State of the art in lifelong learning of robots

Learning offers a useful tool for the automatic design of autonomous robots. These Robots show an unprecedented capability to change to a greater extent as they can learn painlessly throughout their life from new tasks and environments. Such an attitude is crucial when deploying independent robots in climates and real-world settings that are dynamic and unpredictable. By learning constantly, the robots are in a position to evolve and improve at every given opportunity. This research outlines shifts in certain pathways at an individual level such as techniques used to prevent drastic memory loss, increasing memory’s fittingness, and usage of two memory systems. Models of human memory and learning as well as specific methods of dual memory structure are utilized to improve the performance of robotic agents and to replicate complex human behaviour. These comparisons allow researchers to get a clear picture of where the field is currently at and also the potential trajectories i t can take. The objective of this paper is to analyse the evolution of various approaches to the task of lifelong learning. It discusses the advantages of lifelong learning of robots and highlights the problems that researchers have faced for this such as catastrophic forgetting, limited scalability, complexity, cost issues, etc., and compares various techniques that have been used to address these problems. This paper discusses the pros and cons of various techniques and provides their comparative analysis.