Reactive Reinforcement Learning in Asynchronous Environments

TL;DR

This paper introduces reactive reinforcement learning algorithms designed for asynchronous environments, enabling agents to act immediately after observation, thereby reducing reaction time and improving safety in robotic tasks.

Contribution

The paper proposes a new class of reactive RL algorithms that act instantly after observations, addressing asynchronous environment challenges and enhancing response times.

Findings

Reactive SARSA reduces reaction time by approximately the duration of the learning update.

Reactive algorithms improve safety and decision speed in robotic tasks.

Comparison shows reactive RL outperforms conventional methods in asynchronous settings.

Abstract

The relationship between a reinforcement learning (RL) agent and an asynchronous environment is often ignored. Frequently used models of the interaction between an agent and its environment, such as Markov Decision Processes (MDP) or Semi-Markov Decision Processes (SMDP), do not capture the fact that, in an asynchronous environment, the state of the environment may change during computation performed by the agent. In an asynchronous environment, minimizing reaction time---the time it takes for an agent to react to an observation---also minimizes the time in which the state of the environment may change following observation. In many environments, the reaction time of an agent directly impacts task performance by permitting the environment to transition into either an undesirable terminal state or a state where performing the chosen action is inappropriate. We propose a class of reactive reinforcement learning algorithms that address this problem of asynchronous environments by immediately acting after observing new state information. We compare a reactive SARSA learning algorithm with the conventional SARSA learning algorithm on two asynchronous robotic tasks (emergency stopping and impact prevention), and show that the reactive RL algorithm reduces the reaction time of the agent by approximately the duration of the algorithm's learning update. This new class of reactive algorithms may facilitate safer control and faster decision making without any change to standard learning guarantees.