Programming a robotic arm to perform a specific task is time-consuming. This project proposes a novel implementation of self-learning and Neural Network (NN) based specifically designed and 3d printed Robotic Arm fixed on a rover which navigates its path through Simultaneous localization and mapping(SLAM) for the multipurpose application using ROS. The arm uses forward and inverse kinematics to calculate the angle of rotation in each motor to get the required position and then picks the object from the detected object which is done by a Convolution Neural Network from the required object’s dataset.