torque [16].
In 2017, Fukushimal and Fujimoto [17] presented a method for pedaling torque estimation using the recursive least square algorithm with multiple forgetting factors and they considered the condition for travelling on upward as well. In 2018, Rallo et al. [18] measured the wheel speed with a magnetic wheel encoder and showed that the pedaling cadence is correlated with the speed oscillation.
In addition, Bertucci et al. [19] analyzed the crank torque in road cycling on level and uphill, and the results show that when the pedaling cadences are the same, the crank torque profile differences are minimal between level ground and uphill road. In this way, we can see that the main factor affecting pedaling torque is cadences. The inertial measurement unit is also used to detect the angular velocity [20], [21].
Nevertheless, none of them focused on the speed control. The purpose of electric assistance is to keep the speed within a certain range in all circumstances. And it is worthwhile to note that although the inertial measurement unit was used, none of the above papers considered the coordinate conversion of angular velocity.
Similarly to the inertial navigation, in order to obtain the accurate positions, angular velocity based on the inertial coordinate system needs to be converted to another coordinate, which is consistent with the movement of the bicycle. For instance, Nilsson et al. [22] presented an open-source wireless foot-mounted inertial navigation module with an intuitive and significantly simplified dead reckoning interface, which can display the data after coordinate conversion.
Disturbance observer is often used to eliminate the effects of interference on system control, considering that there are many kinds of disturbances during cycling [23].
Therefore, in order to achieve a safe and convenient riding, it is very important to design an improved highly accurate speed control method for electric bicycles.
This paper proposes an improved speed control method for the electric assisted bicycle. A second-order low-pass filter is designed in disturbance observer to estimate and compensate disturbances during cycling. The quaternion method is used to convert the coordinate and users can set a motor auxiliary ratio which is variable with the real-time speed of bicycle. The result finally comes out that the speed can be self-adaptively stabilized in a set safety range despite of the disturbances during cycling.
The rest of the paper is organized as follows. Section II describes the dynamic model of the electric assisted bicycle system. In Section III, the quaternion method is used to re-coordinate the angular velocity which is measured by inertial measurement units (IMU). Besides, we simplify the calculation of pedaling torque. Section IV presents the design of low-pass filter (LPF) in disturbance observer to estimate and compensate the disturbances during riding. Section V is the introduction of motor assisted control. The stability of the proposed method is analyzed in Section VI. Simulation and results comparison are presented in Section VII. Finally, some conclusions are drawn and summarized in Section VIII.
Table I lists the acronyms and symbols explanations used in this paper.
Table
I.
List of Parameter
Parameter
Description
${E_t}$
Translational kinetic energy
${E_r}$
Rotational kinetic energy
$M$
Mass of an electric assisted bicycle and a cyclist
$V$
Velocity of bicycle
$J$
Moment of inertia of wheel
$\omega $
Angular velocity of wheels
$T$
Total kinetic energy of bicycle
$r$
Radius of rear wheel
${P_h}$
Gravitational potential energy of a bicycle
$h$
Vertical height of cycling
$l$
Length of a wheel movement
$\varphi$
Slope angle of uphill
$\theta$
Rotation angle of wheel
${F_{\rm{man}} }$
Pedaling force of cyclist
${F_{\rm{motor}} }$
Assisted power of motor
${T_{\rm{fric}} }$
Disturbances during cycling
$B$
Friction coefficient
${T_{\rm{man}} }$
Pedaling torque
${J_m}$
Moment of inertia of crank shaft
${\omega _0}$
Pedaling angular velocity
${\omega _1}$
Angular velocity of the large gear
${\omega _2}$
Angular velocity of pinion
${\omega _3}$
Angular velocity of rear wheel
$D$
Quaternion from inertial coordinate to geographic coordinate
$N$
Quaternion from geographic coordinate to carrier coordinate
$C$
Quaternion from inertial coordinate to carrier coordinate
$\Omega $
Absolute angular velocity vectors of carrier coordinate
${\Omega _g}$
Absolute angular velocity vectors of geographic coordinate
${\omega _X}$, ${\omega _Y}$, ${\omega _Z}$
Absolute angular velocity vectors on pedal carrier coordinate
Assuming that an electric assisted bicycle is running on an uphill road, the longitudinal dynamic model of the electric bicycle is shown in Fig. 1 and the forces during riding are indicated. We consider a rear-wheel-assisted electric bicycle, the motor is installed on the rear wheel, so as to collect the angular velocity and provide assisted power directly.
where $ T $ is the total kinetic energy of the bicycle.
Considering $V = \omega \times r$, where $ r $ is the radius of rear wheel, here we suppose that the front wheel has the same radius as the rear wheel, T can be written as
$$ T = \frac{1}{2}M{(\omega \times r)^2} + J{\omega ^2}. $$
(4)
Due to the uphill road, the gravitational potential energy of the bicycle $ {P_h} $ can be expressed as
$$ {P_h} = Mgh $$
(5)
where $ h $ is the vertical height of cycling.
Fig. 2 shows the uphill road riding, $ l $ is the length of the wheel movement, $ \varphi $ is the slope angle of the uphill, and $ \theta $ is the rotation angle of the wheel.
The Lagrange function satisfies $ \dfrac{d}{{dt}}\left(\dfrac{{\partial L}}{{\partial \theta }}\right) - \dfrac{{\partial L}}{{\partial \theta }} = Q $. Considering there are two major powers during cycling, which are the pedaling force of cyclist ${F_{\rm{man}}}$ and the assisted power of motor ${F_{\rm{motor}}}$, we have
$$ Q = {T_{\rm{man}}} + {T_{\rm{motor}}} . $$
(10)
At the same time, considering the disturbances during cycling ${T_{\rm{fric}}}$ and the impedance force between tires and ground, $ B $ is friction coefficient, the final Lagrange equation can be written as follows:
To get the pedaling torque ${T_{\rm{man}}}$, we can simplify the pedaling process into a circular motion and use moment formula $ M = {J_m}\dfrac{{d\omega }}{{dt}} $ to replace it, where $ {J_m} $ is the moment of inertia and $ \omega $ is the angular velocity of the pedal rotation.
