Research Article  Open Access
Keke Zhang, Chaoming Si, Zhencai Zhu, Chongbin Guo, Qi Shi, "A TwoDimensional Solar Tracking Stationary Guidance Method Based on FeatureBased Time Series", Mathematical Problems in Engineering, vol. 2018, Article ID 3420649, 12 pages, 2018. https://doi.org/10.1155/2018/3420649
A TwoDimensional Solar Tracking Stationary Guidance Method Based on FeatureBased Time Series
Abstract
The amount of satellite energy acquired has a direct impact on operational capacities of the satellite. As for practical high functional density microsatellites, solar tracking guidance design of solar panels plays an extremely important role. Targeted at stationary tracking problems incurred in a new system that utilizes panels mounted in the twodimensional turntable to acquire energies to the greatest extent, a twodimensional solar tracking stationary guidance method based on featurebased time series was proposed under the constraint of limited satellite attitude coupling control capability. By analyzing solar vector variation characteristics within an orbit period and solar vector changes within the whole life cycle, such a method could be adopted to establish a twodimensional solar tracking guidance model based on the featurebased time series to realize automatic switching of featurebased time series and stationary guidance under the circumstance of different angles and the maximum angular velocity control, which was applicable to nearearth orbits of all orbital inclination. It was employed to design a twodimensional solar tracking stationary guidance system, and a mathematical simulation for guidance performance was carried out in diverse conditions under the background of inorbit application. The simulation results show that the solar tracking accuracy of twodimensional stationary guidance reaches 10^{∘} and below under the integrated constraints, which meet engineering application requirements.
1. Introduction
As microsatellites are featured with being light, small, smart, and cheap, they not only have become the research hotspot of satellite technology development and application, but are growing towards a direction of high functional density and practical applications. In order to adapt to multitype tasks and diversified applications, the requirement for higher payload power appropriate for high functional density microsatellites has been increasingly intense for the consideration of higher cost performance, rapid deployment, and extensive applicability. However, energy maximization must be taken into account with an aim to make microsatellites applied in highpower loads. Specific to application characteristics of microsatellites cluster or network, the scheme of energy maximization should be universal. Especially for the typical nonsolar synchronous inclined orbits such as Globalstar system, the solar panels adopting a solar tracking revolving mechanism is the only approach to solving energy acquisition issues provided that the platform itself remains an earthoriented demand.
In order to maximize the output power improving the efficiency, many methods and devices of sun tracking and maximum power point tracking were proposed, which belong to the special issue of solar energy application in ground photovoltaic systems. There are some fuzzy techniques that were proposed in photovoltaic systems [1–3], such as a method based on simultaneous use of two fuzzy controllers in order to maximize the generated output power of a solar panel. The sun tracking is performed by changing the solar panel orientation in horizontal and vertical directions by two DC motors properly designed. Several solutions for two axis solar tracking systems based on solar maps or based on tetrahedron geometry were studied, which can predict the exact apparent position of the sun or the strongest intensity of visible light [4–6]. Dualaxis solar tracking design utilizes a fourquadrant light dependent resistor (LDR) sensor or adaptive solar sensor and so on. The control of moving the mechanical structure uses a low power microcontroller or a traditional PLC [7–10].
For satellite, the maximum energy acquisition purpose is the same that is achieved generally by solar array drive assembly (SADA). As the microsatellites pursue a high performanceprice ratio strictly limiting development costs, it is impossible to select expensive SADA with slip ring. The lightsmall twodimensional turntable could be used to carry the solar panels to realize solar tracking. For the convenience of openloop control, stepping motor was utilized as a drive element. Nevertheless, the stepping motor still has some shortcomings such as poor dynamic behavior, high pulse step overshoot, and great rate ripple, and it is a nonnegligible interference source as far as satellites with a high requirement for attitude stabilization, such as earth observation satellites. Now, satellite rotating mechanism has been extensively investigated at home and abroad, covering mechanism models, groundbased validation, controller design, and attitude control coupling [11–14].
