Figures & data
Figure 1. Flowchart of the RBFNN optimal control algorithm with model reduction and transfer learning for linear systems.
![Figure 1. Flowchart of the RBFNN optimal control algorithm with model reduction and transfer learning for linear systems.](/cms/asset/f52bae53-0bc1-47d2-af40-35171edeebac/tcon_a_2328687_f0001_oc.jpg)
Figure 2. Quanser-Servo2 Inverted Pendulum system hardware setup (Quanser, Citation2022).
![Figure 2. Quanser-Servo2 Inverted Pendulum system hardware setup (Quanser, Citation2022).](/cms/asset/80127aff-5ccc-4521-934b-bd9903eaf48c/tcon_a_2328687_f0002_oc.jpg)
Table 1. Parameters of the rotary pendulum system.
Table 2. Summary of reduced order model matrices from model-based gramians and empirical gramians.
Table 3. Hankel singular values of the rotary pendulum.
Figure 3. Top: Relative output errors of the reduced order model by the balanced truncation and empirical balanced truncation. Below: The input signal.
![Figure 3. Top: Relative output errors of the reduced order model by the balanced truncation and empirical balanced truncation. Below: The input signal.](/cms/asset/5b24b451-876c-48d9-b194-2aaf64703d63/tcon_a_2328687_f0003_oc.jpg)
Figure 4. Tracking the square wave of the rotary army in simulations. In the legend, ‘LQR’ denotes the LQR control designed with the original model; ‘RBFNN’ denotes the RBFNN control designed with the model-based BT; ‘Empirical RBFNN’ denotes the RBFNN control designed with the empirical BT.
![Figure 4. Tracking the square wave of the rotary army in simulations. In the legend, ‘LQR’ denotes the LQR control designed with the original model; ‘RBFNN’ denotes the RBFNN control designed with the model-based BT; ‘Empirical RBFNN’ denotes the RBFNN control designed with the empirical BT.](/cms/asset/33bf50e7-2fbb-4df1-867c-d8fd84170826/tcon_a_2328687_f0004_oc.jpg)
Figure 5. Comparisons of tracking response of the rotary army in simulations. In the legend, ‘RBFNN_Empirical’ denotes the RBFNN control designed with the empirical BT; ‘LQR_Empirical BT’ denotes the LQR control designed with the empirical BT; ‘LQR_Model-based BT’ denotes the LQR control designed with the model-based BT.
![Figure 5. Comparisons of tracking response θ(t) of the rotary army in simulations. In the legend, ‘RBFNN_Empirical’ denotes the RBFNN control designed with the empirical BT; ‘LQR_Empirical BT’ denotes the LQR control designed with the empirical BT; ‘LQR_Model-based BT’ denotes the LQR control designed with the model-based BT.](/cms/asset/80cf788b-4528-498a-b3f2-f7f021f72349/tcon_a_2328687_f0005_oc.jpg)
Figure 6. Comparisons of tracking response of the rotary army in simulations. Legends are the same as in Figure .
![Figure 6. Comparisons of tracking response α(t) of the rotary army in simulations. Legends are the same as in Figure 5.](/cms/asset/1832b323-59df-4991-bc75-af2aede4138b/tcon_a_2328687_f0006_oc.jpg)
Figure 7. The closed-loop tracking responses of the rotary arm of Quanser-Servo2. Legends are the same as in Figure .
![Figure 7. The closed-loop tracking responses θ(t) of the rotary arm of Quanser-Servo2. Legends are the same as in Figure 4.](/cms/asset/6705ac65-2779-442f-8ba5-9e7ce099d1c1/tcon_a_2328687_f0007_oc.jpg)
Figure 8. The closed-loop responses of the rotary arm under various controls for balancing the inverted pendulum of Quanser-Servo2. Top: Responses before retraining. Bottom: Responses after retraining. Legends are the same as in Figure .