Inertial measurement unit (IMU) is used to collect the pedal angular velocity, it is mainly composed of a three axes of gyroscope and an three directional accelerometer, which can measure the data of the three axes of the $ OXYZ $ coordinate system.
In order to measure the angular velocity of the pedal and the rear wheel, one IMU is installed directly on the right pedal of the bicycle, and the other IMU is firmly connected with the rear wheel to detect the angular velocity of the electric assisted bicycle.
Firstly, the coordinate must be mentioned. In this paper, there are three kinds of coordinate systems. The first one is the earth centered inertial coordinate system ($ i $ system). The origin of coordinate is located at the earth’s center, $ {O_i}{Z_i} $ is along the polar axis of the earth, $ {O_i}{X_i} $ and $ {O_i}{Y_i} $ are on the equatorial plane of the earth.
The second one is the geographic coordinate system ($ t $ system), where the coordinate origin is also the barycenter of the carrier. $ {O_t}{Z_t} $ axis points to the zenith along the local reference ellipsoid’s normal line, $ {O_t}{X_t} $ axis is along the local latitude line to the east and the $ {O_t}{Y_t} $ axis points to the north along the local meridian.
The third is the carrier coordinate system ($ b $ system), which is firmly connected with a carrier. In this paper, the inertial measurement unit is loaded on the pedal, so the pedal is the carrier. The origin of this coordinate is located at the barycenter of pedal. The $ {O_b}{X_b} $ axis points to the direction of bicycle heading on the plane of the carrier, the $ {O_b}{Y_b} $ axis is perpendicular to the $ {O_b}{X_b} $ axis and points to the right side. The $ {O_b}{Z_b} $ axis is perpendicular to the plane that consists of the $ {O_b}{X_b} $ axis and $ {O_b}{Y_b} $ axis.
$ {\omega _0} $ is the pedaling angular velocity, $ {\omega _1} $ is the angular velocity of the large gear and $ {\omega _2} $ is the angular velocity of the pinion. The angular velocity $ \omega $ that we mentioned in Section II is the angular velocity $ {\omega _3} $ of the rear wheel, which can be detected by the IMU. However, IMU measures the projection of the absolute angular velocity vector on the axis of the carrier coordinate system. To be exact, they are not equal to the previous $ {\omega _0} $ and $ {\omega _3} $.
In order to obtain the accurate angular velocity and acceleration to control the speed of the electric bicycle, it is necessary to get the real-time data of the carrier coordinate system (pedal) relative to the geographic coordinate system. In this paper, the quaternion method is used to convert the data measured in the carrier coordinate system to the geographic coordinate system [24].
Matveev et al. [25] proposed a coordinate transformation method in 2009. Assume that the quaternion of the transformation from the inertial coordinate system to the geographic coordinate system is $ D $, the quaternion of the transformation from the geographic coordinate system to the carrier coordinate system is $ N $, and $ C $ is the quaternion that only needs to be transformed once from the inertial coordinate system to the carrier coordinate system. It can be composed of the two intermediate coordinate transformations. Based on the principle of quaternion, we get
$$ C = D \circ N $$
(14)
where “$ \circ $” is the product symbol of the quaternions.
The purpose of the paper is to get the data of the coordinate system relative to the geographic coordinate system, so we need to calculate the quaternion $ N $.
Left multiplying both sides of (14) by the conjugate quaternion $ \bar D $, we can get
$$ \bar D \circ C = \bar D \circ D \circ N .$$
(15)
Considering $ \bar D \circ D = 1 $ and reversing the equation, we get
$$ N = \bar D \circ C . $$
(16)
Take the derivative of both sides of (16)
$$ \dot N = \dot {\bar D} \circ C + \bar D \circ \dot C . $$
(17)
Using the dynamic equation of quaternion $ 2\dot N = N \circ \Omega $ and $ 2\dot {\bar N} = - \Omega \circ \bar N $, $ \dot C $ and $ \dot {\bar D} $ are written as follows
$$ \dot N = - \frac{1}{2}{\Omega _g} \circ \bar D \circ C + \bar D \circ \frac{1}{2}C \circ \Omega . $$
(20)
Substituting (16) into (20), the final expression can be written as
$$ 2\dot N = N \circ \Omega - {\Omega _g} \circ N $$
(21)
where $ \Omega $ and $ {\Omega _g} $ are the absolute angular velocity vectors of the carrier coordinate system and geographic coordinate system. The three components of $ \Omega $ can be measured directly by the three axes gyroscopes in IMU.
Assuming $ {\omega _X} $, $ {\omega _Y} $ and $ {\omega _Z} $ are the absolute angular velocity vectors on the three axes of the pedal carrier coordinate system, $ \Omega $ can be expressed as
Assuming that $ {\omega _{Xg}} $, $ {\omega _{Yg}} $ and $ {\omega _{Zg}} $ are the absolute angular velocity vectors on the three axes of the geographic coordinate system, $ {\Omega _g} $ can be expressed as
where $ {\omega _{ie}} $ is the rotational angular velocity of the earth and the value is $7.292 \times {10^{ - 5}}\;{\rm{rad/s}}$, $ \lambda $ is the longitude and $ \alpha $ is the latitude.
Assuming $ N = {\lambda _0} + i{\lambda _1} + j{\lambda _2} + k{\lambda _3} $ and substituting $ \Omega $ and $ {\Omega _g} $ into (21), we can get
By solving the first order equation group (29), we can get the values of $ {\lambda _0} $,$ {\lambda _1} $,$ {\lambda _2} $,$ {\lambda _3} $, and the expression $ N = {\lambda _0} + i{\lambda _1} + j{\lambda _2} + k{\lambda _3} $.
$ N $ is the quaternion that transforms from the geographic coordinate system to the carrier coordinate system, while our purpose is to convert the coordinate reversely. Therefore, after measuring the absolute angular velocity $ {\omega _t} $ in the carrier coordinate system, the angular velocity $ {\omega _{gt}} $ of the carrier coordinate system relative to the geographic coordinate system can be obtained by
$$ {\omega _{gt}} = N \circ {\omega _t} \circ \bar N . $$
(30)
For the convenience of calculation, we simplify the human pedal model in this paper and the pedaling torque formula can be described as
where $ {\omega _0} $ is the angular velocity of pedal, $ {J_m} $ is the moment of inertia of the crank shaft. $ d $ and $ m $ are the length and mass of crank shaft respectively.
