Processing math: 100%
Zhi-Ren Tsai, Yau-Zen Chang. Model-Based Adaptive Predictive Control with Visual Servo of a Rotary Crane System[J]. Journal of Electronic Science and Technology, 2018, 16(2): 169-174. DOI: 10.11989/JEST.1674-862X.7092809
Citation: Zhi-Ren Tsai, Yau-Zen Chang. Model-Based Adaptive Predictive Control with Visual Servo of a Rotary Crane System[J]. Journal of Electronic Science and Technology, 2018, 16(2): 169-174. DOI: 10.11989/JEST.1674-862X.7092809

Model-Based Adaptive Predictive Control with Visual Servo of a Rotary Crane System

doi: 10.11989/JEST.1674-862X.7092809
Funds: This work was supported by MOST for the support under Grants No. MOST 104-2632-B-468-001, No. MOST 103-2221-E-468-009-MY2, No. MOST 104-2221-E-182-008-MY2, No. MOST 105-2221-E-468 -009, No. MOST 106-2221-E-468-023, No. MOST 106-2221-E-182-033, and Chang Gung Memorial Hospital under Grant No. CMRPD2C0053
More Information
  • Author Bio:

    Zhi-Ren Tsai received the Ph.D. degree in electrical engineering from Chang Gung University, Taoyuan in 2001. He is currently an associate professor with the Department of Computer Science and Information Engineering, Asia University, Taichung. His research interests include computer vision, adaptive control, artificial intelligence, robotics application, Internet of things, big data, embedded system, and power electronics

    Yau-Zen Chang received his Ph.D. degree in robotics from University of Oxford, Oxford in 1992. Since August 1993, he has been with Chang Gung University, and is currently a professor. His research interests include artificial neural network, machine vision, and intelligent control. He is also an IEEE senior member

  • Authors’ information: Z. R. Tsai is with the Department of Computer Science and Information Engineering, Asia University, Taichung 41354 and also with the Department of Medical Research, China Medical University Hospital, China Medical University, Taichung 41354 Y.-Z. Chang (corresponding author) is with the Department of Mechanical Engineering, Chang Gung University, Taoyuan 33302 and also with the Department of Neurosurgery, Chang Gung Memorial Hospital, Chiayi 61363 (e-mail: zen@mail.cgu.edu.tw)
  • Received Date: 2017-02-07
  • Rev Recd Date: 2017-08-03
  • Available Online: 2019-12-25
  • Publish Date: 2018-07-04
  • This paper investigated the implementation of an adaptive predictive controller using nonlinear dynamic echo state neural (ESN) model for a rotary crane system by the visual servo method. The control sequences within the control horizon were described using cubic spline interpolation to enlarge the predictive horizon. Verification of the proposed scheme in the face of exogenous disturbances and modeling error with inaccurate string length was demonstrated by both simulations and experiments.
  • Many papers in the field of stereo vision[1]-[6] have been published in the past decade. The progress has been made in improving accuracy as well as achieving real-time performance by using multiple cameras. However, using cameras to compute 3D position coordinate of targets is still challenging due to the unpredictable lighting conditions, unavoidable noise, and interferences of the surrounding objects.

    In this paper, the proposed rotary crane system is equipped with a camera for payload monitoring. This arrangement is important and novel in building construction, factories, and harbors as it avoids the installation of durable encoders at the hanging base of string. To achieve the reliable position estimation, a white light emitting diode (LED) is attached on the payload and a camera is installed on the rotary frame to derive swing angles for control feedback. The model predictive control (MPC)[7],[8] is applied on this system. Importantly, the motors in the system are controlled in velocity mode, which much simplifies the consideration of nonlinear effects caused by friction and load. Furthermore, to compensate for the time-delay effects in visual servo, a robust adaptive predictive controller is proposed by using echo state neural (ESN) network[9],[10].