Solving the problem related to rotating mechanism motion stability of solar panels is a key to guaranteeing high pointing accuracy and high attitude stabilization of satellites. Factors that affect rotating mechanism motion stability can be elaborated from the following two aspects. Firstly, cogging torque of the stepping motor leads to instability of control, which can be resolved by means of compensation, and research on this issue has been relatively mature. SPOT satellite of France adopts a great subdivision sine/cosine driving strategy and an accelerometer to measure perturbed moment and implement compensation [15]. Engineers of NASA also introduce an input molding technique [16]. Additionally, an adaptive uniaxial driving strategy based on current compensation has been explored [17]. Secondly, due to instability of input angular velocity guiding turntable motion, the accurate angular velocity in the case of twodimensional solar tracking becomes unstable. However, such an issue is usually neglected under the circumstance of unlimited capacity so that relevant investigations can be rarely seen as well.
A solar tracking stationary guidance method based on featurebased time series was proposed under the background of microsatellite solar panels loaded with twodimensional turntable. Through phase division for time within an orbit period, the target velocity was divided into several segments of steady speeds, so as to solve guidance input instability problems and substantially reduce variable speed control. The highstability and smalldisturbance control of the twodimensional turntable can be fulfilled.
2. Problem Description and Modeling
Concerning a microsatellite that adopts orbital inclination of 55^{∘}, angle (the included angle between solar vector and orbital plane) varies between +78^{∘} and . Simultaneously, the corresponding tasks request the satellite to remain earthoriented.
Specific to the microsatellite characteristics, when angle is close to , the included angle formed by solar vector and normal of the solar panels should be greater than if the conventional onedimensional SADA has been utilized. In this case, solar panels are deemed to acquire no energy. Considering that the satellite runs in a 55^{∘} inclined orbit, such a phenomenon exists within a period long enough to incur unbalanced satellite energy supply making the entire satellite unable to run normally. Resultantly, twodimensional rotation must be conducted to realize solar tracking orientation of the panels under the circumstance that earthoriented status of the satellite is maintained, which is able to maximize energy acquired. As the twodimensional turntable is employed, 360degree continuous rotation cannot be completed; thus an appropriate twodimensional solar tracking stationary guidance strategy should be formed to guarantee that the satellite can acquire enough energies for operation in a condition of different angles.
2.1. Basic Assumption
The twodimensional turntable mechanism makes use of a stepping motor. Without loss of generality, friction moment, fluctuating moment, cogging torque, and mechanism dynamics were taken into account, while mutual induction winding, highorder harmonic torque, and external disturbance torque of the motor were ignored. Based on a cogging moment compensation method, velocity stability of the turntable could be improved. In addition, it has been assumed that the angular velocity of the turntable can be matched with the guidance input.
2.2. Constraint Condition
2.2.1. TwoDimensional Rotation Angle Limitation Constraint
With regard to the microsatellite, area of its solar panels unfolded is very large so that it is able to occlude payload or sensor field of view of its planes and . The maximum rotation angle of solar panels defined according to structural configuration and field of view requirements for satellites can be denoted by . In this case, principal axis of the solar panels (normal direction of the panel) is only able to rotate within a circular cone of semicone angle that adopts the satellites radius vector (i.e., the earth’s core points at the satellite) as its axis. Furthermore, the value of was defined to be 90^{∘} dependent on physical design outcomes.
2.2.2. Attitude Control Capacity Constraint
Due to the influence of the capacity possessed by an actuator selected for the microsatellite, the maximum angular velocity and the maximum angular acceleration of the turntable should be no more than 0.2^{∘}/s and 0.01^{∘}/, respectively, for the purpose of ensuring that disturbance to the platform during turntable running satisfies the earthoriented stabilization control requirement.
2.2.3. Pointing Error Constraint
Within the range of limited maximum rotation angle denoted as , angle tracking error of solar tracking should be no more than 14^{∘} to guarantee that the amount of energy acquired can be maximized.
2.3. Model Input
refers to a normalized solar vector in an orbital coordinate system.
Not only can solar vector be acquired by an onboard sun sensor or obtained through calculations based on sun ephemeris, but its expression in an orbital coordinate system is obtained in line with an attitude transformation relation.
stands for onboard cumulative seconds relative to onboard starting time.
2.4. Model Output
In conformity with input and constraint conditions, featurebased time series and angle/angular velocity series required by turntable control can be autonomously calculated to form target angles and angular velocities controlling the turntable in diverse time points. These series consist of pitching direction control time series , azimuth direction control time series , pitching angular velocity series , and azimuth angular velocity series . The corresponding output model can be expressed in
3. Guidance Law Design
Time series is a set of variables ordered chronically and the solar tracking angular relation can be described by multiple time series in one orbit. Therefore, characteristic time points should be selected to form time series for guidance implementation. Then, guidance requirement of the specified accuracy is fulfilled through density control over the series.