The chain transmission model of electric bicycle is shown in Fig. 4. With the principle of chain transmission, the rotational angular velocity of the rear wheel $ {\omega _3} $ can be calculated by the pedaling angular velocity $ {\omega _0} $. However, $ {\omega _3} $ is not proportional to $ {\omega _0} $ because of the disturbance factors such as inertia and friction during cycling. When the pedal stops rotating, $ {\omega _3} $ will still exist. That is why in this paper, an inertial measurement unit is loaded on the rear wheel to detect the angular velocity $ {\omega _3} $ in real time.
Figure
4.
Chain transmission model of the electric bicycle
As we mentioned in Section II, the electric bicycle will be disturbed by many kinds of frictions during riding. This paper mainly considers three kinds of frictions, namely the friction between tires and road, the wind resistance and the impedance force due to the gravitational potential energy.
On the basis of kinematic model, Huang et al. [26] used a sensor fusion algorithm as an observer. In this paper, in order to compensate the disturbances of frictions during riding, a disturbance observer is used to estimate the interference, as shown in Fig. 5 [27]. Considering the variability of frictions, we set a special cut-off frequency in LPF so that the majority of disturbances can be observed and compensated.
$ {G_1} $ is the model of the bicycle, we assume that the bicycle system is a first order model defined as $ {G_1}(s) = \dfrac{{\rm{1}}}{{{J_1}s + B}} $, $ {G_{\rm{2}}}(s) = {G^{{\rm{ -1}}}(s)} $, where $ {J_1} = M{r^2} + 2J $.
$ \tau $ is the disturbance during cycling and $\tau = - {T_{\rm{fric}}}$, it can be observed and compensated by disturbance observer described in this paper. In this way, the transfer function from $ \tau $ to the angular velocity of the wheel $ {\omega _3} $ can be written as
$ {Q_1} $ is a second-order low-pass filter (LPF), it is used in observer to filter out high-frequency [28], [29]. As shown in (36), when $ {Q_{\rm{1}}}(s) = 1 $ in the low-frequency stage, we get $\hat \tau = \tau$. The disturbance can be estimated in this way. Fig. 6 shows the circuit diagram of LPF.
where $ {A_u} $ is the pass-band gain, $ Q $ is quality factor, $ {\omega _c} $ is the cut-off angular frequency of LPF and $ {f_c} $ is the cut-off frequency.
The filtering effect on a certain frequency section can be achieved by selecting the different cut-off angular frequencies. In the low-frequency stage, $ {Q_{\rm{1}}}(s) = 1 $, $ {G_{T{w_{\rm{3}}}}} = {G_{\rm{2}}}^{{\rm{ - 1}}} $ and $ {G_{\tau {\omega _{\rm{3}}}}} = 0 $, the effect of disturbance can be compensated by $ {G_{\rm{2}}} $.
Similarly, in the high-frequency stage, $ {Q_{\rm{1}}}(s) = 0 $, the disturbance which is higher than $ {f_c} $ is effectively suppressed.
In general, people often pedal 65−80 times in a minute when they ride bicycles in a normal frequency. The corresponding frequency is from 1.083 Hz to 1.33 Hz. But the frequency can not be cut off immediately in the low-pass filter, because of the transition-band. Therefore, the cut-off frequency of LPF is set to be 4 Hz in this paper.
Taking the characteristics of the circuit into account, the resistance value cannot be too large, and the $ Q $ value is preferably less than 3. We set $ Q = 0.7 $, then the value of $ {R_1} $, $ {R_2} $, $ {C_1} $, $ {C_2} $ can be calculated.
The expression of LPF can be described as follows:
Fig. 7 shows the amplitude-frequency characteristic of designed LPF. We can see clearly that the curve shows a declining trend after 4 Hz. As indicated in the label, when it is at the cut-off frequency, the gain value is −2.93 dB, which is in accordance with the design of the low-pass filter.
Figure
7.
The amplitude-frequency characteristic of LPF
In order to test and verify the accuracy of disturbance observer $ \tau $, we consider the disturbance as sine wave, the expression is $ \tau = {\rm{2}}0\sin t $.
The simulation results are shown in Fig. 8. The dotted line represents the simulated disturbance $ \tau $ and the solid line is the estimated disturbance $ \hat \tau $ through observation. As we can see, there are small phase and amplitude differences between the two waveforms.
Fig. 9 captures and amplifies the waveforms from 0.6 s to 2 s. The two dashed lines indicate the amplitude values of the waveform vertexes respectively. It is clear that the difference in amplitude is 0.4, which proves that the observer can estimate the majority of disturbance.
Figure
9.
Difference between disturbance and observation
To achieve the purpose of the assisted riding, a permanent magnet brushless DC motor is used in this paper and installed in the rear wheel.
When the riding speed decreases and riding becomes more laborious, the motor can provide assistance based on the real-time angular velocity of wheel. As shown in Fig. 5, $ {\omega _{\rm{3}}} $ is fed back to the motor-assist module, and the auxiliary equation can be expressed as
$$ {T_{\rm{motor}}} = c{T_{\rm{man}}} $$
(41)
where $ c $ is the auxiliary coefficient determined in real-time by the wheel angular velocity.
Japanese industry proposed the PAP method, which has been adopted by regulation. The PAP method can be formulated by the linear piecewise equation [30]. Considering the motor assistance, we rewrite the auxiliary coefficient, and the specific relationship can be expressed as
where $ {\omega _{\min }} $ and $ {\omega _{\max }} $ are the pre-defined angular velocity limits. Once the wheel angular velocity $ {\omega _{\rm{3}}} $ is lower than the minimum limit $ {\omega _{\min }} $, the motor will provide assistance and the auxiliary coefficient $ c = 1 $. The motor and cyclist will provide power together with a proportion of 1:1.
When $ {\omega _{\rm{3}}} $ exceeds the maximum limit $ {\omega _{\max }} $, the motor will be reversed and $ c = {\rm{ - }}1 $. In this way, the electric bicycle will not be overspeed and become too dangerous.
