Zhong-Ping Jiang - Selected Publications#
[1] Z. P. Jiang and T. Liu, Small-gain theory for stability and control of dynamical networks: A survey, Annual Reviews in Control, Vol. 46, pp. 58-79, Oct. 2018. DOI: 10.1016/j.arcontrol.2018.09.001
Significance: This paper surveys the seminal contributions made by Prof. Jiang to the nonlinear small-gain theory for interconnected systems, based on input-to-state stability and input-to-output stability dealing with both internal and external stability properties. This theory is now a common tool used within the nonlinear control community, and it has received international recognition. His 1994 paper on the first truly nonlinear small-gain theorem has been cited over 1200 times on Google Scholar and it is "the reference" in the field.
[2] Z. P. Jiang, T. Bian and W. Gao, Learning-based control: A tutorial and some recent results. Foundations and Trends in Systems and Control, Vol. 8, No. 3, pp. 176–284, 2020. DOI: 10.1561/2600000023
Significance: This monograph develops and surveys data-driven control tools and methods for continuous-time linear and nonlinear systems with completely unknown dynamics. An entanglement of techniques from reinforcement learning and model-based control theory is advocated to show that adaptive optimal controllers can be learned directly from real-time input–output data with stability and robustness guarantees. The effectiveness of the proposed learning-based control framework is demonstrated via its applications to theoretical optimal control problems tied to various important classes of continuous-time dynamical systems and practical problems arising from biological motor control, connected and autonomous vehicles.
[3] T. Bian, D. Wolpert and Z. P. Jiang, Model-free robust optimal feedback mechanisms of biological motor control, Neural Computation, Vol. 32, pp. 562-595, Mar. 2020. doi: 10.1162/neco\_a\_01260
Significance: This is the first-ever paper to propose learning-based optimal control as a theory for the computational principle of human movement. It reveals that internal models alone may not be adequate to explain the motor adaptation behavior during the early phase of learning. The proposed model successfully reproduces the experimental phenomena reported in the work of others, and resolves the apparent inconsistency between existing motor control theories and the experimental observation of the positive impact of motor variability.
[4] T. Bian and Z. P. Jiang, Continuous-time robust dynamic programming, SIAM J. Control and Optimization, 57 (6), pp. 4150--4174, Dec. 2019. DOI: 10.1137/18M1214147
Significance: This paper introduces a new theory of robust dynamic programming for continuous-time dynamical systems subject to a wide spectrum of static and dynamic uncertainties. It creatively integrates modern robust nonlinear control techniques into the study of dynamic programming and reinforcement learning algorithms to conquer two fundamental limitations of existing machine learning and optimization methods: robustness and closed-loop performance analysis. This paper proposes the first robust learning result for continuous-time dynamical systems, as opposed to Markov decision processes and discrete-time models often used previously. The new theory has been applied to solve challenging problems arising from finance and engineering that cannot be dealt with by means of traditional dynamic programming methods.
[5] T. Liu and Z. P. Jiang, Event-triggered control of nonlinear systems with state quantization, IEEE Transactions on Automatic Control, Vol. 64, No. 2, pp. 797-803, Feb. 2019. DOI: 10.1109/TAC.2018.2837129.
Significance: This paper presents the first solution to event-triggered control for nonlinear systems with quantized feedback signals. By means of input-to-state stability and nonlinear small-gain techniques, new event-triggering mechanisms have been developed to handle the complex interaction between the quantizer and the sampler and avoid infinitely fast sampling (especially, Zeno phenomenon), while maintaining asymptotic convergence of the states to an equilibrium of interest.
