Design Of Natural Language Understanding On ROS Based Robot

With the development of science and technology, human has stepped into intelligent society, as a result, robots have been widely concerned and the technology about robot has been developing rapidly. Due to the tireless efforts of people, the functions which can be realized by robots become more and more complicated. However, almost all these functions are realized through the interaction between users and robots in natural language. Especially nowadays robots have been extended to families, which need robots to have the ability to understand natural language, which will determine whether the robots can execute user's commands correctly.As an open source robot operating system, ROS(Robot Operating System) caused a great impact in the field of robotics research, because of its advantages of simple to use and easy to develop. ROS has integrated the mainstream function libraries, such as Open CV, Open GL, etc. And it has lots of drivers of sensor, including Kinect, laser range finder etc. In the open source software of ROS, now we can only find the chatbot written by Justin Huang which relates to Natural Language Understanding. However, chatbot has been designed to realize the chat function of robot. Besides, it is only based on English, which is quite different with Chinese in processing.In this thesis, we mainly study the application of the Natural Language Understanding of Chinese to the robot based on ROS. First, we write the corresponding Server and Client to make the robot provide services like weather forecast, English-Chinese translation, playing music, etc. Then, we build order library for each service of the robot, including all functions, such as Face Recognition, Rotation and Moving, which involve voice commands. Here each service corresponds to numbers of commands to refine every service. To let robots answer some common questions from people, next we build the database of FAQ(Frequently Asked Questions). Finally, we use the method of sentence similarity calculation to calculate the similarity between the user's natural language and the questions in FAQ database or the natural language commands in the order library, to acquire the right answer or service. In this way, users can interact with robots in natural language, then make them complete various functions.