As the permanent magnet brushless DC motor is used in this paper, it is equipped with position sensors. According to the detected difference of the rotor position, each phase coil will be electrified in turns, making the direction change of magnetic field, which is produced by stators. Therefore, we can use logic circuits to control the conduction sequence, so as to achieve motor inversions.
In the same way, if $ {\omega _{\rm{3}}} $ is in a range of $ {\omega _{\min }} \leq \omega \leq {\omega _{\max }} $, the auxiliary coefficient will vary between 0 and 1, so as to provide assistance intelligently.
As shown in Fig. 10, the predefined wheel angular velocity limits are $ {\omega _{\min }}{\rm{ = 12}} $ rad/s and $ {\omega _{\max }} = 16 $ rad/s respectively. The auxiliary coefficient values at different wheel angular velocities are shown in the figure.
where ${\omega _{\rm{wheel}}}$ is the angular velocity of the rear wheel, $ k $ is the transmission ratio between gears and ${\omega _{\rm{motor}}}$ is the angular velocity of motor.
In this paper, $ {\omega _{\rm{3}}} $ is ${\omega _{\rm{wheel}}}$, and the value of $ k $ is related to the radii of the gears.
For the permanent magnet brushless DC motor, the electromagnetic torque of the motor can be expressed as
$$ {T_{\rm{motor}}} = {K_t} \times I $$
(44)
where $ {K_t} $ is the motor constant, and $ I $ is the current of the motor winding.
The motor winding current $ I $ can be calculated as
where $ U $ is the average voltage of motor, $ {K_e} $ is the motor constant, ${\omega _{\rm{motor}}}$ is the angular velocity of motor and $ R $ is the winding resistance.
where $ A = \left[ c(x+\omega ^*)-c(\omega ^{*}) \right ] $. Since $ c $ is a function of $ \omega_3 $, $ \omega ^* $ is set between $\omega_{\rm{min}}$ and $\omega_{\rm{max}}$. Then $ x\neq 0 $, $ Ax <0 $. We also have $ Ax = 0 $ when $ x = 0 $.
Due to the fact that $J_m > 0,\; \dot{\omega}_0>0,\;B>0$, then $ \dot{V}(x) $ is negative-definte and $ \dot{V}(x) = 0 $ only when $ x = \omega_3-\omega^* = 0 $. Therefore, the system is asymptotically stable at $ \omega^* $.
VII.
Simulation
To show the validity of the proposed method, several simulations are implemented. For the convenience of this research, the signals for the input angular velocities of the three directions $ XYZ $ are all assumed to be sinusoid.
The expressions are depicted as $ {\omega _x} = 10\sin (2\pi t) + 10 $, $ {\omega _y} = 12\sin (2\pi t) + 12 $ and ${\omega _\textit{z}} = 14\sin (2\pi t) + 14$, respectively. The waveform is shown in Fig. 11.
Fig. 12 shows the angular velocity and three components curve diagram after coordinate transformation. We can see clearly that the value of angular velocity has changed from the original carrier coordinate system to the geographic coordinate system, so as to provide the subsequent calculation of the pedaling torque.
Take a set of data of the initial angular velocity as an example, $ {\omega _x} = 10 $, $ {\omega _y} = 12 $ and ${\omega _\textit{z}} = 14$. They make up an angular velocity matrix [0 10 12 14]T and the transformed matrix we get is [0 5.836 9.814 17.68]T. Thus, the quaternion from the geographic coordinate system to the carrier coordinate system can be calculated and expressed as N = [–0.9265 –0.216 –0.08526 –0.3018]T.
As we have got the value of transformed angular velocity, the pedaling torque can be obtained by the torque formula. The result is shown in Fig. 13.
With ${T_{\rm{man}}}$ and the angular velocity of the wheel, the motor module can provide assistance in real time.
Article 58 of China’s road traffic safety law stipulates that the maximum velocity of an electric bicycle riding on a non-motorized lane must not exceed 15 km/h.
In 2013, Zhao et al. [31] studied that the average velocity of a bicycle is 15 km/h on an isolated lanes between motor vehicles and non-motor vehicles with better riding conditions in China.
The electric bicycle safety specification stipulates that the maximum velocity of an electric bicycle shall not exceed 25 km/h, which is issued in China in May 2018. Before this specification, the Chinese national standard stipulated that the maximum velocity of electric bicycles was 20 km/h.
Considering that the electric assisted bicycle discussed in this paper is not a fully electric powered bicycle, we believe that velocity should be controlled according to the standard of the bicycle.
Therefore, the safe velocity of ordinary people riding a bicycle is between 12 km/h and 16 km/h. Assuming the radius of wheel $ r $ is 0.3 m and the formula $v = \omega \times r$, we can get the range of angular velocity is 11.1−14.8 rad/s.
Accordingly, the minimum limited angular velocity for the wheel can be set to 10 rad/s and the maximum limit is 13 rad/s. If the wheel angular velocity is lower than 10 rad/s, the motor and manpower will provide power together by the ratio of 1:1. If the wheel angular velocity is higher than 13 rad/s, motor will be reversed, so as to avoid danger.
The angular velocity variation curves of the rear wheel are shown in Figs. 14 and 15.
Figure
14.
Angular velocity of the wheel in previous methods
Three kinds of methods are simulated in our system, two of which are the previous method. A low pass filter is used in method 1. In method 2, low pass filter and band pass filter are used in combination to estimate disturbance and human pedaling torque.
In Fig. 14, the dotted line indicates the result of previous method 1, and the solid line represents the previous method 2. After setting the same safety angular velocity range for both of methods, the rapidity of method 2 is faster but fluctuation is more turbulent.
Fig. 15 shows the contrast curve between the results of this paper and the previous method 1. It can be seen that they can both well control the angular velocity but the rapidity and stability are not the same. Obviously, the method proposed in this paper is much faster and more stable, it takes reaction in the first 5 seconds.
As the cyclist begins to step on, the angular velocity value is the largest at the beginning in Fig. 15. When the value exceeds 13 rad/s, motor will turn to the reverse direction to decrease the value and the curve will show a downward trend. Later, when the value is between 10 rad/s and 13 rad/s, the curve rises up due to the motor’s assistance.
With the control and assistance, we can see clearly that the wheel angular velocity can still be effectively stabilized in a safe range, even though there are disturbances such as frictions and slopes during riding, so the same as the velocity of an electric bicycle.
