Pick and Place with ROS in Unity. AI for motion planning and control of a…

The demo shows the integration of ROS with Unity. A trained deep-learning model is used to predict the position of the cube to perform object pickup and placement using computer vision with a robotic arm in Unity. The robotics system runs in a virtual container and Unity is connected to the ROS endpoint. Each time a pose estimation request is generated, we send an image from the observer camera to the pose estimation service in the ROS workspace that runs a neural network. The pose estimation model takes the image as input and determines the relative pose of the target object, which is used in the MoveIt planner service to determine the robot arm's trajectories for grasping and dropping.