    The coordinate systems are shown in Fig. 1. The coordinate transformation of the coordinates attached on the rotary crane and the camera as in Fig. 1, is defined as follows:

    Figure  1.  Coordinates of the proposed rotary crane: (a) picture of the crane system showing the coordinate systems and (b) schematic diagram of the coordinate systems.
    TBT=[cosαcosβsinαcosβsinβ0cosαsinβsinγsinαcosγsinαsinβsinγ+cosαcosγcosβsinγ0cosαsinβcosγ+sinαsinγsinαsinβcosγcosαsinγcosβcosγ00001],
    BP=[lsinϕcosθ+plsinϕsinθ+dllcosθh1],CP=[xyz1],CP=CBT(α,β,γ)BP(θ,φ,d,h,l,p).

    For the camera coordinate system, as shown in Fig. 2, there are an image plane π and a camera frame center O. The distance between O and π is the focal length f, and the projective point p is the intersection of the line connecting point P and O with the image plane π. By the properties of similar triangles, the 2D image coordinate is calculated as follows:

    Figure  2.  Coordinate of the camera.
    [pxpy]=[xf/yzf/y].

    In order to find the solutions of α,β,γ,d,h,l,p,andf of this 3D position coordinate model, the downhill simplex method is used with the experimental measurements served as training data, as listed in Table 1. The results are shown in Table 2. Next, considering that

    Table  1.  Relationships between (θ,ϕ) and (x,y) using experimental measurements
    θ (degree)θ (degree)
    203040
    x180–114.300–133.630–164.630
    135–86.970–93.630–115.300
    90–19.300–19.800–24.630
    4566.20070.37086.700
    0115.030128.533166.030
    –4592.370108.370143.030
    –9010.37012.20019.370
    –135–70.133–98.300–113.630
    y18013.07017.07029.233
    13553.90060.57086.070
    9074.57081.400108.070
    4563.73073.07097.070
    029.40032.83354.570
    –45–16.500–15.767–14.330
    –90–38.600–42.930–49.600
    –135–29.517–30.767–32.930
     | Show Table
    DownLoad: CSV
    Table  2.  Optimal device parameters of the rotary crane and the camera
    d (mm)h (mm)l (mm)p (mm)
    558.78351.04302.8225.78
    β (degree)γ (degree)α (degree)f
    –6.1518.43–9.50603.14
     | Show Table
    DownLoad: CSV
    y=lsinφsinθ,x=lsinφcosθ,β=tan1(ylcosφ),α=tan1(xlcosφ).

    Moreover, we notice that the position of the swinging load can be uniquely defined by α and β, as shown in Fig. 3. Hence, we use the Kriging algorithm to find the 3D Kriging surface model for interpolating 3D coordinates, using αandβ as input, for the purpose of minimizing the visual delay, as shown in Fig. 4.

    Figure  3.  Definition of swinging angles.
    Figure  4.  Model of position and swinging angles: (a) and (b) are the model between the position (px,py) and the swinging angles α and β, respectively; (c) and (d) are the 3D Kriging surface fitting model for the position (px,py) and the swinging angles α and β, respectively.

    In this section, a general scheme of Fig. 5 is proposed in the predictive control to extend the control horizon for more than one step in order to compensate for the delayed sensor feedback using a camera[7],[8],[11].

    Figure  5.  Control structure of this study.

    The physical model of the rotary crane system is described as follows:

    l¨θ=1lsinφ[l¨αsinφ2l˙φ˙αcosφ+r˙α2cosθ2˙r˙αsinθr¨αsinθ¨rcosθ2l˙φ˙θcosφ]+b1˙θ (1)
    l¨φ=1l[(gsinφ)+(¨rsinθcosφ)+2l˙θ˙αcosφsinφ+l˙θ2cosφsinφ+l˙α2cosφsinφ+r˙α2sinθcosφ+2˙r˙αcosθcosφ+r¨αcosθcosφ]+b2˙φ (2)

    where b1 and b2 are damping coefficients to be decided. Equations (1) and (2) are in the standard state space form:

    ˙x(t)=fp(x(t),u(t));y(t)=gp(x(t)) (3)

    where the state vector x=[θ,˙θ,φ,˙φ]T, the control input vector u=[˙α,¨α,˙r,¨r]T, and the output vector y=[αg,γg]T, as shown in the experimental platform of Fig. 6.

