DAI Lab@RUC

[20230330] The paper titled "Learning ability analysis for linear discrete delay systems with iteration-varying trial length" has been accepted for publication in top journal Chaos, Solitons & Fractals. This work was led by Prof. JinRong Wang.
[20230314] The paper titled "Distributed Learning Control for High-Speed Trains With Operation Safety Constraints" has been accepted for publication in IEEE Transactions on Cybernetics. This work was collaborated with Mr. Gao and Prof. Song.
[20230223] The paper titled "Enhancing Iterative Learning Control With Fractional Power Update Law" has been accepted for publication in IEEE/CAA Journal of Automatica Sinica. This work was collaborated with Mr. Li and Prof. Yu.
[20230112] The Annual Report of DAI Lab has been released. Please click here for the PDF file. NEW
[20230109] The manuscript titled "Adaptive Finite-Time Fuzzy Control for Hybrid Levitation System of Maglev Trains With Active Anti-lock Constraints" has been accepted for publication in Journal of the Franklin Institute. This work was led by Dr. Zhang.
[20230101] The manuscript titled "Fixed-time Adaptive Consensus for Stochastic Multi-Agent Systems with Uncertain Actuator Faults" has been accepted for publication in ISA Transactions. This work was led by Prof. Wang.
[20230101] We, DAI Lab, wish you Happy 2023.

Our recent focus is learning control with fading channels. A fading channel indicates the unreliable communication network that the transmitted signal would suffer multiplicative randomness. Such randomness is generally modeled by a random variable with its expectation away from 1; thus, the received signal is usually biased in the sense that its expectation is not the original signal. Consequently, a correction is necessary for the received signal before the following procedures. The primary issue in this direction is how to correct the received signals.

If the statistics information (especially the mean) of the fading channel is known, a direct correction can be made by multiplying the mean inverse to the received signals. The design and analysis are given in [TSMC: Syst-2021]. We notice that the corrected input is involved with large disturbances, which may significantly affect the dynamics of the plant. To address this issue, we proposed an iteration domain moving averaging operator for the received inputs. The results are elaborated in [TNNLS-2020]. Motivated by the averaging idea, we proceed to investigate how the averaging techniques affect the learning and tracking abilities of a conventional learning scheme, where the learning ability is reflected by the convergence rate and tracking ability is reflected by the final tracking precision. To this end, we studied three specific averaging techniques, namely, moving averaging, general average with all historical information, and forgetting-based average. The results demonstrate that the forgetting-based average operator-based scheme can connect the other two schemes by tuning the forgetting factor. The results are provided in [TAC-2021]. In addition, we have successfully extended the learning control scheme to a distributed version for multi-agent systems, where each agent is controlled by the input signal generated using the faded neighborhood information (FDI). The iterationwise asymptotic consensus tracking is strictly established for both linear and affine nonlinear multi-agent systems. The results are provided in [TNNLS-2021c]

If the statistics information is unknown, a promising approach is to estimate the mean of the fading channel. This idea is conducted in [TNNLS-2021a], where an iterative estimation mechanism is proposed using a unit pilot signal in each iteration. This mechanism provides necessary statistical information such that the biased signals after transmission can be corrected before being utilized. All the above advances assume that the system information is available for the control design. If both system information and fading statistics are unknown, a natural idea is to estimate them simulatenously. To this end, we propose an error transmission mode and an iterative gradient estimation method. Using the faded tracking error data only, the gradient for updating input is iteratively estimated by a random difference technique along the iteration axis. This gradient acts as the updating term of the control signal. The results are summarized in [TNNLS-2021b].

Furthermore, we consider the general case that the fading channels' statistics, i.e., mean and covariance, are nonrepetitive in the iteration domain. This nonrepetitive randomness introduces non-stationary contamination and drifts to the actual signals, yielding essential challenges in signal processing and learning control. we propose a practical framework constituted by an unbiased estimator of the mean inverse, a signal correction mechanism, and learning control schemes. The convergence and tracking performance are established for both constant and decreasing step-lengths. If the statistics satisfy asymptotic repetitiveness in the iteration domain, a consistent estimator applies to the framework while retaining the framework's asymptotic properties. These results are demonstrated in [TAC-2022].

All the above publications are regular papers in IEEE Transactions. For a complete list, please refer to the LazyPack under the Topic: ILC with fading channels.

Dong SHEN (M'10-SM'17) received the B.S. degree in mathematics from Shandong University, Jinan, China, in 2005. He received the Ph.D. degree in mathematics from the Academy of Mathematics and Systems Science, Chinese Academy of Sciences (CAS), Beijing, China, in 2010 (supervised by Prof. Han-Fu Chen, IEEE/IFAC Fellow).

From 2010 to 2012, he was a Post-Doctoral Fellow with the Institute of Automation, CAS (advised by Prof. Fei-Yue Wang, IEEE/IFAC Fellow). From 2016 to 2017, he was a visiting scholar at National University of Singapore (with Prof. Jian-Xin Xu, IEEE Fellow). From 2019 July to August, he was a visiting scholar at RMIT University (with Prof. Xinghuo Yu, IEEE Fellow). From 2012 to 2019, he has been with College of Information Science and Technology, Beijing University of Chemical Technology, Beijing, China. Now, he is a Full Professor with the School of Mathematics, Renmin University of China, Beijing, China.

His current research interests include iterative learning control, stochastic control and optimization, machine learning and its applications. He has published more than 160 refereed journal and conference papers. He is the (co-)author of Iterative Learning Control for Systems with Iteration-Varying Trial Lengths (Springer, 2019), Iterative Learning Control with Passive Incomplete Information (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 Henan Provincial Natural Science Award in 2022, Shandong Provincial Natural Science Award in 2021, IEEE CSS Beijing Chapter Young Author Prize in 2014, and Wentsun Wu Artificial Intelligence Science and Technology Progress Award in 2012. He is on the Editorial Board of International Journal of Robust and Nonlinear Control, IEEE Access, and IET Cyber-Systems and Robotics.