Prioritized Constraints in Optimization-Based Control

TL;DR

This paper develops a theoretical framework and an efficient solver for optimization-based control with prioritized constraints, enabling real-time applications like autonomous driving to handle conflicting constraints effectively.

Contribution

It introduces the concept of prioritized intersections, extends existing methods, and provides a tailored dual active-set solver optimized for real-time control scenarios.

Findings

Successfully applied in real-time MPC for autonomous driving

Outperforms existing solvers for hierarchical quadratic programming

Handles six levels of conflicting constraints efficiently

Abstract

We provide theoretical foundations and computational tools for the systematic design of optimization-based control laws with constraints that have different priorities. By introducing the concept of prioritized intersections, we extend and unify previous work on the topic. Moreover, to enable the use of prioritized intersection in real-time applications, we propose an efficient solver for forming such intersections for polyhedral constraints. The solver in question is a tailored implementation of a dual active-set quadratic programming solver that leverages the particular problem structure of the optimization problems arising for prioritized intersections. The method is validated in a real-time MPC application for autonomous driving, where it successfully resolves six different levels of conflicting constraints, confirming its efficiency and practicality for control. Furthermore, we show that the proposed solver outperforms existing solvers for hierarchical quadratic programming, making it relevant beyond control applications.