    Figure  6.  Picture of the experimental prototype setup.

    During the time interval between t and t+TC, the ESN-model-based controller u simulates a physical model described by:

    {˙^x(τ)=ˆf(ˆx(τ),u(τ))ˆy(τ)=ˆg(ˆx(τ)),t+TCτt+TC+TP (4)

    using estimated initial state vector ˆx(t+TC), where t=kTs, Ts is sampling period and k=0,1,2,; TP is the prediction horizon, and yr is the desired output of model output ˆy(τ). Relations between these variables and parameters are shown in Fig. 7.

    Figure  7.  Relations between the reference trajectory, candidate control inputs, and estimated system output during the predictive control.

    As shown in Fig. 7, ˆx(t+TC) is obtained by the simulation of (4) using the control input sequence calculated in the last time interval [t,tTC]. For each trial, taking ui as the i-th trial, different weights ϖ, ψ and ζ are applied in the two orthogonal directions, α and r, in different stages. In doing so, the cost function is defined as:

    Cost=ψFα+ζFγ,
    Fα=t+TC+TPt+TC||αd(τ)ˆαg(τ)||dτ+ϖt+TC+TPt+TC||¨α(τ)||dτ,
    Fγ=t+TC+TPt+TC||γd(τ)ˆγg(τ)||dτ+ϖt+TC+TPt+TC||¨r(τ)||dτ,

    where αd(τ) and γd(τ) are desired trajectories; ˆαg(τ) and ˆγg(τ) are estimated trajectories of αg and γg, respectively .

    In this section, a tapped-delay[12] ESN network[13]-[15] is designed as an adaptive model for cubic spline interpolation[16],[17] to enlarge the predictive horizon[18],[19], where the physical model-based controller[20] u in (4) is replaced by a discrete-time ESN mode composed of 10 neurons each equipped with 9 tapped-delay.

    Furthermore, an adaptation theorem is used for modeling to enhance the accuracy of ESN-based model. First, the input sequence is assumed to be a K×1 matrix, ESN has N neurons and the output is an L×1 matrix, moreover, the state of each neuron is x(k)=tanh(Winu(k)+Wx(k1)+Wback), where the length of Win is an N×K matrix, W is N×N and the Wback is N×L. The final outputs of model is ˆy(k)=Wout[u(k),x(k),ˆy(k1),,ˆy(kn)].

    To complete the neural plant modeling design, a convergence theory for this model is provided in the following theorem:

    Theorem 1. If the sampling time Ts is selected such that ˉy(k)y(k)ˉε and 0<η(k)<2/dˆy(k)dWout(k)2, is satisfied, where

    dˆy(k)dWout(k)=ˆy(k)Wout(k)+ni=1ˆy(k)ˆy(ki)dˆy(ki)dWout(k);

    where ˉy(k) is the optimal output of model, we have that ˆy(k)ˉy(k)y(k).

    Proof of Theorem 1. First, the Lyapunov candidate[21] is designed as follows,

    V1(k)=1/2(ˆy(k)y(k))T(ˆy(k)y(k))=V2(k)+ε(k),

    where V2(k)=1/2ˆy(k)ˉy(k)2, ε(k)=1/2ˉy(k)y(k)2+(ˆy(k)ˉy(k))T(ˉy(k)y(k)).

    This model is then trained such that V2(k) is minimized. The following derivation involves a convergence condition for selecting appropriate learning rates η(k). Firstly we notice that

    e(k+1)=e(k)(1[dˆy(k)dWout(k)]Tη(k)dˆy(k)dWout(k))

    where e(k)=ˆy(k)ˉy(k). Thus

    V2(k+1)V2(k)=12e(k)2[(1[dˆy(k)dWout(k)]Tη(k)dˆy(k)dWout(k))21].