VIII.
Conclusion
The contribution of this paper lies in the presenting of an improved speed control method for the electric assisted bicycle without torque sensor installed.
On the basis of dynamic model system, we combine the disturbance observer, coordinate transformation and motor assistance to realize the speed control.
Based on the simulation results above, the proposed method can well estimate and compensate the majority of disturbances during cycling, and improve the measurement and control accuracy by means of coordinate system conversion. Finally, the bicycle speed can be intelligently stabilized within a safe range with the motor assistance.
Compared with the previous methods, this new proposed method has a higher rapidity and stability than that of the previous methods. Meanwhile, the stability of the method is also analyzed. It further validates that the speed control method can not only assist the cycling under different road conditions, but also improve the accuracy.
S. Arimoto, S. Kawamura, and F. Miyazaki, "Bettering operation of robots by learning, " J. Robotic Syst. , vol. 1, no. 2, pp. 123-140, Jan. 1984. doi: 10.1002/rob.4620010203/abstract
[2]
D. A. Bristow, M. Tharayil, and A. G. Alleyne, "A survey of iterative learning control, " IEEE Control Syst., vol. 26, no. 3, pp. 96-114, Jan. 2006. http://ieeexplore.ieee.org/document/1636313/
[3]
H. S. Ahn, Y. Q. Chen, and K. L. Moore, "Iterative learning control: Brief survey and categorization, " IEEE Trans. Syst. Man Cybern. C, vol. 37, no. 6, pp. 1099-1121, Nov. 2007. http://ieeexplore.ieee.org/document/4343981/
H. S. Ahn and D. Bristow, "Special issue on 'iterative learning control', " Asian J. Control, vol. 13, no. 1, pp. 1-2, Jan. 2011. doi: 10.1002/asjc.334/abstract
[7]
C. Freeman and Y. Tan, "Iterative learning control and repetitive control, " Int. J. Control, vol. 84, no. 7, pp. 1193-1295, Aug. 2011.
[8]
D. Y. Meng and K. L. Moore, "Robust iterative learning control for nonrepetitive uncertain systems, " IEEE Trans. Autom. Control, vol. 62, no. 2, pp. 907-913, Feb. 2017. http://ieeexplore.ieee.org/document/7463016/
M. Yu and Y. C. Li, "Robust adaptive iterative learning control for discrete-time nonlinear systems with time-iteration-varying parameters, " IEEE Trans. Syst. Man Cybern. : Syst., vol. 47, no. 7, pp. 1737-1745, Jul. 2017. http://ieeexplore.ieee.org/document/7880611/
[11]
L. Hladowski, K. Galkowski, W. Nowicka, and E. Rogers, "Repetitive process based design and experimental verification of a dynamic iterative learning control law, " Control Eng. Pract. , vol. 46, pp. 157-165, Jan. 2016. https://www.sciencedirect.com/science/article/pii/S0967066115300344
[12]
H. F. Tao, W. Paszke, E. Rogers, H. Z. Yang, and K. Galkowski, "Iterative learning fault-tolerant control for differential time-delay batch processes in finite frequency domains, " J. Process Control, vol. 56, pp. 112-128, Aug. 2017.
[13]
S. Mandra, K. Galkowski, and H. Aschemann, "Robust guaranteed cost ILC with dynamic feedforward and disturbance compensation for accurate PMSM position control, " Control Eng. Pract., vol. 65, pp. 36-47, Aug. 2017. https://www.sciencedirect.com/science/article/pii/S0967066117301144
[14]
B. Altin and K. Barton, "Exponential stability of nonlinear differential repetitive processes with applications to iterative learning control, " Automatica, vol. 81, pp. 369-376, Jul. 2017.
[15]
Y. Q. Wang, H. Zhang, S. L. Wei, D. H. Zhou, and B. Huang, "Control performance assessment for ILC-controlled batch processes in a 2-D system framework, " IEEE Trans. Syst. Man Cybern.: Syst., 2017, doi: 10.1109/TSMC.2017.2672563.
T. T. Meng and W. He, "Iterative learning control of a robotic arm experiment platform with input constraint, " IEEE Trans. Ind. Electron., vol. 65, no. 1, pp. 664-672, Jan. 2018.
H. Kim, J. S. Lee, J. S. Lai, and M. Kim, "Iterative learning controller with multiple phase-lead compensation for dual-mode flyback inverter, " IEEE Trans. Power Electron. , vol. 32, no. 8, pp. 6468-6480, Aug. 2017. http://ieeexplore.ieee.org/document/7579576/
[20]
C. T. Freeman, "Robust ILC design with application to stroke rehabilitation, " Automatica, vol. 81, pp. 270-278, Jul. 2017.
[21]
H. S. Ahn, Y. Q. Chen, and K. L. Moore, "Intermittent iterative learning control, " in Proc. 2006 IEEE Conf. Computer Aided Control System Design, 2006 IEEE Int. Conf. Control Applications, 2006 IEEE Int. Symp. Intelligent Control, Munich, Germany, 2006, pp. 832-837.
[22]
H. S. Ahn, K. L. Moore, and Y. Q. Chen, "Discrete-time intermittent iterative learning controller with independent data dropouts, " IFAC Proc. Vol., vol. 41, no. 2, pp. 12442-12447, 2008. doi: 10.3182/20080706-5-KR-1001.02106
[23]
H. S. Ahn, K. L. Moore, and Y. Q. Chen, "Stability of discretetime iterative learning control with random data dropouts and delayed controlled signals in networked control systems, " in Proc. 10th Int. Conf. Control Automation, Robotics, and Vision, Hanoi, Vietnam, 2008, pp. 757-762.
