Abstract
Deep learning has provided new ways of manipulating, processing and analyzing data. It sometimes may achieve results comparable to, or surpassing human expert performance, and has become a source of inspiration in the era of artificial intelligence. Another subfield of machine learning named reinforcement learning, tries to find an optimal behavior strategy through interactions with the environment. Combining deep learning and reinforcement learning permits resolving critical issues relative to the dimensionality and scalability of data in tasks with sparse reward signals, such as robotic manipulation and control tasks, that neither method permits resolving when applied on its own. In this paper, we present recent significant progress of deep reinforcement learning algorithms, which try to tackle the problems for the application in the domain of robotic manipulation control, such as sample efficiency and generalization. Despite these continuous improvements, currently, the challenges of learning robust and versatile manipulation skills for robots with deep reinforcement learning are still far from being resolved for real-world applications.
Highlights
Robots were originally designed to assist or replace humans by performing repetitive and/or dangerous tasks which humans usually prefer not to do, or are unable to do because of physical limitations imposed by extreme environments
A robotic arm can be described as a chain of links that are moved by joints which are actuated by motors
Reinforcement learning [20] is a subfield of machine learning, concerned with how to find an optimal behavior strategy to maximize the outcome though trial and error dynamically and autonomously, which is quite similar with the intelligence of human and animals, as the general definition of intelligence is the ability to perceive or infer information, and to retain it as knowledge to be applied towards adaptive behaviors in the environment
Summary
Robots were originally designed to assist or replace humans by performing repetitive and/or dangerous tasks which humans usually prefer not to do, or are unable to do because of physical limitations imposed by extreme environments. If a component is slightly shifted, the control system may have to stop and to be recalibrated Under this strategy, the robot arm performs by following a series of positions in memory, and moving to them at various times in their programming sequence. With the advancements in modern technologies in artificial intelligence, such as deep learning, and recent developments in robotics and mechanics, both the research and industrial communities have been seeking more software based control solutions using low-cost sensors, which has less requirements for the operating environment and calibration. We follow the discussion and present other real-world challenges of utilizing DRL in robotic manipulation control in the forth section, with a conclusion of our work in the last section
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
