Reinforcement learning and adaptive optimization of a class of Markov jump systems with completely unknown dynamic information

Abstract

providing relevant details, so we can investigate your claim. Download date:03. Nov. 2020 Abstract-In this paper, an online adaptive optimal control problem of a class of continuous-time Markov jump linear systems (MJLSs) is investigated by using a parallel reinforcement learning (RL) algorithm with completely unknown dynamics. Before collecting and learning the subsystems information of states and inputs, the exploration noise is firstly added to describe the actual control input. Then, a novel parallel RL algorithm is used to parallelly compute the corresponding N coupled algebraic Riccati equations (AREs) by online learning. By this algorithm, we will not need to know the dynamic information of the MJLSs. The convergence of the proposed algorithm is also proved. Finally, the effectiveness and applicability of this novel algorithm is illustrated by two simulation examples. Index Terms-Markov jump linear systems (MJLSs); adaptive optimal control; online; reinforcement learning (RL); coupled algebraic Riccati equations (AREs).

Distributed Event-Triggered Adaptive Formation Tracking of Networked Uncertain Stratospheric Airships Using Neural Networks

“…, A i,1 > 0 and A i,2 > 0 are design parameters, (12) and (13) indicate that when the triggering condition E i ≥ A i,1 z i,2 + A i,2 is satisfied, the input τ i updates its value at t i,a+1 and then is set toτ i (t i,a ) during t ∈ [t i,a , t i,a+1 ). Here, the signalτ i for the event-triggered update of the control input is defined as…”

“…Remark 3: Compared with the existing works [8]- [10], [14], [15] for single stratospheric airships, the proposed control scheme (13)-(16) deals with the distributed formation control problem for networked multiple stratospheric airships and the local controller (12) updates its value according to the proposed asynchronous triggering rule (13). Although the local controllers (12) are intermittently updated and the leader output signal is only available for some of the followers, the stability of the resulting formation control system can be ensured, as presented in the following section.…”

“…Theorem 1: Consider the networked multiple stratospheric airships (4) with unknown nonlinearities under the directed graph. For all initial conditions that satisfy V (0) ≤ , the proposed event-triggered adaptive formation tracker (i.e., (12)-(16)) with event-triggering laws (13) ensures that (i) all the closed-loop signals are semi-globally uniformly ultimately bounded; (ii) lim t→∞ − (1 N ⊗ 0 ) −ō ≤ε; (iii) there exist minimum inter-event times…”

“…For airship models subject to unknown system parameters, a nonlinear trajectory tracking control approach was studied in [10] where the unknown parameters were compensated by an adaptive mechanism. Adaptive approximation techniques [11]- [13] have been employed to estimate unknown nonlinearities of airships in adaptive backstepping and sliding mode control frameworks [14], [15]. Despite these efforts, the existing tracking control strategies [8]- [10], [14], [15] for nonlinear airship models have the following restrictions.…”

The Small-Signal Stability of Offshore Wind Power Transmission Inspired by Particle Swarm Optimization

“…A lowpass filter was introduced to improve the system damping, but the effect of different grid-connected positions and capacity on the small signal of the system was not considered. At present, neural networks [16,17], deep learning, and machine learning methods have been widely used in various fields, and remarkable results have been achieved [18,19].…”