[24]
S. S. Saab, "A discrete-time stochastic learning control algorithm, " IEEE Trans. Autom. Control, vol. 46, no. 6, pp. 877-887, Jun. 2001. http://ieeexplore.ieee.org/document/928588/
[25]
X. H. Bu and Z. S. Hou, "Stability of iterative learning control with data dropouts via asynchronous dynamical system, " Int. J. Autom. Comput., vol. 8, no. 1, pp. 29-36, Feb. 2011. doi: 10.1007/s11633-010-0551-3
[26]
X. H. Bu, Z. S. Hou, and F. S. Yu, "Stability of first and high order iterative learning control with data dropouts, " Int. J. Control Autom. Syst. , vol. 9, no. 5, pp. 843-849, Oct. 2011. doi: 10.1007%2Fs12555-011-0504-9
[27]
X. H. Bu, F. S. Yu, Z. S. Hou, and F. Z. Wang, "Iterative learning control for a class of nonlinear systems with random packet losses, " Nonlin. Anal. : Real World Appl., vol. 14, no. 1, pp. 567-580, Feb. 2013. https://www.sciencedirect.com/science/article/pii/S1468121812001423
[28]
X. H. Bu, Z. S. Hou, F. S. Yu, and F. Z. Wang, "H∞ iterative learning controller design for a class of discrete-time systems with data dropouts, " Int. J. Syst. Sci., vol. 45, no. 9, pp. 1902-1912, 2014. doi: 10.1080/00207721.2012.757815
[29]
X. H. Bu, Z. S. Hou, S. T. Jin, and R. H. Chi, "An iterative learning control design approach for networked control systems with data dropouts, " Int. J. Robust Nonlin. Control, vol. 26, pp. 91-109, Jan. 2016.
[30]
A. Hassibi, S. P. Boyd, and J. P. How, "Control of asynchronous dynamical systems with rate constraints on events, " in Proc. 38th IEEE Conf. Decision and Control, Phoenix, USA, 1999, pp. 1345-1351.
[31]
X. H. Bu, H. Q. Wang, Z. S. Hou, and Q. Wei, "Stabilisation of a class of two-dimensional nonlinear systems with intermittent measurements, " IET Control Theory Appl., vol. 8, no. 15, pp. 1596-1604, Oct. 2014.
[32]
J. Liu and X. E. Ruan, "Synchronous-substitution-type iterative learning control for discrete-time networked control systems with Bernoullitype stochastic packet dropouts, " IMA J. Math. Control Inf., 2017, doi: 10.1093/imamci/dnx008.
[33]
J. Liu and X. E. Ruan, "Networked iterative learning control for discrete-time systems with stochastic packet dropouts in input and output channels, " Adv. Differ. Equat., 2017, doi: 10.1186/s13662-017-1103-8.
[34]
J. Liu and X. E. Ruan, "Networked iterative learning control design for nonlinear systems with stochastic output packet dropouts, " Asian J. Control, vol. 20, no. 3, pp. 1077-1087, May 2018. doi: 10.1002/asjc.1457/full?scrollTo=references
[35]
D. Shen and Y. Q. Wang, "Iterative learning control for networked stochastic systems with random packet losses, " Int. J. Control, vol. 88, no. 5, pp. 959-968, 2015. doi: 10.1080/00207179.2014.986762?src=recsys
D. Shen, C. Zhang, and Y. Xu, "Intermittent and successive ILC for stochastic nonlinear systems with random data dropouts, " Asian J. Control, vol. 20, no. 3, May 2018. doi: 10.1002/asjc.1480/full
[39]
D. Shen, Y. Q. Jin, and Y. Xu, "Learning control for linear systems under general data dropouts at both measurement and actuator sides: a Markov chain approach, " J. Franklin Inst., vol. 354, no. 13, pp. 5091-5109, Sep. 2017. https://www.sciencedirect.com/science/article/pii/S0016003217302594
[40]
D. Shen and J. X. Xu, "A novel Markov chain based ILC analysis for linear stochastic systems under general data dropouts environments, " IEEE Trans. Autom. Control, vol. 62, no. 11, pp. 5850-5857, Nov. 2017. http://ieeexplore.ieee.org/document/7779121/
[41]
Y. Jin and D. Shen, "Iterative learning control for nonlinear systems with data dropouts at both measurement and actuator sides, " Asian J. Control, 2017, doi: 10.1002/asjc.1656.
[42]
D. Shen and J. X. Xu, "A framework of iterative learning control under random data dropouts: mean square and almost sure convergence, " Int. J. Adapt. Control Sign. Process., vol. 31, no. 12, pp. 1825-1852, Dec. 2017. http://shendongacademy.com/Publications/J-2017-14.pdf
[43]
Y. J. Pan, H. J. Marquez, T. W. Chen, and L. Sheng, "Effects of network communications on a class of learning controlled non-linear systems, " Int. J. Syst. Sci. , vol. 40, no. 7, pp. 757-767, Jan. 2009. https://dl.acm.org/citation.cfm?id=1568373.1568381
[44]
L. X. Huang and Y. Fang, "Convergence analysis of wireless remote iterative learning control systems with dropout compensation, " Math. Probl. Eng., vol. 2013, pp. Article No. 609284, Mar. 2013.
[45]
C. P. Liu, J. X. Xu, and J. Wu, "Iterative learning control for remote control systems with communication delay and data dropout, " Math. Probl. Eng., vol. 2012, pp. Article No. 705474, Jan. 2012.
[46]
W. J. Xiong, L. Xu, T. W. Huang, X. H. Yu, and Y. H. Liu, "Finiteiteration tracking of singular coupled systems based on learning control with packet losses, " IEEE Trans. Syst. Man Cybern.: Syst., 2018, doi: 10.1109/TSMC.2017.2770160.
[47]
T. Zhang and J. M. Li, "Iterative learning control for multi-agent systems with finite-leveled sigma-delta quantization and random packet losses, " IEEE Trans. Circuit. Syst. -I: Regul. Papers, vol. 64, no. 8, pp. 2171-2181, Aug. 2017. http://ieeexplore.ieee.org/document/7914678/
[48]
T. H. Gronwall, "Note on the derivatives with respect to a parameter of the solutions of a system of differential equations, " Ann. Math. , vol. 20, no. 4, pp. 292-296, Jul. 1919.
