The shopping assistant Robot design based on ROS and deep learning

As the traditional service robots' artificial intelligence bottlenecks, there is a huge gap between service robots and human intelligence in the cognitive and learning discipline. So the service robots could not be widely applied. As deep learning theory is proposed in 2012, it may lead to a generation leap forward in machine learning discipline, so as to improving the traditional robot's cognitive algorithms. This paper studies the principle of three-dimensional Kinect sensor and the deep learning framework of CNN. We proposed a shopping assistant robot designing method which combined Robot Operation System and deep learning method. Firstly, the ROS packages for the service robot are designed. Secondly, Kinect sensors are used for acquiring the information in the robot. Finally, we use simulation to evaluated the design.