Robot Learning is a challenging task. It is hard for a beginner to start the process of training a robot to learn skills from scratch. I will try to use some basic code to explain this procedure.
You need some basic skills, such as C++/Python, Deep Learning, ROS, Robotics, Reinforcement Learning.
The basic is a URDF robot model. We will transfer it from ROS to MuJoCo and train it with the PPO algorithm.
We show how to change the URDF model to MuJoCo xml model.
We define the OpenAI gym robot environment and a pick&place task.
We use PPO to train the robot.
We use the policy trained in MuJoCo to control the robot in Gazebo.