A Predictive Deep Learning Approach to Output Regulation: The Case of Collaborative Pursuit Evasion

Abstract

In this paper, we consider the problem of controlling an underactuated system in unknown, and potentially adversarial environments. The emphasis will be on autonomous aerial vehicles, modelled by Dubins dynamics. The proposed control law is based on a variable integrator via online prediction for target tracking. To showcase the efficacy of our method, we analyze a pursuit evasion game between multiple autonomous agents. To obviate the need for perfect knowledge of the evader's future strategy, we use a deep neural network that is trained to approximate the behavior of the evader based on measurements gathered online during the pursuit. 1 S. Shivam, and Y. Wardi are with the

Tracking control by the Newton–Raphson method with output prediction and controller speedup

“…Our technique is not concerned with the computation of the reference trajectory but only with its tracking, for which it uses a Newton–Raphson flow, not optimal control. If the reference trajectory has to be computed in real time then various numerical techniques can be used including interpolation as in Section 5, below, or learning methods based on neural nets as in References 4,19,20.…”

“…Preliminary results concerning the material in this paper can be found in the following four conference papers 4,15‐17 . This paper extends them in the following ways: (1) It presents a new version of the controller which can yield perfect asymptotic tracking, in contrast with the above‐mentioned references where only approximate tracking is obtained.…”

“…The system-diagram that we consider is depicted in Figure 1, where the reference signal r(t), control input u(t), and system output y(t) are all in R m for a given m ∈ {1, 2, … , }. The condition that the reference, control, and output have the same dimension is essential for the present discussion, and although ad-hoc ways to circumvent the effects of its absence have begun to emerge, 4 we defer their general exposition to a future publication.…”

“…4,[15][16][17] This paper extends them in the following ways: (1) It presents a new version of the controller which can yield perfect asymptotic tracking, in contrast with the above-mentioned references where only approximate tracking is obtained. (2) Its main results, concerning asymptotic convergence and error analysis in a general setting, and contained in Sections 3 and 4, are new.…”

The Effect of Using Social Engineering for Cybersecurity on the Internet of Things Environment

“…The study by [14] touched on the importance of data privacy, and that data security is essential. All technology companies prefer their data to everything.…”