Yanqi Zhang

Robotic operations in constrained workspaces are highly susceptible to collisions under unknown adversarial disturbances, primarily because existing cost functions fail to account for dynamic uncertainties. To ensure safety, the framework integrates predefined state and torque bounds. The framework then employs a minimax game between control inputs and adversarial disturbances to derive optimal real-time policies. Specifically, a nonlinear transformation function maps constrained states onto an unconstrained manifold. At the same time, torque constraints are embedded directly within the cost functional. Through remote imitation of expert behavior, the learner autonomously refines its policy-evaluation mechanism, thereby compensating for cost-function biases in adversarial environments. Moreover, a Dual Event-Triggered (DET) mechanism is synthesized to identify high-value expert demonstrations, effectively minimizing computational and communication overhead. Extensive simulation results validate the proposed approach, demonstrating superior convergence precision, resource efficiency, and robustness across diverse adversarial scenarios.

In this work, a novel full state constraint scheme for a class of high-order discrete-time nonlinear systems subjected to unknown parameters is investigated. The new prescribed performance functions are used to ensure that all states of systems can be constrained in a predefined region. Compared with the existing state constraint control and prescribed performance function schemes, the new prescribed performance functions have the advantage of not requiring the feasibility conditions that the virtual controllers are assumed to be bounded. Unlike the conventional switch multiple model set, the proposed multiple model method not only does not rely on a large number of identification models, but also it avoids the potential chattering problem from the switch multiple model method. It is demonstrated that the presented control method can ensure the global boundedness of all the closed-loop signals and that all states converge to the predefined regions. The demonstrated control scheme is implemented on one simulation example for illustrating the effectiveness of the reported scheme.

Summary: In this work, an adaptive multiple model pre‐assigned exponential convergence control scheme based on the time varying exponential function is demonstrated for a class of discrete‐time nonlinear systems. For coping with the noncausality problem from previous adaptive backstepping design method of nonlinear discrete‐time systems, a first‐order recursive filter is used in controller design procedures and the n‐step prediction model is not needed, thereby reducing the computational burden effectively. The time varying function is employed to force all states variables to convergence to zero with a pre‐defined rate which does not rely on the initial conditions of systems. In addition, the multiple model set is constructed to address unknown parameters problem and further the second level adaptive algorithm of unknown parameters is developed. A characteristic feature of the second‐level multi‐model adaptation algorithm is that it dynamically adapts to time changes in the identified model (it uses identification information better than the switching multiple model control). In a unified framework of second level adaption, a first‐order recursive filter and backstepping technique, a pre‐assigned exponential convergence controller is presented. Finally, the results about discrete‐time pre‐assigned exponential convergence method are compared with other methods using two case studies.

The finite-time control problem of the generalized discrete-time systems is a major challenging problem to be solved. In this work, the discrete-time adaptive fuzzy finite-time control scheme is presented for nonlinear systems with discrete form and uncertain dynamics. The fuzzy logic systems (FLSs) are employed to approach uncertain dynamics of systems. Then, the sufficient condition of semiglobal finite-time stability of discrete-time nonlinear systems (DTNSs) is constructed. The discrete-time filters are used to overcome the noncausal problem from the traditional backstepping controller design processes of DTNSs. With the help of the discrete finite-time theorem, discrete-time filters, backstepping techniques, and FLSs, an adaptive fuzzy discrete finite-time control scheme is proposed in a unified framework. The stability analysis can ensure the boundedness of all signals of the closed-loop systems and the achieving the desired tracking control performance in the finite-time. Finally, two examples are adopted to verify abovementioned claims.

This paper investigates the problem of the multiple model control of nonlinear full state constrained systems with a novel barrier Lyapunov function. To handle the problem of unknown parameters, the identification model set containing identification models is established. The novel barrier Lyapunov functions (BLFs) are employed to guarantee that all system states do not violate their boundaries and can remove the restricted condition associated with traditional BLFs. By using the first order filter technique, the "explosion of complexity" difficulty caused by the repeated differential from the backstepping technique is removed. In addition, the estimation of unknown parameters is obtained by using the convex hull property. Based on the second level adaptation technique, the convex hull parameters are designed by using the identification error dynamics, and this algorithm has better convergence property than other algorithms. Then, a multiple model control strategy is developed, which not only guarantees that the requirement of all state constraints is realized, but also ensures that the boundedness of all signals of closed‐loop systems are not violated. The key advantage is that the proposed strategy does not require enormous identification models, that is, the number of identification model is reduced significantly. Finally, the effectiveness of the presented scheme is illustrated through the simulation results.

In this study, a multiple model control scheme based on a second-level adaption design approach is established for a class of nonlinear discrete time systems. The nonlinear function of systems is relaxed to zero-order bounded. In the first-level adaption, N identification models are employed to estimate unknown parameters, which can weaken the effect of unknown parameters. The convex hull property based on all identification models is constructed, which ensures that the convex hull identified parameters converge to the real value rapidly. Compared with the existing multiple model controller, the actual controller is developed by a convex hull for the identification of unknown parameters and it can avoid some potential drawbacks of the switch mechanism. In the constructed scheme, the discrete time system is bounded input bounded output (BIBO) and the tracking error is bounded. Finally, a simulation example is presented to validate the proposed scheme.

The finite-time control design problem of half-vehicle active suspension systems with uncertain dynamics and external disturbances is investigated in this work. The unknown functions, which caused by uncertain parameters and unknown dynamics, are approximated with help of neural networks (NNs). An extended Lyapunov condition of finite-time stability is employed to achieve the control of the vertical and pitch motions more quickly. Then, assisted by the practical finite-time theory, the finite-time controller is proposed. It can ensure that half-vehicle active suspension systems achieve the stability in a finite time and the ride comfort can be enhanced. In addition, the developed adaptive finite-time control approach is performed to half-vehicle active suspension systems. By comparing analysis of simulation results, the validity of the established scheme is demonstrated and the performance of half-vehicle active suspension systems is exhibited.