Reinforcement learning (RL) has emerged as one of the most important fields of AI and represents a step towards building autonomous systems with a higher-level understanding of the visual world. For instance, one of the recent advances is the development of AlphaGo – a computer programme that combines advanced search tree with deep neural networks. These neural networks use RL techniques to take a description of the Go board as an input and process it through a number of different network layers containing millions of neuron-like connections. In this research, we seek to explore the use of RL techniques in the field of Operations Management and replicate its successes in this area.

Real-world applications in Operations Management require RL algorithms to act in the presence of model and data uncertainty. During learning process, it is likely that the agent executes sub-optimal actions that may lead to unsafe/poor states of the system. Exploration is particularly brittle in high-dimensional state/action space due to an increased number of low performing actions and noisy observations. We seek to build on Robust Optimisation techniques to build exploration in approximate RL setting. To ensure robustness during learning, we will explore the robust policy and value iteration algorithms.