Advancements in Reinforcement Learning (2019)

This chapter explores the significant advancements and trends in reinforcement learning observed in 2019, highlighting the shift from traditional methods to new approaches. It discusses the evolution of benchmarks, the complexities of real-world applications, and contrasts simulation-based learning with physical robot constraints. The chapter also emphasizes the growing attention to model-based reinforcement learning and its implications for vision-based tasks, showcasing the potential of neural networks in predicting game dynamics.