• The Journal of Korea Robotics Society
• /
• v.14 no.2
• /
• pp.81-86
• /
• 2019

Robots are used in various industrial sites, but traditional methods of operating a robot are limited at some kind of tasks. In order for a robot to accomplish a task, it is needed to find and solve accurate formula between a robot and environment and that is complicated work. Accordingly, reinforcement learning of robots is actively studied to overcome this difficulties. This study describes the process and results of learning and solving which applied reinforcement learning. The mission that the robot is going to learn is bottle flipping. Bottle flipping is an activity that involves throwing a plastic bottle in an attempt to land it upright on its bottom. Complexity of movement of liquid in the bottle when it thrown in the air, makes this task difficult to solve in traditional ways. Reinforcement learning process makes it easier. After 3-DOF robotic arm being instructed how to throwing the bottle, the robot find the better motion that make successful with the task. Two reward functions are designed and compared the result of learning. Finite difference method is used to obtain policy gradient. This paper focuses on the process of designing an efficient reward function to improve bottle flipping motion.

• The Journal of Korea Robotics Society
• /
• v.17 no.1
• /
• pp.1-7
• /
• 2022

Robots are widely used in industries and services. Traditional robots have been used to perform repetitive tasks in a fixed environment, and it is very difficult to solve a problem in which the physical interaction of the surrounding environment or other objects is complicated with the existing control method. Reinforcement learning has been actively studied as a method of machine learning to solve such problems, and provides answers to problems that robots have not solved in the conventional way. Studies on the learning of all physical robots are commonly affected by noise. Complex noises, such as control errors of robots, limitations in performance of measurement equipment, and complexity of physical interactions with surrounding environments and objects, can act as factors that degrade learning. A learning method that works well in a virtual environment may not very effective in a real robot. Therefore, this paper proposes a weighted sum method and a linear regression method as an effective and accurate learning method in a noisy environment. In addition, the bottle flipping was trained on a robot and compared with the existing learning method, the validity of the proposed method was verified.