Study And Implementation Of Robot Task Re-planning Mechanism

Robots nowadays are increasingly applied in non-industrial domains to serve human beings in their daily life.Compared to industrial uses,a robot providing non-industrial services is typically situated in a more dynamic environment in which it may encounter unexpected changes that may make its current plan infeasible.Therefore,during task execution,the robot must timely respond to these unexpected issues via some form of re-planning,in which it can reactively adapt its current plan in response to environment changes,in order to improve the chance of successfully accomplishing its task.Current re-planning approaches focus on reactive control-level planning(e.g.,motion or trajectory planning),in which the plan being adapted is normally described as a sequence of low-level instructions directly controlling joint motors and other physical actuators of the robot.Researchers studying task re-planning primarily describes a task as a movement of the robot after confining sequence of tasks,and use methods and techniques of motion re-planning to solve task re-planning.Most of these approaches depend on specific scenarios and mathematical models,and lack sufficient representation of high-level knowledge associated with the goals,resources,and structure of the task.In order to overcome these shortcomings,in this paper,we propose a novel task-level re-planning approach,which uses events to represent the uncertain issues that trigger re-planning,and which describes each task as a sequence of high-level actions.These tasks,according to their functions,are further categorized into manipulation and locomotion tasks.Our main contributions in this paper are summarized as below:(1)We propose ETRM(Event-based Task Re-planning Mechanism)which uses Task Feasibility Evaluation Algorithm(TFEA)to enable runtime task selection.(2)Based on ETRM,we propose a framework called SAPE(Sensor-Actor-Planner-Executor)which is implemented on top of ROS(Robot Operating System).This framework is based on the distributed communication features of ROS,and uses “topic” mechanism of ROS to realize event-flow among multiple functional modules of the framework.This framework also provides two concurrent Actor modules which respectively evaluate the aforementioned two types of actions in order to improve re-planning efficiency.