Workshop #2

Prev Next

Real-Time Optimization Based Autonomous Control

Organizers: Behçet Açıkmeşe, Eric Feron, Alexander Domahidi, and John Hauser

Abstract: Optimization algorithms, when they are used in a real-time and safety-critical context, offer the potential for considerably advancing robotic and autonomous systems by improving their ability to initiate, plan, and execute complex missions. To meet distinctly challenging performance and reliability requirements, such automated systems must optimally utilize their full performance envelopes in real-time, while simultaneously satisfying critical mission and environmental constraints. Hence real-time optimization-based control would be a game changing capability, providing solutions to the challenging optimization problems that are ubiquitous in a wide range of applications. Optimization-based problem formulations explicitly consider the problem constraints, objectives, and level of uncertainty involved, so that we can accurately quantify and utilize the system’s performance envelope. These formulations are not only necessary for mathematical completeness and rigor, but also for the accurate quantification of the autonomous systems capabilities. Though optimization provides a powerful formulation framework, there are important challenges in the following areas that must be studied to fully realize this potential: accurate formulations of the control problems as tractable optimization problems, robust real-time implementable numerical optimization methods, and systematic verification for real-time optimization-based control methods and software. To this end, this workshop aims to introduce (i) example applications where real-time optimization can provide significant performance gains or new capabilities; (ii) recent advances in robust real-time optimization algorithms; (iii) methods of verification and certification for real-time algorithms; and (iv) recent demonstrations of real-time optimization based control.

Workshop goals: The main goal of the workshop is to introduce the role of real-time optimization in modern autonomous control applications via real-world examples. We aim to discuss the theoretical, algorithmic, and engineering challenges in realizing real-time optimization based control. We will also present some important recent advances made in realizing and demonstrating real-time optimization in actual engineering applications.

Workshop content and schedule:
  • Introduction and overview (20 mins).
  • Receding horizon control of nonlinear systems (30 mins).
  • Convexification and real-time convex optimization for control (30 mins).
  • Iterative optimization algorithms (30 mins).
  • Autocoding control and optimization systems with code-level guarantees (30 mins).
  • Applications and Demonstrations (20 mins).

Intended audience: This workshop is intended for a broad range of researchers and engineers, from industry to academia. The example applications will be of strong interest to the audience not only because they serve as a proof of concept, but also due to the experience obtained in the transition of this technology to practice. The theoretical results that convert control problems into tractable convex optimization problems (i.e., convexification), the theoretical basis of receding horizon optimal control, IPM algorithms, and the theory behind the software certification methods will all be of interest to researchers from both a theoretical and practical point of view.

Prerequesites: This workshop requires a basic knowledge of optimization and control theory. Most of the discussions are selfcontained, and they are motivated with application examples.