Research On Human-Computer Interaction Method Of Mobile Robot Based On Natural Language Understanding

The Research in the field of mobile robots is in full swing.In daily voice command control,it is a challenging problem to convert Chinese commands with flexible structures and containing a large number of synonyms into machine language.The current mainstream instruction parsing is mainly centered on key information extraction.The current common keyword extraction methods are rule-based key information extraction and statistics-based key information extraction.Rule-based key information extraction is based on linguistics.A series of language rules are summarized to analyze and process natural language.This method has strong expansion capabilities,but it consumes manpower and material resources,workload,and the probability of errors varies with instructions.The size of the set grows linearly.The key information extraction based on statistics is to use machine learning methods to establish the mapping relationship between instructions and machine language with the help of a large-scale corpus.However,this method is not very expandable and needs to continuously update the machine learning model.In response to the above problems,this article combines the rule-based method with the statistics-based method to complete the human-computer interaction strategy of mobile robots in a complexand flexible command environment.The main research contents are as follows:The paper designs a Bayesian classifier of Bernoulli model to complete the task of Chinese instruction classification,and on this basis,proposes a Chinese instruction parsing method based on syntactic analysis.Based on the CBOW continuous word vector model,the Chinese instructions of the mobile robot are converted into 200-dimensional vectors that the computer can understand,and the classification task of the Chinese instructions is completed according to the Bayesian classifier based on the Bernoulli model.According to specific types of Chinese instructions,using the dependency syntax tree and the Chinese syntactic knowledge base built by itself,it dynamically analyzes the structure of Chinese instructions and extracts key information.A synonym mapping method based on word vector contrast similarity is designed,which converts Chinese information with the same meaning and different expressions into the same machine instructions.Using the continuous word vector model based on word meaning,the synonym mapping is completed by comparing the similarity between the standard information word vector and the extracted key information word vector.Design a cross-platform remote mobile robot control system,adopt a micro-service architecture as a whole,use TCP transport layer protocol for communication between services,and design an application layer protocol by itself.In the Windows system,design the human-computer interaction client platform and the interactive algorithm server equipped with Chinese instruction parsing algorithms;in the Linux system,design and realize the overall structure of the two-wheel differential mobile robot,the control interface and the interactive server of the system control center.Taking into account the performance requirements of the control center,the thread pool is designed to support the intensive computing power of the system and optimize the efficiency of the system.The simulation experiment in the ROS system verifies the validity of the design.