Training Agents to Play 2D Games Using Reinforcement Learning
Harshil Jhaveri, Nishay Madhani, Narendra M. Shekokar · 2021
This chapter provides an important step in the field of reinforcement learning, and their agent outperforms all human benchmarks on Atari games. Effective emulation, simulation and agent-based application of learning is a very sophisticated criterion of establishing the supremacy and applicability of visual and spatial parameters. The advent of artificial intelligence can be effectively used to create advanced systems that are capable of decision-making, solve complex issues and emulate human processes of learning and decision-making virtually and comparably accurate. A special type of algorithm called reinforcement learning enables an agent to learn within an environment and earn the maximum reward available. AI-based training of a CPU agent to simulate and play games in diverse environments has been thought of for quite some time but has been relatively paused due to the unavailability of GPUs and fast processors owing to its high computation requirement.