Design And Implementation Of Task Planning Framework For Service Robot In Incomplete Knowable Environment

Due to the limitation of perception,indoor service robots may not be able to obtain complete taskrelated information.Therefore,indoor service robots are faced with a kind of incomplete knowable environment.According to the requirements of task planning in such environments,a task planning system for service robots in an incomplete knowable environment is proposed and implemented,which improves the task planning and execution ability of indoor service robots when task information is missing.ROS,the Robot Operation System,is introduced as the software middleware of the task planning system in incomplete knowable environment,which ensures the compatibility of different perception and execution devices and meet the communication needs between distributed nodes.Web Ontology Language(OWL)is introduced to formalize the representation of indoor common sense,environment information and robot behavior.The semantic knowledge base module and historical knowledge base module of service robot are designed and implemented to realize the storage,sharing and utilization of the above knowledge.In view of environment information,environmental hypothesis based on common sense is proposed and its credibility is calculated based on historical knowledge.The deterministic reasoning service based on Semantic Web Rule Language(SWRL)rules and the probabilistic reasoning service based on Markov Logic Network(MLNs)are designed,and the function of task autonomous reasoning and room semantic type reasoning was realized on this base.An environmental hypothesis action is proposed to model and represent the environmental hypothesis,and the task planning problem for service robot in the incomplete knowable environment was modeled as Markov Decision Process(MDPs)problem.An extended Semantic Action Model(SAMs)is proposed to realize the semantic description and storage of behavior and skills of service robots.The Probabilistic Planning Domain Definition Language(PPDDL)is used to describe the task planning problem,and then the Labled Real Time Dynmic Programming(LRTDP)algorithm is used to solve the optimal action of the service robot in the current state.The dispatch node and update node are developed to realize the dispatch of action and update of the state.Finally,an exception feedback processing framework is proposed and implemented to deal with possible exceptions in the process of planning,dispatching and execution,which improves the stability of the system.Finally,a demonstration experiments based on the robot simulation software Gazebo is designed,and a simulated indoor four-room environment is built.The feasibility and effectiveness of this system is verified through indoor services,such as object delivery,object search.