An Idiosyncrasy of Time-discretization in Reinforcement Learning
This work solves a practical problem for researchers and practitioners applying RL to real-world systems with time-discretization choices, though it is incremental in nature.
The paper addresses the mismatch between continuous-time physical systems and discrete-time reinforcement learning algorithms, showing that a simple modification can better align return definitions when discretization granularity is variable or stochastic.
Many reinforcement learning algorithms are built on an assumption that an agent interacts with an environment over fixed-duration, discrete time steps. However, physical systems are continuous in time, requiring a choice of time-discretization granularity when digitally controlling them. Furthermore, such systems do not wait for decisions to be made before advancing the environment state, necessitating the study of how the choice of discretization may affect a reinforcement learning algorithm. In this work, we consider the relationship between the definitions of the continuous-time and discrete-time returns. Specifically, we acknowledge an idiosyncrasy with naively applying a discrete-time algorithm to a discretized continuous-time environment, and note how a simple modification can better align the return definitions. This observation is of practical consideration when dealing with environments where time-discretization granularity is a choice, or situations where such granularity is inherently stochastic.