Study And Design Of Natural Language Understanding And Mission Planning For Home Service Robot

Because of the looking forward to liberation from the heavy simple labor, service robots become a recent hot research project. Robotics is a comprehensive discipline, so the service robot also includes a wide range of technical. The natural language understanding is mainly responsible for solving the problem of human-computer interaction. It provides a convenient means for communication between humans and robots. And mission planning helps to improve the operating efficiency of the robot, making it complete instructions faster and more accurately issued by human beings.By researching and improving dependency grammar and valence theory of the natural language processing method, and evolutionary algorithm of optimization algorithm, this paper develops a natural language understanding algorithms and a mission planning algorithm for service robots.Firstly, this paper describes some of the basic methods of natural language understanding and robotics mission planning, by discovery its development status, and explaining the significance of these two technologies in service robots. Then, the dependency grammar and valence theory, and evolutionary algorithm are analyzed in detail to identify their advantages and disadvantages, as well as their applicability to to the service robot applications environment.Secondly, the dependency grammar and valence theory is applied to the service robot natural language understanding. Using RoboCup@home simulation as an example, we analyze the vocabulary sentence characteristics, summarized the vocabulary commonly used method, and build the valence glossary based on the idea of dependency grammar and valence theory. Based on the valence glossary to determine the dependency structure of the sentence, the transformation is achievable between natural language and machine language of imperative sentence and declarative sentence.The third part is about improving the traditional evolutionary algorithm, making it suitable to solve robot mission planning problems. This article builds a new coding structure called step based on the characteristics of mission planning. So the missions can be expressed by gene on the chromosome, thereby the method of the completion of mission can be changed by changing the structure of the chromosome. At the same time, this article changes the way to take evolutionary way of allogeneic parity crossover and random variation in traditional evolutionary algorithm. The new method uses autologous ectopic crossover and restricted choice variant to solve the problem of mission scheduling and item selection. We also add the concept of filtering to eliminate the individual does not comply with the rules of logic before evaluation, thereby improved the operation efficiency and the accuracy of the results. Besides we design the fitness function for the evaluation of individual fitness——the degree of the planning, which can be used to determine the phase-out of the individual.Finally, this article demonstrates the computing process of the two algorithms with examples. The effectiveness and the superiority of them are verified by statistical comparative analysis of the results of the competition with the other teams using RoboCup@home simulation as experimental material. And the achievements in the real competition show the practicality of the algorithms.