Design purpose of the twodimensional solar tracking stationary guidance law is to guarantee maximization of energies acquired by the solar panels under constraints.
3.1. Optimal Pointing Guidance Law in Angle Limitation Constraint
Under constraint described in Section 2, that is, in a condition of twodimensional rotation angle limitation, conical surface formed by the maximum limited angle constituted by the principal axis of solar panels intersects with that formed by rotation of the sun around orbital plane normal of the satellite. Optimal pointing of the solar panels principal axis is the combination of two segments of arcs, as shown in Figure 1. In general cases, two conical surfaces mentioned above directly intersect with each other or intersect in a translation manner among the conical surfaces, as points 1–4 presented in Figure 1.
When the satellite is in a light region, solar tracking guidance strategy of the panels can be illustrated as follows. On one hand, while the solar vector is within a constrained conical plane formed by the maximum rotation angle of the solar panels principal axis, the panel can track the sun, which is referred to as sunoriented; on the other hand, if the solar vector is outside the constrained conical plane described above, solar panels can track projection of the solar vector on the conical plane, which is known as quasisunoriented.
When the satellite is in a shadow region, solar tracking of panels principal axis has no significance. Considering this, only a guidance strategy of “rotating to the symmetric position in constant speed via the shortest path” is proposed. In other words, when the satellite is in a shadow region, rotating mechanism of the solar panels rotates its principal axis to the pointing at the moment of moving out of the shadow region from that when it moves into the shadow region along the shortest path. As azimuth pointing of the solar panels entering or exiting the shadow region is the complementary opposite number and the pointing and their pitch angle are identical, they are called symmetric positions.
By analyzing different temporal intervals, the included angle between sun pointing and zenith direction is the target attitude angle . In addition, the maximum included angle formed by panel pointing and the zenith direction is the limited angle . As for the target attitude angle of demarcation points distinguishing shadow and light, it is denoted by representing earth shielding angle. Then, the following conclusions can be drawn.
Point 1 to 2, the light region, : the solar vector is located in the range of solar panels’ maximum rotation angle, which indicates that the panel points at the solar vector.
Point 2 to 3, the light region, : solar panels point at projection of the solar vector in the conical plane of the maximum rotation angle.
Point 3 to 4, the shadow region, : the panel rotates to the pitch symmetrical attitude at a uniform velocity along the shortest path.
Point 4 to 1, the light region, : the panel points at the projection of solar vector in the conical plane of the maximum rotation angle.
To sum up, solar tracking guidance strategy of panels can be concluded as follows.
If , sunoriented mode can be established.
If , projection axisoriented mode can be established.
If , it rotates to the symmetrical position at a uniform velocity along the shortest path.
3.2. TwoDimensional Solar Panel Guidance Law in a Synthetic Constraint Condition
Subjected to synthetic constraint conditions described in Section 2, azimuth angular velocities are very high (see simulation results in Section 4) in terms of azimuthelevation type turntables when angle is rather small. It fails to satisfy torque requirement of attitude control disturbances. Therefore, only rotation in pitching direction is conducted under the circumstance that angle is small. As for large and small angles, the pattern partition strategy is utilized to carry out independent guidance.
According to the constraint of angle tracking error should be no more than described in Section 2, in order to retain a certain allowance, and below is defined as small angle, while that greater than or equal to is defined as large angle. Furthermore, calculation formula of angle is given below.where is component of solar vector along direction in an orbital coordinate system.
To meet relevant synthetic constraint conditions, autonomous switching of featurebased time series was implemented for solar tracking of panels according to large and small angles for the purpose of performing independent guidance. Within an orbit period, several characteristic time slots were segmented to adjust angular velocities. A twodimensional rotation mode should be adopted in the case of large angle to realize twodimensional featurebased time series guidance. By contrast, if angle is small, onedimensional rotation mode was used to conduct onedimensional featurebased time series guidance.
3.2.1. Large Angle Guidance Strategy
The difficulty of the twodimensional rotation mode lies in constraints over angular velocity and angular acceleration. Besides, it is less likely for azimuth angular velocity across the zenith to meet demands if angle is small. On this basis, the strategy can be elaborated as follows.
