GSoC’22 RoboComp project: Reinforcement Learning for pick and place operations
30th June 2022
Hi! I am Daniel Peix, graduated in Computer Engeneering at ‘Universidad de Extremadura’ in Cácerew, Spain. Since I wass a child, I had facilities for Maths and for solving logic and ingenuity problems. Over time, this has led to love programming, robotics and artificial intelligence. This opportunity working in a project at RoboComp offers me the option to explore and learn more about the fields of computer science that interest me.
About the Project
A robot can face a lot of different situations, because the real world is not and ideal environment. If anyone wants to implement a software solution for any situation that could happen, an insane amount of code would be needed. In order to avoid that issue, reinforcement learning is used. This kind of algorithm gives the robot a way to be more flexible when facil the different situations that can take place in the real world, as well as it reduces de efficiency, because a learnt task will alway sbe less efficient that the one implemented in a dedicated code.
The goal of this project is exploring some reinforcement learning algorithms, in order to solve the robotic arm’s grasping task. In order to so, it will be necessary to develop an environment implementing the OpenAI Gym interface, to be able to support reinforcement learning processes. Then, different RL algorithms will be tested and the best ones will be chosen to explore the solutions to the problem.