    Hence, if 1<1[dˆy(k)dWout(k)]Tη(k)dˆy(k)dWout(k)<1 and ˉy(k)y(k)ˉε, we have that V2(k+1)<V2(k). This is equivalent to V2(k)0, thereby the proof is completed.

    Input to the ESN model are past positions, velocities, and accelerations of motor and payload. Outputs of the ESN model are now positions of payload. Each trial control sequence is encoded into 3 parameters which define a cubic spline.

    In order to find the best solution on-line, the downhill simplex method is used. Considering that the method suffers from the possibility of being trapped in local minima, we restart each interval with four polytopes of 4 vertices randomly and make it calculate for 20 iterations, and then start a polytope with the 4 best results of these polytopes as vertices and run for 20 more iterations.

    Experimental prototype of Fig. 6 was used for several verification experiments. In the experimental settings, the cable length varies between 0.175 m and 0.400 m, and the payload is assumed to be 2.851 kg hanging from a cable of 920 mm with nominal length being 310 mm, and the control signal is updated with a sampling period of 0.05 s, according to the experimental prototype, and TP=10TC. Longer TC alleviates the demand on computing power of the controller.

    Firstly, the effect of disturbance is studied by assuming an extra velocity of 57.3 deg/s in the θ direction during the period of 5.5 s to 6.5 s. The best control horizon TC=0.3 s without inducing significant vibration is selected through testing the tracking performance of the system for control horizons ranging from 0.1 s to 0.6 s. Fig. 8 shows the responses when disturbance is applied. Finally, the advantage of adaptive predict control is dictated by comparing the results with the conventional predictive control[11] in Fig. 9.

    Figure  8.  Tracking performance of the predictive controller when an extra disturbance is introduced between 5.5 s to 6.5 s: (a) ag and (b) yg.
    Figure  9.  Comparison and weights: (a) and (b) are comparison of swing ranges between the conventional predictive controller and the proposed adaptive predict controller, where (a) shows the movement of the payload in the left and right direction and (b) in the forward and backward direction, both are expressed in meters; (c) weights of ESN model are added up showing the trends of adaptation. The inputs to the ESN model are normalized that the weights have no units.

    The control problems of modeling error and time-varying delay in anti-swing control were handled with the adaptive predictive control strategy. Importantly, a robust LED vision servo was proposed, and a tapped-delay ESN network was designed as an adaptive model to enlarge prediction horizon in the control law. Performances of the proposed scheme were verified by experimental results.