[49]
J. Liu and X. E. Ruan, "Networked iterative learning control approach for nonlinear systems with random communication delay, " Int. J. Syst. Sci., vol. 47, no. 16, pp. 3960-3969, Apr. 2016. doi: 10.1080/00207721.2016.1165894?src=recsys
[50]
J. Liu and X. E. Ruan, "Networked iterative learning control design for discrete-time systems with stochastic communication delay in input and output channels, " Int. J. Syst. Sci., vol. 48, no. 9, pp. 1844-1855, Feb. 2017. doi: 10.1186/s13662-017-1103-8
D. Shen, "Data-driven learning control for stochastic nonlinear systems: multiple communication constraints and limited storage, " IEEE Trans. Neural Netw. Learn. Syst., vol. 29, no. 6, pp. 2429-2440, Jun. 2018. http://ieeexplore.ieee.org/document/7920392/
[53]
T. Seel, T. Schauer, and J. Raisch, "Iterative learning control for variable pass length systems, " IFAC Proc. Vol., vol. 44, no. 1, pp. 4880-4885, Jan. 2011.
[54]
T. Seel, C. Werner, and T. Schauer, "The adaptive drop foot stimulator - multivariable learning control of foot pitch and roll motion in paretic gait, " Med. Eng. Phys., vol. 38, no. 11, pp. 1205-1213, Nov. 2016. https://www.sciencedirect.com/science/article/pii/S1350453316301230
[55]
T. Seel, C. Werner, J. Raisch, and T. Schauer, "Iterative learning control of a drop foot neuroprosthesis - generating physiological foot motion in paretic gait by automatic feedback control, " Control Eng. Pract. , vol. 48, pp. 87-97, Mar. 2016. https://www.sciencedirect.com/science/article/pii/S0967066115300484
[56]
R. W. Longman and K. D. Mombaur, "Investigating the use of iterative learning control and repetitive control to implement periodic gaits, " in Fast Motions in Biomechanics and Robotics. Berlin, Heidelberg: Springer, 2014, pp. 189-218. doi: 10.1007/978-3-540-36119-0_9
[57]
M. Guth, T. Seel, and J. Raisch, "Iterative learning control with variable pass length applied to trajectory tracking on a crane with output constraints, " in Proc. 52nd IEEE Ann. Conf. Decision and Control, Florence, Italy, 2013, pp. 6676-6681. https://www.sciencedirect.com/science/article/pii/S0016003216302381
[58]
T. Seel, T. Schauer, and J. Raisch, "Monotonic convergence of iterative learning control systems with variable pass length, " Int. J. Control, vol. 90, no. 3, pp. 393-406, 2017. doi: 10.1080/00207179.2016.1183172
[59]
X. F. Li, J. X. Xu, and D. Q. Huang, "An iterative learning control approach for linear systems with randomly varying trial lengths, " IEEE Trans. Autom. Control, vol. 59, no. 7, pp. 1954-1960, Jul. 2014. http://ieeexplore.ieee.org/document/6682999/
[60]
X. F. Li, J. X. Xu, and D. Q. Huang, "Iterative learning control for nonlinear dynamic systems with randomly varying trial lengths, " Int. J. Adapt. Control Sign. Process., vol. 29, no. 11, pp. 1341-1353, Nov. 2015. doi: 10.1002/acs.2543/abstract
[61]
X. F. Li and J. X. Xu, "Lifted system framework for learning control with different trial lengths, " Int. J. Autom. Comput., vol. 12, no. 3, pp. 273-280, Jun. 2015.
[62]
D. Shen, W. Zhang, Y. Q. Wang, and C. J. Chien, "On almost sure and mean square convergence of P-type ILC under randomly varying iteration lengths, " Automatica, vol. 63, pp. 359-365, Jan. 2016.
Y. S. Wei and X. D. Li, "Varying trail lengths-based iterative learning control for linear discrete-time systems with vector relative degree, " Int. J. Syst. Sci., vol. 48, no. 10, pp. 2146-2156, Apr. 2017.
[67]
S. D. Liu, A. Debbouche, and J. R. Wang, "On the iterative learning control for stochastic impulsive differential equations with randomly varying trial lengths, " J. Comput. Appl. Math., vol. 312, pp. 47-57, Mar. 2017. https://www.sciencedirect.com/science/article/pii/S0377042715005385
L. J. Wang, X. F. Li, and D. Shen, "Sampled-data iterative learning control for continuous-time nonlinear systems with iteration-varying lengths, " Int. J. Robust Nonlin. Control, vol. 28, no. 8, pp. 3073-3091, May 2018.
[70]
K. Abidi and J. X. Xu, "Iterative learning control for sampled-data systems: from theory to practice, " IEEE Trans. Ind. Electron., vol. 58, no. 7, pp. 3002-3015, Jul. 2011. http://ieeexplore.ieee.org/document/5559422/
[71]
J. X. Xu, K. Abidi, X. L. Niu, and D. Q. Huang, "Sampled-data iterative learning control for a piezoelectric motor, " in Proc. 2012 IEEE Int. Symp. Industrial Electronics, Hangzhou, China, 2012, pp. 899-904.
[72]
J. X. Xu, D. Q. Huang, V. Venkataramanan, and T. C. T. Huynh, "Extreme precise motion tracking of piezoelectric positioning stage using sampled-data iterative learning control, " IEEE Trans. Control Syst. Technol., vol. 21, no. 4, pp. 1432-1439, Jul. 2013. http://ieeexplore.ieee.org/document/6228523/
[73]
D. Q. Huang, J. X. Xu, V. Venkataramanan, and T. C. T. Huynh, "Highperformance tracking of piezoelectric positioning stage using currentcycle iterative learning control with gain scheduling, " IEEE Trans. Ind. Electron., vol. 61, no. 2, pp. 1085-1098, Feb. 2014. http://ieeexplore.ieee.org/document/6480838/
[74]
C. J. Chien, "The sampled-data iterative learning control for nonlinear systems, " in Proc. 36th Conf. Decision and Control, San Diego, California, USA, 1997, pp. 4306-4311.
[75]
C. J. Chien, "A sampled-data iterative learning control using fuzzy network design, " Int. J. Control, vol. 73, no. 10, pp. 902-913, Nov. 2000.
[76]
C. J. Chien, Y. C. Hung, and R. H. Chi, "Sample-data adaptive iterative learning control for a class of unknown nonlinear systems, " in Proc. 13th Int. Conf. Control, Automation, Robotics & Vision, Singapore, 2014, pp. 1461-1466.
