Fault- Tolerance Support for Mobile Robotic Applications

摘要

The advent of powerful yet also affordable mobile robotic platforms will spur a new generation of computer-driven applications that employ multiple mobile robots to perform a large variety of missions. This prospect has lead to the emergence of corresponding programming models that aim to ease the development of such applications, allowing a more straightforward exploitation of the resources that are available on these platforms. But fault-tolerance remains a challenging issue due to the inherent dynamics of such cyber-physical systems. Notably, some of the actions that are performed in the context of such executions cannot be easily undone, while in other cases it may be problematic to automatically redo certain actions without the explicit control of the application. The desired functionality cannot be achieved by applying traditional fault-tolerance schemes out-of-the-box; these need to be combined and extended in a suitable manner. In this paper, we present system-level support designed to provide transparent fault-tolerance for applications that coordinate multiple mobile robots in a centralized way. We discuss a concrete implementation of the proposed approach as part of an existing programming framework. Also, we present an evaluation of the respective overhead for individual runtime operations as well as for indicative application scenarios with and without failures.