Online Next best view (NBV) planner for a 6-DOF robotic arm using Deep Reinforcement Learning for 3D scene reconstruction

BSc assignment

How can you control a 6-DOF robotic arm to tackle the single robot Next best view (NBV) problem using a Reinforcement Learning model for 3D scene reconstruction?

At the end of the assignment, I will try to have succesfully trained and deployed the RL model on the robot arm. This robot arm should be able to carry out semi-optimized trajectories which allows the robot to reconstruct a 3D scene using the attached Intel real sense RGB-D camera.

I hope that the robot will be sufficiently trained such that it can generalise well to unknown objects and scenes and that the trained simulation will carry over to the real life robot.