[77]
C. J. Chien and C. L. Tai, "A DSP based sampled-data iterative learning control system for brushless DC motors, " in Proc. 2004 IEEE Int. Conf. Control Applications, Taipei, China, 2004, pp. 995-1000.
[78]
C. J. Chien and K. Y. Ma, "Feedback control based sampled-data ilc for repetitive position tracking control of dc motors, " in Proc. 2013 CACS Int. Automatic Control Conference, Nantou, China, 2013, pp. 377-382.
[79]
C. J. Chien, Y. C. Hung, and R. H. Chi, "On the current error based sampled-data iterative learning control with reduced memory capacity, " Int. J. Autom. Comput., vol. 12, no. 3, pp. 307-315, Jun. 2015. doi: 10.1007/s11633-015-0890-1
[80]
M. X. Sun, D. W. Wang, and G. Y. Xu, "Sampled-data iterative learning control for SISO nonlinear systems with arbitrary relative degree, " in Proc. 2000 American Control Conf. , Chicago, USA, 2000, pp. 667-671.
M. X. Sun, D. W. Wang, and Y. Y. Wang, "Sampled-data iterative learning control with well-defined relative degree, " International Journal of Robust And Nonlinear Control, vol. 14, no. 8, pp. 719-739, May 2004.
[83]
S. Zhu, X. X. He, and M. X. Sun, "Initial rectifying of a sampled-data iterative learning controller, " in Proc. 6th World Congress on Intelligent Control and Automation, Dalian, China, 2006, pp. 3829-3833.
[84]
M. X. Sun, Z. L. Li, and S. Zhu, "Varying-order sampled-data iterative learning control for MIMO nonlinear systems, " Acta Autom. Sinica, vol. 39, no. 7, pp. 1027-1036, 2013.
[85]
T. Oomen, M. van de Wal, and O. Bosgra, "Design framework for highperformance optimal sampled-data control with application to a wafer stage, " Int. J. Control, vol. 80, no. 6, pp. 919-934, Jul. 2007. doi: 10.1080/00207170701216329?scroll=top&needAccess=true
T. Oomen, J. van de Wijdeven, and O. H. Bosgra, "System identification and low-order optimal control of intersample behavior in ILC, " IEEE Trans. Autom. Control, vol. 56, no. 11, pp. 2734-2739, Nov. 2011.
[88]
T. Sogo and N. Adachi, "A limiting property of the inverse of sampleddata systems on a finite-time interval, " IEEE Trans. Autom. Control, vol. 46, no. 5, pp. 761-765, May 2001. http://ieeexplore.ieee.org/document/920797/
[89]
Y. Fan, S. P. He, and F. Liu, "PD-type sampled-data iterative learning control for nonlinear systems with time delays and uncertain disturbances, " in Proc. 2009 Int. Conf. Computational Intelligence and Security, Beijing, China, 2009, pp. 201-205.
[90]
P. Sun, Z. Fang, and Z. Z. Han, "Sampled-data iterative learning control for singular systems, " in Proc. 4th World Congress on Intelligent Control and Automation, Shanghai, China, 2002, pp. 555-559.
[91]
S. H. Zhou, Y. Tan, D. Oetomo, C. Freeman, and I. Mareels, "On on-line sampled-data optimal learning for dynamic systems with uncertainties, " Proc. 9th Asian Control Conf., Istanbul, Turkey, 2013, pp. 1-7, 2013. http://ieeexplore.ieee.org/document/6606377/
[92]
D. W. Wang, Y. Q. Ye, and B. Zhang, Practical Iterative Learning Control with Frequency Domain Design and Sampled Data Implementation. Singapore: Springer, 2014.
Y. Xu, D. Shen, and X. H. Bu, "Zero-error convergence of iterative learning control using quantized error information, " IMA J. Math. Control Inf., vol. 34, no. 3, pp. 1061-1077, Sep. 2017.
X. H. Bu, Z. S. Hou, L. Z. Cui, and J. Q. Yang, "Stability analysis of quantized iterative learning control systems using lifting representation, " Int. J. Adapt. Control Sign. Process. , vol. 31, no. 9, pp. 1327-1336, Sep. 2017.
[97]
W. J. Xiong, X. H. Yu, R. Patel, and W. W. Yu, "Iterative learning control for discrete-time systems with event-triggered transmission strategy and quantization, " Automatica, vol. 72, pp. 84-91, Oct. 2016.
[98]
T. Zhang and J. M. Li, "Event-triggered iterative learning control for multi-agent systems with quantization, " Asian J. Control, vol. 20, no. 3, pp. 1088-1101, May 2018.
[99]
W. J. Xiong, X. H. Yu, Y. Chen, and J. Gao, "Quantized iterative learning consensus tracking of digital networks with limited information communication, " IEEE Trans. Neural Netw. Learn. Syst., vol. 28, no. 6, pp. 1473-1480, Jun. 2017. http://ieeexplore.ieee.org/document/7425248/
[100]
D. Shen and J. X. Xu, "Zero-error tracking of iterative learning control using probabilistically quantized measurements, " Proc. 11th 2017 Asian Control Conf., Gold Coast, Australia, 2017, pp. 1029-1034.
[101]
Z. S. Hou and Z. Wang, "From model-based control to data-driven control: Survey, classification and perspective, " Inf. Sci., vol. 235, pp. 3-35, Jun. 2013.
[102]
E. Rogers, K. Galkowski, and D. H. Owens, Control Systems Theory and Applications for Linear Repetitive Processes. Berlin Heidelberg: Springer-Verlag, 2007.
Tinghua Li, Qinghua Yang, Xiaowei Tu and Bin Ren, "An Improved Torque Sensorless Speed Control Method for Electric Assisted Bicycle With Consideration of Coordinate Conversion," IEEE/CAA J. Autom. Sinica, vol. 7, no. 6, pp. 1575-1584, Nov. 2020. doi: 10.1109/JAS.2020.1003360
Tinghua Li, Qinghua Yang, Xiaowei Tu and Bin Ren, "An Improved Torque Sensorless Speed Control Method for Electric Assisted Bicycle With Consideration of Coordinate Conversion," IEEE/CAA J. Autom. Sinica, vol. 7, no. 6, pp. 1575-1584, Nov. 2020. doi: 10.1109/JAS.2020.1003360