eprintid: 10194559 rev_number: 9 eprint_status: archive userid: 699 dir: disk0/10/19/45/59 datestamp: 2024-07-16 09:07:58 lastmod: 2024-07-16 09:07:58 status_changed: 2024-07-16 09:07:58 type: article metadata_visibility: show sword_depositor: 699 creators_name: De Lellis, Francesco creators_name: Coraggio, Marco creators_name: Russo, Giovanni creators_name: Musolesi, Mirco creators_name: di Bernardo, Mario title: Guaranteeing Control Requirements via Reward Shaping in Reinforcement Learning ispublished: inpress divisions: UCL divisions: B04 divisions: F48 keywords: Computational control, deep reinforcement learning (RL), learning-based control, policy validation, reward shaping note: © 2024 The Authors. This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see (https://creativecommons.org/licenses/by/4.0/). abstract: In addressing control problems such as regulation and tracking through reinforcement learning (RL), it is often required to guarantee that the acquired policy meets essential performance and stability criteria such as a desired settling time and steady-state error before deployment. Motivated by this, we present a set of results and a systematic reward-shaping procedure that: 1) ensures the optimal policy generates trajectories that align with specified control requirements and 2) allows to assess whether any given policy satisfies them. We validate our approach through comprehensive numerical experiments conducted in two representative environments from OpenAI Gym: the Pendulum swing-up problem and the Lunar Lander. Utilizing both tabular and deep RL methods, our experiments consistently affirm the efficacy of our proposed framework, highlighting its effectiveness in ensuring policy adherence to the prescribed control requirements. date: 2024-05-17 date_type: published publisher: Institute of Electrical and Electronics Engineers official_url: http://dx.doi.org/10.1109/tcst.2024.3393210 oa_status: green full_text_type: other language: eng primo: open primo_central: open_green verified: verified_manual elements_id: 2115444 doi: 10.1109/TCST.2024.3393210 lyricists_name: Musolesi, Mirco lyricists_id: MMUSO05 actors_name: Musolesi, Mirco actors_id: MMUSO05 actors_role: owner funding_acknowledgements: [European Union] full_text_status: public publication: IEEE Transactions on Control Systems Technology pages: 12 citation: De Lellis, Francesco; Coraggio, Marco; Russo, Giovanni; Musolesi, Mirco; di Bernardo, Mario; (2024) Guaranteeing Control Requirements via Reward Shaping in Reinforcement Learning. IEEE Transactions on Control Systems Technology 10.1109/TCST.2024.3393210 <https://doi.org/10.1109/TCST.2024.3393210>. (In press). Green open access document_url: https://discovery.ucl.ac.uk/id/eprint/10194559/2/Musolesi_Guaranteeing%20Control%20Requirements%20via%20Reward%20Shaping%20in%20Reinforcement%20Learning_AAM.pdf