{"id":736,"date":"2021-08-29T23:10:39","date_gmt":"2021-08-29T23:10:39","guid":{"rendered":"https:\/\/elo-x.eu\/?p=736"},"modified":"2023-10-16T17:31:46","modified_gmt":"2023-10-16T17:31:46","slug":"rudolf-reiter","status":"publish","type":"post","link":"https:\/\/elo-x.eu\/?p=736","title":{"rendered":"Rudolf Reiter"},"content":{"rendered":"\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\n\t\t
\n\n\t\t\t\t\t\t\t\t\t\t\t\n\t\t\t\t\t\t

\n\t\t\t\t\t\t\t\t\n\t\t\t\tRudolf Reiter \n\t\t\t<\/h2 > \n\t\t\t\t\t\t\t\t\t<\/a>\n\t\t\t\t\t<\/div>\n\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t
\n\t\t\t\n\t\t\t\t\t\t<\/span>\n\t\t<\/div>\n\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

PhD Candidate in Engineering Science<\/span><\/span><\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

SYSCOP Laboratory<\/b><\/span><\/p>

University of Freiburg<\/b><\/span><\/p><\/div>

\u00a0<\/div><\/div>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\"\"\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

Rudolf Reiter graduated 2016 with a Master\u2019s degree in electrical engineering with a focus on control systems from the Graz University of Technology in Austria. During that time he did internships at Bernecker & Rainer (ABB) and an exchange semester at the University of Utah, USA. Afterwards, he joined an international company for measurement devices (Anton Paar GmbH) where he worked as a specialist for control systems and signal processing. In 2018 he started working as a researcher for autonomous driving at a vehicle research company. There he also joined the Autonomous Racing Graz team, which competes at real-world autonomous racing events.<\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t

\n\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t
\n\t\t\t\n\t\t\t\t\t\t<\/span>\n\t\t<\/div>\n\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t

Project description<\/h2>\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

Rudolf Reiter is actively engaged in cutting-edge research in the field of autonomous driving and autonomous racing as part of his pursuit of a Ph.D. His work primarily focuses on several key areas, including model predictive control, mixed integer optimization, optimal control, and reinforcement learning. By combining expertise in these domains, he aims to develop innovative solutions that can enhance the performance and safety of autonomous vehicles, especially in strategic racing scenarios and combinatorial sophisticated multi-lane traffic. His research promises to contribute significantly to the advancement of autonomous systems, making them more efficient and capable in navigating complex environments while pushing the boundaries of autonomous racing technology.<\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t

\n\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\n\t\t\t\t\t\t\n\t\t\t\t\t\t\t\t\tRead more about this project<\/span>\n\t\t\t\t\t<\/span>\n\t\t\t\t\t<\/a>\n\t\t\t\t<\/div>\n\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t
\n\t\t\t
<\/div>\t\t<\/div>\n\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t
\n\t\t\t\n\t\t\t\t\t\t<\/span>\n\t\t<\/div>\n\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t

Publications<\/h2>\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t
<\/a>
<\/div><\/div><\/form>
1.<\/div>

Reiter, Rudolf; Ghezzi, Andrea; Baumg\u00e4rtner, Katrin; Hoffman, Jasper; McAllister, Robert D; Diehl, Moritz<\/p>

AC4MPC: Actor-Critic Reinforcement Learning for Nonlinear Model Predictive Control<\/a> Working paper<\/span> <\/p>

2024<\/span>.<\/p>

Abstract<\/a><\/span> | Links<\/a><\/span> | BibTeX<\/a><\/span><\/p>

@workingpaper{reiter2024ac4mpc,
\r\ntitle = {AC4MPC: Actor-Critic Reinforcement Learning for Nonlinear Model Predictive Control},
\r\nauthor = {Rudolf Reiter and Andrea Ghezzi and Katrin Baumg\u00e4rtner and Jasper Hoffman and Robert D McAllister and Moritz Diehl },
\r\nurl = {https:\/\/doi.org\/10.48550\/arXiv.2406.03995},
\r\nyear  = {2024},
\r\ndate = {2024-06-06},
\r\nabstract = {Ac{MPC} and ac{RL} are two powerful control strategies with, arguably, complementary advantages. In this work, we show how actor-critic ac{RL} techniques can be leveraged to improve the performance of ac{MPC}. The ac{RL} critic is used as an approximation of the optimal value function, and an actor roll-out provides an initial guess for primal variables of the ac{MPC}. A parallel control architecture is proposed where each ac{MPC} instance is solved twice for different initial guesses. Besides the actor roll-out initialization, a shifted initialization from the previous solution is used. Thereafter, the actor and the critic are again used to approximately evaluate the infinite horizon cost of these trajectories. The control actions from the lowest-cost trajectory are applied to the system at each time step. We establish that the proposed algorithm is guaranteed to outperform the original ac{RL} policy plus an error term that depends on the accuracy of the critic and decays with the horizon length of the ac{MPC} formulation. Moreover, we do not require globally optimal solutions for these guarantees to hold. The approach is demonstrated on an illustrative toy example and an ac{AD} overtaking scenario.},
\r\nkeywords = {},
\r\npubstate = {published},
\r\ntppubtype = {workingpaper}
\r\n}
\r\n<\/pre><\/div>
Close<\/a><\/p><\/div>
Ac{MPC} and ac{RL} are two powerful control strategies with, arguably, complementary advantages. In this work, we show how actor-critic ac{RL} techniques can be leveraged to improve the performance of ac{MPC}. The ac{RL} critic is used as an approximation of the optimal value function, and an actor roll-out provides an initial guess for primal variables of the ac{MPC}. A parallel control architecture is proposed where each ac{MPC} instance is solved twice for different initial guesses. Besides the actor roll-out initialization, a shifted initialization from the previous solution is used. Thereafter, the actor and the critic are again used to approximately evaluate the infinite horizon cost of these trajectories. The control actions from the lowest-cost trajectory are applied to the system at each time step. We establish that the proposed algorithm is guaranteed to outperform the original ac{RL} policy plus an error term that depends on the accuracy of the critic and decays with the horizon length of the ac{MPC} formulation. Moreover, we do not require globally optimal solutions for these guarantees to hold. The approach is demonstrated on an illustrative toy example and an ac{AD} overtaking scenario.<\/div>
Close<\/a><\/p><\/div>