Expression of solar vector in the orbital coordinate system is used to figure out the next time point when the sun moves across zenith of the satellite and the number of characteristic time points within an orbit period. Regarding the latter, it incorporates the maximum angular velocity equilibrium point, the maximum limited angle time point, and the light and shadow region demarcation time point. Then, time of one orbit is divided into several time slots according to characteristic time points to calculate the mean angular velocity under the circumstance that two points on both ends of each time slot remain consistent with the target attitude, to update the time series dependent on the calculated maximum angular acceleration, and to finally output time series and angular velocity series needed by the turntable. Moreover, time slots in the time series can be further segmented by means of dichotomy to increase the number of series subsections, so that angular accuracy in the entire guidance process can be improved.
As regards characteristic time points in the minimum featurebased time series, they can be computed in the following ways.
(1) Time point across zenith of the satellite:where refers to orbit period and can be figured out in line with satellite orbit by the following equation:where km, , and ; furthermore, and are the semimajor axis and mean elements of orbital inclination.
(2) Maximum angular velocity equilibrium time point : it is calculated by a Newton iteration method and the time function can be written as the following equation.
(3) Maximum limited angle time point :
(4) Light and shadow region demarcation time point :
(5) Other time points are calculated by a symmetric method as follows:
Solar panels guidance procedure in a condition of large angle has been shown in Figure 2.
Due to time point calculation error accumulation and that of angular deviation in a condition of angular velocity control, time points in the time series and angular velocities in angular velocity series should be corrected during execution of guidance strategy. While correction of time points was realized by updating covering time, angular velocities were corrected by actual turntable feedback angle recalculations of switching time points.
3.2.2. Small Angle Guidance Strategy
Under the circumstance of small angle, time of an orbit period was divided into four segments according to the light and shadow region demarcation point and limited angle switching point (see Section 3), so as to work out their angular velocities. Moreover, two of such four segments fall into the category of quasisunoriented. In a condition of small angle, the solar panels guidance procedure is illustrated in Figure 3.
Under the circumstance of onedimensional rotation, the solar vector rotates within the orbital plane at a uniform velocity. Therefore, the sunoriented pitch angular velocity can be calculated by the following equation.
Angular velocity of the shadow region means going back to the symmetric position within the time slot of this region. In the onedimensional condition, pitch angle of such a position is the limited angle. For pitch angular velocity of the shadow region, it can be figured out by the following equation.where is calculated based on the earth shielding angle; that is, .
4. System Simulation and Analysis
4.1. SunOriented Angle of Limited Angle Constraint Based on a Pattern Partition Strategy
Under constraint described in Section 2, according to guidance strategy design results, simulation analysis was conducted for two modes of large and small angles under the limited angle constraint. In the case that the twodimensional turntable rotates along two directions, angles, angular velocities, and angular accelerations were computed. Corresponding simulation results have been shown in Figure 4.
If , the azimuth angular velocity up to /s and above is very large. Consequently, it fails to meet low disturbance requirement of microsatellites. The angular velocities should be subjected to smoothing to reduce solar tracking accuracy and further realize the purpose of low disturbance.
4.2. Tracking Situations Based on FeatureBased Time Series
In terms of the guidance strategy design based on featurebased time series, time of an orbit period was divided into seven segments according to seven characteristic time points under the circumstance that switching point of two modes was . In this manner, large disturbance torque that may be incurred targeted at the satellite platform attitude control could be avoided. Subsequently, twodimensional solar tracking simulation was conducted and the relevant results have been given in Figure 5.
Figure 5 signifies that the angular deviation between two dimensionalities and sunoriented direction are both small in consistency with a guidance strategy based on the featurebased time series, guaranteeing the acquisition of satellite energy. Concerning details of the integrated deviation, please refer to Section 4. As for the angular velocity, the azimuth angular velocity that originally fails to meet relevant requirements have been controlled below /s and values of angular velocities within the entire orbit are those of several speeds matched with time series. This is beneficial for lowering bearing losses of the turntable. In terms of angular acceleration reflecting disturbance torque, solar tracking curve angular acceleration without optimization changes frequently. Consequently, not only is longplaying disturbance generated, but reciprocating disturbance gives rise to waste of satellite control resources. In comparison, disturbance can be rather concentrated after the optimization of such an acceleration, which makes the disturbance to satellite minimized.
