Human-robot Interaction Based On Battle Management Language For Multi-robot System

Commanding and controlling a multi-robot system is a challenging task.Static control commands are difficult to fully meet the requirements of controlling different robots.This thesis uses restricted natural language to control multi-robot systems,and proposes a framework based on Battle Management Language(BML)to control multi-robot systems.BML human-robot interaction framework consists of three parts,BML markup,vocabulary tree and role tree.Each robot has a role tree and a vocabulary tree.When inputing BML commands,the BML commands can be converted into instructions that the robot can execute by searching the role tree and vocabulary tree.Based on the BML human-robot interaction framework,this thesis proposes a contextfree language BML to control multi-robot systems.Based on context-free language,text and speech can be used for human-robot interaction.When the input text conforms to the grammatical rules of BML,it can be recognized by the robot and transformed into instructions executable for the robot.In order to make the control system only obey the command of the commander,this thesis uses Gaussian Mixture Models-Universal Background Model(GMM-UBM)to recognize the speaker,and let the control system only respond to the command of the commander's voice.The advantage of GMM-UBM is that after training a general model,recognition can be realized by a voice of about 10 s.The GMM-UBM trained in this thesis has an equal error rate of 0.218% on our own data set.Multi-robot systems can accomplish many complex tasks.In actual use,they will face unexpected problems.They cannot use preset commands to accomplish tasks.In order to solve this problem,more basic instructions are needed to control the robot,and a combination of these basic instructions is used to accomplish complex tasks.This thesis proposes RML(Robot Management Language)and RVM(RML Virtual Machine)to control the robots.RVM is the running environment of RML.The robots can be controlled by registering functions in RVM.RML supports conditional statements,loop statements,and function calls.By writing RML,robots can accomplish non-preset complex tasks without presets.The experiments based on simulation system and real robot systems are performed using Robo Cup robot soccer as the application.1?Human-robot interaction using text and speech was performed to realize tasks such as such as controlling a single robot,robot separating from and joining into the group,commanding the multi-robot system and comparative experiments with traditional remote control methods.The results show the BML-based human-robot interaction framework proposed in this thesis is a universal framework,which allows users to control multi-robot systems using task-level and motion-level commands,and has obvious advantages when completing complex tasks.2?The human-robot interaction using RML was carried out in the simulation system.The grammar and computational performance of RML and the interaction capabilities between RVM and robots were verified.The results show that,RML allows robots to complete non-preset complex tasks.