Vehicle Driving Control Based On Natural Language Instructions

Using natural language instructions to control the vehicle traveling frees the human from the tedious task of driving and improves the traffic safety, which has important application value. In this paper, an analog car is the targeted research object in order to research about using natural language instruction to control the analog car driving, by means of the analysis and conversion of natural language instruction, using the result controls the analog car traveling automatically under the indoor simulated traffic environment.In this paper, the main research work including:Firstly, analyze and transform the natural language instructions. I.e. receiving the unconstrained natural language instruction in the form of text from PC and extracting the linguistic unit which is necessary to the vehicle traveling process, and converts it to the standard format command which is necessary to the vehicle. When extracting the language unit information from natural language instructions, we use the sequence labeling method, we define nine kinds of semantic categories according to the language unit required by the vehicle traveling, and we use part of speech and syntactic structure component as the feature and label out the semantic category sequence with CRF toolkit. The semantic category sequence is then cut into a specific operation, and according to a certain format converts it to the control command. Using Prefixspan algorithm to mining frequent sequential patterns, with the help of Label Propagate algorithms and artificial constructed field dictionary, using pattern-matching algorithm, the semantic category sequences are cut into one or more of the driving operation, for each running operation, it will be converted to a predetermined structure which is related to the traveling and includes six parts of the information, each structure including:a starting point, speed, direction, distance, the action and the termination point. Using rules, statistics and reasoning method can standard each section of the structure in order that the analog car can receive simple and no ambiguity information.Secondly, the natural language instruction executes. I.e. transforming the structured information into controlling signal that the analog car is able to identify and send it to the analog car, lastly, the instruction is executed under the indoor simulated traffic environment. In this paper, it is using Wireless communication to realize the communication between the PC and the analog car. Communication device at one end is the serial port, the other end connected to Bluetooth. Programming on the serial port of the PC, it can realize sending out the controlling signal from the serial port. The analog car receives the controlling signal and resolves it with the command resolution module. The analog car completes the resolved driving operations with the help of relevant information obtained by the camera, RFID, digital compass and so on.