Additionally, matching between angles at both ends of each sequence can ensure that angle errors of the entire orbit cannot accumulate. Meanwhile, error precision of the angle can be improved along with the increase in density of the time series.
4.3. Error Analysis
Based on joint debugging of MATLAB/Simulink and STK, joint simulation was conducted for guidance law design. While a guidance law calculation generation model, a turntable control model, and a turntable output model were all constructed by Simulink, STK played a role in analog input and solar tracking situation visualization when utilized to output realtime orbit and onboard time. Through calculations of realtime data, realtime tracking conditions were observed to obtain angle errors of solar tracking subjected to a guidance strategy, as shown in Figures 6 and 7.
A onedimensional solar tracking pattern was utilized under the circumstance of small angle. The solar tracking error of panels at the stable tracking stage is one angle. In addition, errors still can be found before or after the light region due to constraint of the limited angle. The maximum value of them can be expressed in .
A twodimensional solar tracking pattern was utilized under the circumstance of large angle. Errors are incurred in the actual velocity and averaged velocity between tracking points of diverse segments. Moreover, such errors arrive at their maximum values as the azimuth angle crosses zero. In the case that angle is equal to , the corresponding error can be up to . Due to constraints of the limited angle, errors still exist before/after the light region. The maximum value of them can be expressed in .
Solar panels fail to reach their target angle physically because of limited angle constraints, which produces the maximum value of angle error denoted as . Besides, the composition error subjected to a pattern based guidance strategy should be controlled below .
5. Physical System Verification
In order to verify the validity of the proposed method, physical system verification was carried on, which was composed of the onboard computer, power supply, and twodimensional turntable. The sun vector was taken by recursive calculation, and the guidance data was calculated by the onboard computer according to former method. The rotation of the turntable was controlled to achieve the sunoriented mode. The turntable was designed for a certain type of microsatellite, and its physical picture has been shown in Figure 8.
The main performance of the turntable has been shown in Table 1.

Based on the microsatellite orbital characteristics, two modes of large and small angles were selected to verify the guidance method, respectively, in oneyear period, the large angle was , and the small angle was . Through the simulation analysis of the two angles, the results have been shown in Figure 9.
Under the circumstance of large and small angles, there was a small difference between the turntable output attitude and the guidance attitude shown in Figure 4. It meets the requirements of ensuring the low disturbance to the platform, and the maximum angular velocity and the maximum angular acceleration of the turntable should be no more than and , respectively. To verify the sunoriented deviation, the comparison of the turntable output attitude and the sunoriented attitude has been shown in Figure 10.
There was a small deviation between the turntable twodimensional angle and sunoriented direction during execution of guidance strategy, which could guarantee the acquisition of satellite energy. The composition angular deviation of sunoriented direction has been shown in Figure 11.
The temporal interval between 2000 s and 4000 s in Figure 11 was in a shadow region. Besides, the composition angular deviation of sunoriented direction in a sunshiny region should be controlled below , which meets the requirement.
6. Conclusion
A stationary low disturbance solar tracking guidance method applicable to twodimensional solar panels was proposed to realize enough energy acquisition based on the microsatellite platform. Moreover, it supports payload power consumption demands required by multitype tasks performed by the microsatellite, such as spacebased space surveillance, earth observation, navigation enhancement, and loworbit communication, which fills in the gap of solutions to engineering applications of the microsatellite in twodimensional solar panels and provides new ideas for replacement of the traditional SADA with twodimensional turntable. Simulation results demonstrate that such a method has the potential to effectively improve guidance stability of the turntable. Therefore, it is a valid solution to solar tracking stationary guidance problems of microsatellite panels.(1)Segmental and autonomous switching modes were utilized to extract time series within the satellite orbit period according to feature points for the purpose of guidance planning. Moreover, patterns and time series lengths can be autonomously switched in line with the current positional relations with the sun so as to form twodimensional solar tracking guidance laws in the period.(2)The guidance method proposed is not only applicable to diverse nearearth orbits of different orbital inclination, but able to guarantee the solar tracking accuracy of the panels to be below and ensure maximization of energy acquired.
Conflicts of Interest
The authors declare that they have no conflicts of interest.
