Figure 1
From: Robustness of quantum reinforcement learning under hardware errors
Summary of the scenarios analysed in the present work. We consider two models for quantum reinforcement learning (QRL) agents and test their performance on two environments, CartPole and the Travelling Salesperson Problem (TSP). We analyse the performance of the agents when these are trained and used in the presence of the most common noise sources found on real quantum hardware, namely statistical fluctuations due to shot noise, coherent errors due to imperfect control or calibration of the device, and incoherent errors coming from the unavoidable interaction of the quantum hardware with its environment