Szczegóły publikacji
Opis bibliograficzny
Robust reinforcement learning for overhead crane control with variable load conditions / Patryk BAŁAZY, Kamil Pieprzycki, Paweł KNAP // W: ICCC 2024 [Dokument elektroniczny] : 25th International Carpathian Control Conference : 22–24 May 2024, Krynica-Zdrój, Poland : proceedings / ed. Andrzej Kot. — Wersja do Windows. — Dane tekstowe. — Piscataway : IEEE, cop. 2024. — Dod. ISBN: 979-8-3503-5069-2, 979-8-3503-5071-5. — e-ISBN: 979-8-3503-5070-8. — S. [1–6]. — Wymagania systemowe: Adobe Reader. — Bibliogr. s. [6]. — Publikacja dostępna online od: 2024-07-01
Autorzy (3)
Słowa kluczowe
Dane bibliometryczne
| ID BaDAP | 154337 |
|---|---|
| Data dodania do BaDAP | 2024-07-10 |
| Tekst źródłowy | URL |
| DOI | 10.1109/ICCC62069.2024.10569379 |
| Rok publikacji | 2024 |
| Typ publikacji | materiały konferencyjne (aut.) |
| Otwarty dostęp |  |
| Wydawca | Institute of Electrical and Electronics Engineers (IEEE) |
Abstract
Overhead cranes are key machines in modern industry, used for transporting and loading heavy loads. However, controlling them is difficult due to the variability of sling length and weight. Most studies to date are based on a fixed load scenario, which is insufficient for practical application. This paper develops a novel reinforcement learning (RL) algorithm that uses a policy that is robust to object nonlinearity. Twin-delayed deep deterministic policy gradient (TD3) with randomization of object features is efficient for problems with stochastic complexity. The algorithm can learn a universal control method that is robust to a wide range of load mass changes. A simulation study was conducted to demonstrate the effectiveness of the proposed method. The results showed that the algorithm achieves satisfactory control performance compared to the classical LQR algorithm.
