Knowledge-Enabled Task Execution And Service Application For Networked Robotic Systems

It is an effective shortcut to improve the intelligent level of service robots by using knowledge to make decision of using auxiliary resources in the environment autonomously when completing complex navigation and operation tasks. Based on the robot and intelligent environment combined network robot system, the task decision-making and execution approach based on knowledge representation and processing is presented, and typical service applications are developed.Firstly, after analyzing the demands of task execution in the network environment for service robots, the structureof the knowledge based robot system is presented.Secondly, on the basis of common sense knowledge provided by KnowRob, the description model of the service and requirement is developed, and a coarse-to-fine task decompositionmethod is proposed. In the step of coarse decomposition, natural language described tasks are translated into plans in the standard CRAM Plan Laguage. In the step of fine decomposition, each plan will be decomposed into a series of service primitives autonomously by reasoning about the implicit information of the task, searching and synthesizing auxiliary resources, and replanning dynamically. This solution solves the problem when the task is difficult to complete relying on the limited resources of the individual robot. Thirdly, the fuctions of plan generation, assignment and execution are realized using the ROS platform.Finally, three typical perception and operating modules are developed, including laser based localization and navigation module, vision based interactive object recognition module and simple grabbing fuction module. In addition, the encapsulationsof the corresponding modules are also described. The feasibility of knowledge based task execution for a single robot is verified by simulation. For the typical applications of home-care robots, a typical scenario for knowledge based network robot task execution is presented, and the experimental results validated the system design from the aspect of the complex task decomposition and decision-making of using auxiliary resources.