This repository provides a simulation environment for training reinforcement learning agents to perform robotic pick-and-place operations. The agent receives visual observations (images) and must predict grasp success chances pixel-wise to determine where to place the gripper. The environment uses MuJoCo for physics simulation and implements a custom controller class for robot control, including multi-discrete action spaces and two-finger gripper mechanics.