Runtime Performance Verification And Abnormality Analysis Of Robot Software

With the continuous development of robotic applications and robot-relative disciplines,robot software is facing more and more uncertain working environment, and dealing withmore and more sophisticate task. Some robot task is strict on the performance of its software,for example, it needs high performance for a patient-care robot to take care of patient, or itmay delay the patient’s condition, even lead to patient’s death. Sometimes people cannotdiscover the abnormal performance of robotic software at runtime and make effectiveintervention in time. The factors make it urgent to implement the runtime self-adaptation onperformance of robot software.This paper discusses the solution to the above problems taking MiniROS as the object ofstudy. MiniROS is a robot middleware platform, which is developed by Intelligent Softwareand Robot Lab of South China University of Technology. Several robot components which arebased behaviors could be assembled into a complete robot software system by MiniROS. Thispaper aims at resolving the runtime verification, abnormality analysis and selecting theappropriate adjustment solution of robot software assembled by MiniROS according to theanalysis results, at last achieving the robot software runtime self-adaption on performance.The main research works are as follow:Firstly, this paper proposes a model named Modularity Probability Transition PetriNetwork, to establish runtime model for runtime system. In other words, each behavior groupwhich may be selected to run in the runtime system is modeled based on ProbabilityTransition Petri Net.Secondly, this paper explains the definition of context in runtime system, analyzes thecontext model and implements monitoring the context. Through the research and analysis, thispaper represents the context based on key-value model for its simplicity, and monitors theruntime context based on aspect oriented programming, to reduce the coupling degreebetween runtime context monitor module and the function module of runtime system. Andthen verifies the context which related to performance with the expected value in the runtimemodel.Thirdly, this paper makes link analysis for transitions in runtime model, which is based on the dependent degree of atomic behavior. Defines the importance of transitions and setspriority for transitions to be analyzed. And then analyzes the abnormal runtime model basedon the model’s structure and property, finds out the abnormal cause, and reason to selectappropriate solution for the abnormality based on rule knowledge base. And at last guides theruntime system to adjust itself with the solution.Finally, this paper takes robot soccer as example, to establish runtime model for soccerrobot software, and verify the behavior groups’ running time, analyze the abnormal module,and reason to select the appropriate solution. This example concludes the solution in thispaper can realize the performance runtime self-adaptation of robot software which isassembled by MiniROS.