MORO:a Self-Adaptive Software Framework For Heterogeneous Mobile Robots

In recent years, the rapid adoption of mobile robots in various domains generates tremendous demands for self-adaptive software for robot applications. However, from a software engineering perspective, the development of such software systems is suf-fered from following two problems:First, software applications are closely dependent on the underlying mobile robot platform, and the porting of software applications to other heterogeneous platforms is extremely painful, especially when QoS issues have to be considered. Second, the self-adaption logic, the sensing and control logic and the business logic are tightly coupled together, which brings unnecessary complexity and makes the reuse of self-adaptation logic impossible.To address the two problems, this thesis proposes a self-adaptive software frame-work to hide the heterogeneity of different mobile robot platforms and facilitate the de-velopment of self-adaptive applications with a rule-based self-adaptation model. The framework clearly separates adaptation logic from sensing and control logic and appli-cation business logic. Elaborately, the main contributions of this work are as follows:Masking the differences of platforms in software and hardware. The framework uses a universal mobile robot model and defines a set of unified driver interfaces to mask the differences in robot hardware platforms. Based on the Linux kernel and the leJOS JVM, the framework also provides a unified programming envi-ronment for application development. In addition, the framework incorporates a self-adaptive mechanism that can ensure same level of QoS across different mobile robot platforms. Decoupling self-adaption logic from business logic with a rule-based model. D-ifferent from traditional rule models, in our model application developers can customize conditions and actions, which makes the framework more flexible.Facilitating application development with two auxiliary tools. One of them is a visual editor for adaptation rules. It creates the adaptation finite-state machine rule model diagrams and automatically detects logic errors, if there are any. The other tool can generate the codes needed by the framework according to user specified adaptation conditions and actions.To evaluate the framework, we developed an environment exploration applica-tion and deployed it on four types of mobile robots with very different software and hardware platforms. Preliminary experimental results confirmed the effectiveness and reliability of the framework.