Reinforcement Learning with Elastic Time Steps

arXiv:2402.14961v1 Announce Type: cross
Abstract: Traditional Reinforcement Learning (RL) algorithms are usually applied in robotics to learn controllers that act with a fixed control rate. Given the discrete nature of RL algorithms, they are oblivious to the effects of the choice of control rate: finding the correct control rate can be difficult and mistakes often result in excessive use of computing resources or even lack of convergence.
We propose Soft Elastic Actor-Critic (SEAC), a novel off-policy actor-critic algorithm to address …

abstract act algorithms arxiv control cs.lg cs.ro effects elastic learn mistakes nature rate reinforcement reinforcement learning robotics type