Component Behavior Selection And Healing For Robotic Software

With the development of robot-relative disciplines and the continuous expanding ofrobotic applications, robot software is facing a variety of sensors and actuators, more andmore uncertain working environment and sophisticate task. Besides, robot software is desiredto meet some important non-functional requirements, such as adaptivity, reliability andreal-time. These factors make it difficult to develop robot software.In the long-term development of robot software engineering history, people realize thatsoftware reuse is an important way to increase the quality and efficiency of software.Component-base software architecture has been proved to a good way for robot software.However, most of the robot software nowadays is only designed for specific robot, whichmakes it hard to be reused. And the component-based software architecture doesn’t considerthe characteristic of robotics.In this thesis, firstly, this paper proposes a behavior-based component model for robotsoftware. In this model, behavior is the basic unit in component organizing. A componentconsists of several behaviors, and a composite behavior can be formed with a few componentsand their connections. It makes software structure hierarchical, high cohesion and lowcoupling, which can well represent the robot software.Second, this paper proposes the idea of behavior selection in component, and designs aflexible framework for it. According to different environments and tasks, component in robotsoftware is required to select the proper behaviors adaptively. The selection module in acomponent monitor the environment and the component itself, and calculate the contextrepresentation using predicate calculus, then resolve the behavior conflicts and activate theproper behaviors. In this thesis, this paper implements two selection strategies based onpriority and link analysis.Third, this paper analyzes the software self-healing problem, and proposes a healingframework for component behaviors, which includes exception detection and healing methods. It can improve the reliability and availability of robot software. Specially, this paper utilizescomponent configuration and operation logging to track exceptions, and handles them throughseparating, user handling and component restarting.Finally, this paper introduces the key technology for implementing the componentbehavior model. Then it takes robot soccer for example, to verify the utility of componentbehavior model, behavior selection and healing, and concludes that the system in this paperfairly has the characteristic of self adaptivity and reliability.