[6] W. Gao, Z. P. Jiang, F. Lewis and Y. Wang, Leader-to-formation stability of multi-agent systems: an adaptive optimal control approach, IEEE Transactions on Automatic Control, vol 63, no. 10, pp. 3581-3587, Oct. 2018. DOI: 10.1109/TAC.2018.2799526
Significance: This paper published in the No. 1 journal in Automatic Control proposes a novel data-driven solution to the cooperative adaptive optimal control problem of leader-follower multi-agent systems under switching network topology and model uncertainty. This is one of the first papers proposing the systematic use of reinforcement learning and approximate/adaptive dynamic programming to learn adaptive/optimal controllers on-line from real-time input-state data. Performance analysis and rigorous stability proofs with robustness to unmeasurable leader disturbance are provided. It goes beyond previous cooperative control algorithms by allowing communication loops among autonomous mobile agents.
[7] W. Gao and Z. P. Jiang, Learning-based adaptive optimal tracking control of strict-feedback nonlinear systems, IEEE Transactions on Neural Networks and Learning Systems, vol. 29, no. 6, pp. 2614-2624, June 2018. DOI: 10.1109/TNNLS.2017.2761718
Significance: This paper studies a fundamentally important problem in control theory that is to design feedback controllers to achieve asymptotic tracking with perfect disturbance rejection, the so-called output regulation problem. This paper has proposed the first learning-based solution to address both adaptivity and optimality properties for the output regulation of uncertain nonlinear systems. Published in the flagship journal of the IEEE Computational Intelligence Society, this paper is cited 65+ times on Google Scholar. Its previous linear counterpart published in the December 2016 issue of IEEE Trans. Automatic Control (the flagship journal of the IEEE Control Systems Society) has been cited 158+ times.
[8] Y. Jiang and Z. P. Jiang, Robust Adaptive Dynamic Programming. Wiley-IEEE Press, 2017.
Significance: This research monograph summarizes the recent developments in robust adaptive dynamic programming that generalize the literature of adaptive dynamic programming to continuous-time systems with unknown system dynamics and unknown system order. The theory relaxes both the curse of dimensionality and the curse of modeling associated with traditional dynamic programming for optimal decision making. The efficacy of the developed framework for learning-based optimal control has been validated by its applications to electric power systems, human motor control and other real-world engineering examples. It is cited 120+ times on Google Scholar.
[9] W. Gao, Z. P. Jiang and K. Ozbay, Data-driven adaptive optimal control of connected vehicles, IEEE Transactions on Intelligent Transportation Systems, Vol. 18, No. 5, pp. 1122-1133, May 2017. DOI: 10.1109/TITS.2016.2597279 9
Significance: This is the first paper that has looked at cooperative driving from a learning-based optimal control perspective. It presents a a data-driven non-model-based control algorithm that minimizes an integral-quadratic cost in terms of the errors of vehicle headway and velocity for a string of mixed traffic human-driven and automated vehicles with vehicle-to-vehicle communication. The result extends the conventional cooperative adaptive cruise control (CACC) system to learning-based CACC system with inherent adaptivity and high-performance. String stability and input-to-state stability with respect to the disturbance of the leading vehicle are guaranteed and validated via the online learning control of connected vehicles in Paramics-based traffic micro-simulation. The paper is published in the No. 1 journal of the IEEE Intelligent Transportation Systems Society and has been cited 80+ times on Google Scholar.
[10] T. Bian and Z. P. Jiang, Value iteration and adaptive dynamic programming for data-driven adaptive optimal control design, Automatica, vol. 71, pp. 348–-360, Sept. 2016. DOI: 10.1016/j.automatica.2016.05.003
Significance: This paper solves a longstanding open problem in dynamic programming, and provides the first continuous-time variant of the popular value iteration method with rigorous convergence analysis for continuous-time unknown dynamical systems described by differential equations. It removes the restrictive assumption on the a priori knowledge of an initial admissible control policy in all policy iteration algorithms previously proposed in the literature, and gives a new tool for data-driven adaptive optimal controller design for continuous-time uncertain systems. Published in a top journal in Automatic Control, this paper has to-date received 90+ citations on Google Scholar.
