Optimization, Systems & Control I

C2L-B: Optimization, Systems & Control I

Session Type: Lecture
Session Code: C2L-B
Location: Room 2
Date & Time: Friday March 24, 2023 (10:00-11:00)
Chair: Holden Lee
Track: 4
Paper IDPaper TitleAuthorsAbstract
3015Zeroing the Output of Nonlinear Systems Without Relative DegreeW. Steven Gray{3}, Kurusch Ebrahimi-Fard{2}, Alexander Schmeding{1}The goal of this paper is to establish some facts concerning the problem of zeroing the output of a system that does not have relative degree. The approach taken is to work purely in the input-output setting using Chen-Fliess series representations. It is first shown that a class of generating series called primely nullable series in some sense provide building blocks for this problem using the shuffle algebra. Next, the focus turns to factoring generating series in the shuffle algebra into its irreducible elements for the purpose of identifying nullable factors. This is achieved by viewing the shuffle algebra as the symmetric algebra over the commutative polynomials in Lyndon words. A specific algorithm based on the Chen-Fox-Lyndon factorization of words is given.
3183Performance Analysis of Model-Free Control and PID Control on a Class of Nonlinear MIMO SystemsJames Cheng Peng{2}, James C. Spall{1}Modern dynamical systems are generally non- linear, which make traditional linear controllers infeasible in some cases. To cope with the need for controlling such modern systems, researchers have developed robust non- linear controllers that leveraged function approximators such as neural networks and polynomials. In this work, we investigate one successful nonlinear controller, the SPSA- based model-free control. We shall make a rigorous performance analysis on a class of nonlinear MIMO systems. As a benchmark, we compare it with PID, one of the most- widely used control technique.
3186Policy Poisoning in Batch Learning for Linear Quadratic Control Systems via State ManipulationCourtney King, Son Tung Do, Juntao ChenIn this work, we study policy poisoning through state manipulation, also known as sensor spoofing, and focus specifically on the case of an agent forming a control policy through batch learning in a linear-quadratic (LQ) system. In this scenario, an attacker aims to trick the learner into implementing a targeted malicious policy by manipulating the batch data before the agent begins its learning process. An attack model is crafted to carry out the poisoning strategically, with the goal of modifying the batch data as little as possible to avoid detection by the learner. We establish an optimization framework to guide the design of such policy poisoning attacks. The presence of bi-linear constraints in the optimization problem requires the design of a computationally efficient algorithm to obtain a solution. Therefore, we develop an iterative scheme based on the Alternating Direction Method of Multipliers (ADMM) which is able to return solutions that are approximately optimal. Several case studies are used to demonstrate the effectiveness of the algorithm in carrying out the sensor-based attack on the batch- learning agent in LQ control systems.