Coherence and Feasibility of Real-Time Software Tasks in Networked Adaptive Systems

This paper deals with the dynamic reconfiguration of networked devices linked by a controller area network (CAN). Each device runs dependent periodic and aperiodic software tasks and can be adapted to any evolution in its environment. A reconfiguration is a dynamic scenario that activates-deactivates the deployed devices, adds-removes software tasks, or changes the network traffic according to user requirements. Nevertheless, such a scenario can trigger the execution of new-old tasks to violate real-time deadlines or to possibly increase the energy consumption. Moreover, a reconfiguration that adapts dependent tasks in different devices can modify the network traffic and some deadlines of frames can be violated too. To resolve all these problems that can happen after concurrent distributed reconfiguration scenarios, we propose a dynamic methodology called Cynapsys-reconfigurable control system that allows coherent distributed behaviors of devices after any scenario. This run-time automatic strategy based on a multi-agent architecture is achieved in five steps: 1) applying reconfiguration scenarios on the system devices to update their services; 2) coordination between devices after any reconfiguration for their coherence; 3) feasibility analysis of each reconfigured device; 4) verification of CAN feasibility; and 5) reconfigurable frame packing. A developed tool is applied to a case study for the evaluation of the proposed contribution.