Subsequent projects

Prof. Dr. Majid Zamani
Technische Universität München
Professur für Hybride Regelungssysteme

Prof. Dr. Murat Arcak
University of California, Berkeley
EECS - Electrical Engineering and Computer Science

Certifiable Synthesis of Autonomous Systems

This project aims at developing a scalable design methodology, in which the embedded controllers are synthesized for stochastic cyber-physical systems (SCPS) from high-level requirements in a push-button and formal manner. Safety-critical and operation-critical systems such as autonomous transportation and power networks are examples of SCPSs. In SCPSs, spatially distributed physical systems operate in uncertain or noisy environments and interact tightly with distributed computational components through embedded controllers. In order to reduce the design costs of embedded control software and guarantee its correctness, we propose a divide and conquer strategy to scale automated synthesis for SCPSs by combining compositional techniques from computer science (e.g. assume-guarantee rules) with those from control theory (e.g. small-gain and dissipativity type theorems).


Primary project: Automated Synthesis of Cyber-Physical Systems: A Passivity-Based Approach

Final Report

Collaboratively led by Prof. Majid Zamani, Prof. Murat Arcak, Mr. Bingzhuo Zhong (a PhD student from TU Munich), and Dr. Paul Griffioen (a postdoc from UC Berkeley), this project aimed to innovate hierarchical control methodologies and develop techniques for computing probabilistic controlled invariant sets for nonlinear systems. The culmination of this project resulted in two significant outcomes, addressing critical challenges in control theory and application.

Outcome 1: Enhanced Hierarchical Control Architectures

This outcome focused on refining hierarchical control methodologies to mitigate errors between abstract and concrete systems. Traditional methods often encounter significant discrepancies, limiting the effectiveness of hierarchical control architectures. To overcome this limitation, a novel simulation relation termed “general approximate alternative simulation relation” was introduced. This relation incorporates input sets from abstract systems, effectively capturing system similarities and reducing error rates. Methods for establishing this relation were provided for discrete-time control systems with finite state and input sets, as well as linear control systems and their continuous abstractions. The practicality and effectiveness of this approach were demonstrated through illustrative examples, showcasing its potential for real-world applications.

Outcome 2: Computing Probabilistic Controlled Invariant Sets

The second outcome focused on computing probabilistic controlled invariant sets for unknown nonlinear systems using data-driven techniques. Specifically, Gaussian process state space models were employed to model nonlinear systems, accounting for unmodeled and unknown dynamics. A semi-definite-programming-based optimization scheme was proposed to compute probabilistic controlled invariant sets, accompanied by state-feedback controllers to maximize system stability subject to input constraints. Additionally, the relationship between probabilistic invariance and finite-time horizon safety was explored, leading to the design of safety controllers to ensure system safety within finite-time horizons. The efficacy of these techniques was validated through simulations and physical experiments on a quadrotor platform.


Privacy Notice