![Figure 8. The closed-loop responses θ(t) of the rotary arm under various controls for balancing the inverted pendulum of Quanser-Servo2. Top: Responses before retraining. Bottom: Responses after retraining. Legends are the same as in Figure 4.](/cms/asset/29c018a3-f218-402f-b990-b981cfeae43f/tcon_a_2328687_f0008_oc.jpg)
Figure 9. The closed-loop tracking response of the rotary arm of Quanser-Servo2. Legends are the same as in Figure .
![Figure 9. The closed-loop tracking response θ(t) of the rotary arm of Quanser-Servo2. Legends are the same as in Figure 4.](/cms/asset/f9bd72e5-33db-450e-8a68-4f922f8c6c76/tcon_a_2328687_f0009_oc.jpg)
Figure 10. The closed-loop response of the pendulum in the rotary arm tracking control of Quanser-Servo2. Legends are the same as in Figure .
![Figure 10. The closed-loop response α(t) of the pendulum in the rotary arm tracking control of Quanser-Servo2. Legends are the same as in Figure 4.](/cms/asset/c3ec05a8-acd9-456c-bb98-be6f53deeb63/tcon_a_2328687_f0010_oc.jpg)
Table 4. Summary of control performance for LQR, RBFNN, empirical RBFNN, retrained RBFNN and retrained empirical RBFNN.
Figure 11. Robustness comparisons of all the controls under consideration. Top: The closed-loop angle response of the rotary arm in balancing control of Quanser-Servo2. Bottom: Disturbance
. Legends are the same as in Figure .
![Figure 11. Robustness comparisons of all the controls under consideration. Top: The closed-loop angle response θ(t) of the rotary arm in balancing control of Quanser-Servo2. Bottom: Disturbance d(t). Legends are the same as in Figure 4.](/cms/asset/4676cb0e-cba5-49e8-8925-1d3ee8365ec3/tcon_a_2328687_f0011_oc.jpg)
Figure 13. Performance comparison of RBFNN, Poly-NN control and LQR controls for the nonlinear system in Equation (Equation61(61)
(61) ). Top: The control
. Middle: The response
. Bottom: The response
.
![Figure 13. Performance comparison of RBFNN, Poly-NN control and LQR controls for the nonlinear system in Equation (Equation61(61) x˙1=x1+x2−x1(x12+x22)x˙2=−x1+x2−x2(x12+x22)+u(61) ). Top: The control u(t). Middle: The response x1(t). Bottom: The response x2(t).](/cms/asset/35a92da5-9956-49a2-afbe-eeb208cc5a4c/tcon_a_2328687_f0013_oc.jpg)
Figure 14. Comparison of spatial distribution of RBFNN and Poly-NN optimal controls u as a function of the state . Left: The control
plotted in the training region
. Right: The control
plotted beyond the training region into the larger region
.
![Figure 14. Comparison of spatial distribution of RBFNN and Poly-NN optimal controls u as a function of the state x. Left: The control u(x) plotted in the training region Xs1∈[−1,1]×[−1,1]. Right: The control u(x) plotted beyond the training region into the larger region Xs2∈[−2,2]×[−2,2].](/cms/asset/db99f4a6-3c95-4203-9bde-f43012e35e31/tcon_a_2328687_f0014_oc.jpg)
Figure 15. Robustness of the RBFNN and LQR controls with respect to the model uncertainty β. The vertical dash lines mark the critical value of β, beyond which the closed-loop system becomes unstable.
![Figure 15. Robustness of the RBFNN and LQR controls with respect to the model uncertainty β. The vertical dash lines mark the critical value of β, beyond which the closed-loop system becomes unstable.](/cms/asset/c1a32244-2e01-401a-bedd-fb12c137ad98/tcon_a_2328687_f0015_oc.jpg)
Figure A1. Comparison of RBFNNs and LQR control performances for the linear 2D system. Top: Control . Middle: Response
. Bottom: Response
. The initial condition of the system is
.
![Figure A1. Comparison of RBFNNs and LQR control performances for the linear 2D system. Top: Control u(t). Middle: Response x1(t). Bottom: Response x2(t). The initial condition of the system is x(0)=[1,1]T.](/cms/asset/2333d903-4fb1-4ce0-ac59-68b69e0a4db8/tcon_a_2328687_f0016_oc.jpg)