I am currently working with classical roboticists who insist on inverse kinematics, and what I (perhaps mistakenly) call the old way of thinking about robots accomplishing tasks. Much of the relatively recent research focuses on Robots using Brain models such as Multiple timescales (Artificial Intelligence models) that segment sequences and reproduce them, having learned them. The problem I face is this bunch of roboticists insist that a robot already knows the sequence, and training it to be reproduced is redundant, since a Robot can already reproduce the sequence anyway. How accurate would you rate this assessment of using AI in robotics? Are there any advantages of using AI to learn sequences for robot control?

  • $\begingroup$ Speaking as a "roboticist" that specializes in reinforcement learning, I am not quite sure what you are asking here. What is the task you are trying to accomplish? Are you training a robot to reproduce a specific movement? Cause in that case using classic control is both convenient and easy, and introducing AI to solve this would be a lot of extra work for little reward. If you however want the robot to generalize over a task, and be adaptable to new situations, then it becomes a problem fit for AI, specifically reinforcement learning. $\endgroup$ – Krrrl Oct 27 '19 at 21:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.