Quantum reinforcement learning in dynamic environments

TL;DR

This paper explores extending quantum reinforcement learning to dynamic environments by modifying a hybrid quantum-classical agent with dissipation, demonstrating improved adaptability and success in time-dependent reward settings.

Contribution

It introduces a dissipation mechanism to a hybrid quantum-classical RL agent, enabling effective learning in dynamic, time-dependent environments.

Findings

Hybrid agent adapts quickly to environmental changes.

Modified hybrid agent outperforms classical RL in success probability.

Demonstrates feasibility of quantum RL in non-stationary settings.

Abstract

Combining quantum computing techniques in the form of amplitude amplification with classical reinforcement learning has led to the so-called "hybrid agent for quantum-accessible reinforcement learning", which achieves a quadratic speedup in sample complexity for certain learning problems. So far, this hybrid agent has only been applied to stationary learning problems, that is, learning problems without any time dependency within components of the Markov decision process. In this work, we investigate the applicability of the hybrid agent to dynamic RL environments. To this end, we enhance the hybrid agent by introducing a dissipation mechanism and, with the resulting learning agent, perform an empirical comparison with a classical RL agent in an RL environment with a time-dependent reward function. Our findings suggest that the modified hybrid agent can adapt its behavior to changes in the environment quickly, leading to a higher average success probability compared to its classical counterpart.