We introduce a reinforcement learning algorithm assisted by a feedback controller. The idea is to enhance tabular learning algorithms by means of a control strategy with limited knowledge of the system model. We show that, by tutoring the learning process, the algorithm converges more quickly than the tabular Q-learning strategy. We use the classical problem of stabilizing an inverted pendulum as a benchmark to numerically illustrate the advantages and disadvantages of the approach.
Tutoring Reinforcement Learning via Feedback Control / De Lellis, F.; Russo, G.; Di Bernardo, M.. - (2021), pp. 580-585. (Intervento presentato al convegno 2021 European Control Conference, ECC 2021 tenutosi a nld nel 2021) [10.23919/ECC54610.2021.9654881].
Tutoring Reinforcement Learning via Feedback Control
De Lellis F.;Di Bernardo M.
2021
Abstract
We introduce a reinforcement learning algorithm assisted by a feedback controller. The idea is to enhance tabular learning algorithms by means of a control strategy with limited knowledge of the system model. We show that, by tutoring the learning process, the algorithm converges more quickly than the tabular Q-learning strategy. We use the classical problem of stabilizing an inverted pendulum as a benchmark to numerically illustrate the advantages and disadvantages of the approach.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