References
 C.H. Huang, H.Y. Pan, and K.C. Lin, “Development of intelligent fuzzy controller for a twoaxis solar tracking system,” Applied Sciences, vol. 6, no. 5, article no. 130, 2016. View at: Publisher Site  Google Scholar
 M. Taherbaneh, A. H. Rezaie, H. Ghafoorifard, K. Rahimi, and M. B. Menhaj, “Maximizing output power of a solar panel via combination of sun tracking and maximum power point tracking by fuzzy controllers,” International Journal of Photoenergy, vol. 2010, Article ID 312580, 13 pages, 2010. View at: Publisher Site  Google Scholar
 W. Batayneh, A. Owais, and M. Nairoukh, “An intelligent fuzzy based tracking controller for a dualaxis solar PV system,” Automation in Construction, vol. 29, pp. 100–106, 2013. View at: Publisher Site  Google Scholar
 F. Duarte, P. D. Gaspar, and L. C. Gonçalves, “Two axis solar tracker based on solar maps, controlled by a lowpower microcontroller,” Renewable Energy & Power Quality Journal, vol. 1, no. 08, pp. 411–415, 2010. View at: Publisher Site  Google Scholar
 Y. Away and M. Ikhsan, “Dualaxis sun tracker sensor based on tetrahedron geometry,” Automation in Construction, vol. 73, pp. 175–183, 2017. View at: Publisher Site  Google Scholar
 Y. Yao, Y. Hu, S. Gao, G. Yang, and J. Du, “A multipurpose dualaxis solar tracker with two tracking strategies,” Journal of Renewable Energy, vol. 72, pp. 88–98, 2014. View at: Publisher Site  Google Scholar
 J.M. Wang and C.L. Lu, “Design and implementation of a sun tracker with a dualaxis single motor for an optical sensorbased photovoltaic system,” Sensors, vol. 13, no. 3, pp. 3157–3168, 2013. View at: Publisher Site  Google Scholar
 A. Argeseanu, E. Ritchie, and K. Leban, “New low cost structure for dual axis mount solar tracking system using adaptive solar sensor,” in Proceedings of the International Conference on Optimization of Electrical and Electronic Equipment, pp. 1109–1114, May 2010. View at: Publisher Site  Google Scholar
 H. Fathabadi, “Novel high efficient offline sensorless dualaxis solar tracker for using in photovoltaic systems and solar concentrators,” Journal of Renewable Energy, vol. 95, pp. 485–494, 2016. View at: Publisher Site  Google Scholar
 Y. M. Roshan and M. Moallem, “Maximum power point estimation and tracking using power converter input resistance control,” Solar Energy, vol. 96, pp. 177–186, 2013. View at: Publisher Site  Google Scholar
 H.Q. Li, X.F. Liu, L.C. Duan, and G.P. Cai, “Deployment and control of spacecraft solar array considering joint stickslip friction,” Aerospace Science and Technology, vol. 42, pp. 342–352, 2015. View at: Publisher Site  Google Scholar
 E. A. Mizrah, S. B. Tkachev, and N. V. Shtabel, “Research of control system stability in solar array simulator with continuous power amplifier of parallel type,” in Proceedings of the Iop Conference, vol. 94, 2015. View at: Publisher Site  Google Scholar
 L. U. Dongning and Y. Liu, “On the control of flexible solar array drive systems,” Aerospace Control and Application, vol. 39, no. 1, pp. 27–33, 2013. View at: Google Scholar
 L.X. Zhang, Study on dynamics and control of complex remote sensing satellite, Ph.D. dissertation, Harbin Institute of Technology, 2014.
 J. P. Chretien, C. Reboulet, P. Rodrigo, and M. Maurette, “Attitude control of a satellite with a rotating solar array,” Journal of Guidance Control Dynamics, vol. 5, no. 6, pp. 589–596, 2015. View at: Google Scholar
 D. J. Michael and T. H. Robert, “Input shaping to reduce solar array structural vibrations,” Federal Register, vol. 75, no. 7, pp. 989990, 1998. View at: Google Scholar
 Z. Si, Y. Liu, and K. Li, “Research on modeling and driver design of solar array drive assembly,” Aerospace Control Application, vol. 36, no. 2, pp. 13–19, 2010. View at: Google Scholar
Copyright
Copyright © 2018 Keke Zhang et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.