Design And Implementation Of Software Robot System For Typical Application

With the continuous breakthrough of artificial intelligence,big data,cloud computing and other information technologies,the digital economy is booming,and the demand of producers for intelligent information in the production process is growing.It is a new trend to use robot process automation to improve production efficiency.However,the intelligent degree of software robots in traditional robot process automation is low.It can't deal with the complicated and diverse operation tasks in the existing software and it can't meet the intelligent needs.For this reason,this paper studies,analyzes requirements,designs and implements the software robot for typical applications,and improves the intelligence of software robot,including the following aspects:Firstly,this paper studies the representation modeling method of software robot behavior in typical applications.In order to solve the defects of the existing model to describe the typical application software behavior,this paper analyzes the users' typical application software behavior,designs a mixed state model combining the finite state and infinite state of the software,and implements the state collection algorithm of software menu area.Comparison experiments show that the proposed model can reduce the time of menu state traversal task by 28% and the memory consumption by 29.91% compared with the existing model.Secondly,in order to solve the problem of software robot in intelligence improvement,a self-learning method of software robot is proposed based on deep reinforcement learning DQN algorithm,which enables software robot to learn and use software independently in the training environment.At the same time,a training environment construction algorithm for typical application is proposed.In the end,the experimental results show that the target reachable rate of the optimized learning algorithm reaches 98.2%.Thirdly,aiming at the interaction problem between users and software robots,the human-computer interaction model of software robot is proposed by analyzing the human-computer interaction relationship.The front-end design of the software robot interaction language is carried out.The lexical analyzer and the grammar analyzer of the interactive language are designed and implemented in detail.The experiment proves that the lexical analyzer can divide words correctly and the grammar analyzer can check the grammar.Finally,based on the above work,this paper designs the overall architecture and main function modules of the software robot system for typical applications,and implement the system.After testing the system,the functions of system meet the requirements,and the system can complete various tasks issued by users.