Reinforcement Learning Algorithms for Autonomous Adaptive Agents

Intelligent agents are being designed to automate many tasks-for e.g., traffic signal control, vehicle driving, inventory control and are also being used in improving lives of people-like in healthcare, agriculture, wildlife protection etc. The widespread deployment of intelligent agents requires that we minimize the bottlenecks which affect their performance and utility. Motivated by this challenge, my thesis proposes new algorithms and methods which helps the agent in efficiently operating in the real-world and also during interaction with humans. My work has shown significant improvements in the performance of deployed agents, when operating in real world.