A Constraint Homotopy Active Set Solver for Linear-Quadratic Optimal Control

• 成果类型: Article
• 署名作者: Buerger, Johannes; Cannon, Mark
• 署名单位: BMW AG; University of Oxford
• 刊物名称: IEEE TRANSACTIONS ON AUTOMATIC CONTROL
• ISSN/ISSBN: 0018-9286
• DOI: 10.1109/TAC.2024.3400356
• 发表日期: 2024
• 页码: 7205-7210
• 关键词: Active set solver; linear-quadratic (LQ) optimal control; parametric programming; quadratic programming (QP)

摘要

An efficient optimization method is proposed for linear-quadratic optimal control problems with state and control constraints. We describe an active set solver that uses Riccati recursions to solve a sequence of equality-constrained subproblems. The main contribution is a homotopy method based on relaxing inequality constraints. This overcomes known shortcomings of Riccati active set solvers relating to their initialization and their application to problems with time-varying model data. It can be used exclusively or in combination with established Riccati active set solvers. The efficiency is demonstrated in numerical examples against state-of-the-art quadratic programming solvers.