  • [1]
    M. Y. Kim, S. M. Ayaz, J. Park, and Y. Roh, " Adaptive 3D sensing system based on variable magnification using stereo vision and structured light,” Optics and Lasers in Engineering, vol. 55, pp. 113-127, Apr. 2014.
    [2]
    R. Correal, G. Pajares, and J. J. Ruz, " Automatic expert system for 3D terrain reconstruction based on stereo vision and histogram matching,” Expert Systems with Applications, vol. 41, pp. 2043-2051, Mar. 2014.
    [3]
    W. Kazmi, S. Foix, G. Aleny, and H. J. Andersen, " Indoor and outdoor depth imaging of leaves with time-of-flight and stereo vision sensors: Analysis and comparison,” ISPRS Journal of Photogrammetry and Remote Sensing, vol. 88, pp. 128-146, Feb. 2014.
    [4]
    M. Camplani and L. Salgado, " Background foreground segmentation with RGB-D Kinect data: An efficient combination of classifiers,” Journal of Visual Communication and Image Representation, vol. 25, pp. 122-136, Jan. 2014.
    [5]
    G. Du and P. Zhang, " Markerless human-robot interface for dual robot manipulators using Kinect sensor,” Robotics and Computer-Integrated Manufacturing, vol. 30, pp. 150-159, Apr. 2014.
    [6]
    B. Tippetts, D. J. Lee, K. Lillywhite, and J. Archibald, " Review of stereo vision algorithms and their suitability for resource-limited systems,” Journal of Real-Time Image Processing, vol. 11, pp. 5-25, Jan. 2016.
    [7]
    S. Iles, M. Lazar, F. Kolonic, and J. Matusko, " Stabilizing model predictive control of a gantry crane based on flexible set-membership constraints,” IFAC PapersOnLine, vol. 48, no. 23, pp. 248-253, 2015.
    [8]
    D. Jolevski and O. Bego, " Model predictive control of gantry/bridge crane with anti-sway algorithm,” Journal of Mechanical Science and Technology, vol. 29, pp. 827-834, Feb. 2015.
    [9]
    R. Couillet, G. Wainrib, H. Sevi, and H. T. Ali, " The asymptotic performance of linear echo state neural networks,” Journal of Machine Learning Research, vol. 17, pp. 1-35, Mar. 2016.
    [10]
    S. Scardapane, D. Wang, and M. Panella, " A decentralized training algorithm for echo state networks in distributed big data applications,” Neural Networks, vol. 78, pp. 65-74, Jun. 2016.
    [11]
    S. W. Su, H. Nguyen, and R. Jarman, " Model predictive control of gantry crane with input nonlinearity compensation,” in Proc. of Intl. Conf. Control, Automation and Systems Engineering, 2009, pp. 312-316.
    [12]
    A. Vahdat, J. Morgan, A. R. McIntyre, M. I. Heywood, and A. N. Zincir-Heywood, " Tapped delay lines for GP streaming data classification with label budgets,” in Proc. of the 18th European Conf. on Genetic Programming, 2015, pp. 126-138.
    [13]
    R. Bhuvana, S. Purushothaman, and P. Rajeswari, " Methodology of echo state neural network to diagnose human depression,” Indian Journal of Science and Technology, vol. 8, pp. 1-4, Nov. 2015.
    [14]
    S. I. Han and J. M. Lee, " Fuzzy echo state neural networks and funnel dynamic surface control for prescribed performance of a nonlinear dynamic system,” IEEE Trans. on Industrial Electronics, vol. 61, pp. 1099-1112, Feb. 2014.
    [15]
    G. Manjunath and H. Jaeger, " Echo state property linked to an input: Exploring a fundamental characteristic of recurrent neural networks,” Neural Computation, vol. 25, pp. 671-696, Dec. 2013.
    [16]
    Q. Dai, A. K. Katsaggelos, S. Yu, W. Kang, J. Jeon, and J. Paik, " Directionally adaptive cubic-spline interpolation using optimized interpolation kernel and edge orientation for mobile digital zoom system,” in Proc. of the 18th IEEE Intl. Symposium on Consumer Electronics, 2014, pp. 1-2.
    [17]
    S. H. Hong, L. Wang, T. K. Truong, T. C. Lin, and L. J. Wang, " Novel approaches to the parametric cubic-spline interpolation,” IEEE Trans. on Image Processing, vol. 22, pp. 1233-1241, Mar. 2013.
    [18]
    X. Liu and X. Kong, " Nonlinear model predictive control for DFIG-based wind power generation,” IEEE Trans. on Automation Science and Engineering, vol. 11, pp. 1046-1055, May 2014.
    [19]
    B. Zhu, H. Tazvinga, and X. Xia, " Switched model predictive control for energy dispatching of a photovoltaic-diesel-battery hybrid power system,” IEEE Trans. on Control Systems Technology, vol. 23, pp. 1229-1236, May 2015.
    [20]
    M. H. Fatehi, M. Eghtesad, and R. Amjadifard, " A novel dynamic model for a crane system with a flexible cable and largeswing angle,” Intl. Journal of Scientific Engineering and Technology, vol. 4, pp. 370-377, Jun. 2015.
    [21]
    A. Derdiyok, and A. Basci, " Speed estimation of an induction machine based on designed Lyapunov candidate functions,” Electrical Engineering, vol. 98, pp. 67-75, Mar. 2016.
  • Catalog

      Figures(9)  /  Tables(2)

      Article Metrics

      Article views (815) PDF downloads (97) Cited by()
      Related
      Proportional views

      /

      DownLoad:  Full-Size Img  PowerPoint
      Return
      Return