A middleware for cooperating mobile embedded systems

The cooperation of mobile embedded systems gives rise to new and promising applications in fields such as industrial automation, logistics, telematics, and team robotics. Such applications are subject to real-time constraints due to the physical interactions between the mobile systems and their environment. Achieving a timely predictable cooperation, however, is a challenging task. The execution times of the tasks and the reliability of the communication links depend on the dynamically changing environment and are hence hard to predict. Moreover, there is an inherent tradeoff between an efficient cooperation and the autonomy of the mobile systems. Managing the interdependencies between the actions of the mobile systems may result in complex runtime interactions, thus rendering the timing behavior of the application hard to analyze and predict, even if the underling services are timely. In this thesis, we present a middleware supporting the application designers in overcoming the abovementioned challenges. On the two lower layers, it provides timely predictable multicast communication and task execution. Both service are not based on worst-case assumptions. To resolve the cooperation / autonomy tradeoff and achieve a timely predictable behavior on the level of the cooperative application too, the two upper layers of the middleware provide common views to the application. Based on these common views, the mobile systems decide about their actions locally so that runtime coordination in the application can be avoided. Common views are available for application-independent aspects of the control system, like group membership, as well as for more the application-specific aspects of the controlled system, such as position and speed of the mobile systems. Two application scenarios guided the design of the middleware and have been implemented as prototypes. In the first scenario, a group of mobile systems coordinate their velocities