Multi-Agent Adaptive Architecture for Flexible Distributed Real-Time Systems

Abstract

Recent critical embedded systems become more and more complex and usually react to their environment that requires to amend their behaviors by applying run-time reconfiguration scenarios. A system is defined in this paper as a set of networked devices, where each of which has its own operating system, a processor to execute related periodic software tasks, and a local battery. A reconfiguration is any operation allowing the addition-removal-update of tasks to adapt the device and the whole system to its environment. It may be a reaction to a fault or even optimization of the system functional behavior. Nevertheless, such scenario can cause the violation of real-time or energy constraints, which is considered as a critical run-time problem. We propose a multi-agent adaptive architecture to handle dynamic reconfigurations and ensure the correct execution of the concurrent real-time distributed tasks under energy constraints. The proposed architecture integrates a centralized scheduler agent (ScA) which is the common decision making element for the scheduling problem. It is able to carry out the required run-time solutions based on operation research techniques and mathematical tools for the system’s feasibility. This architecture assigns also a reconfiguration agent (RA <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">p</sub> ) to each device <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$p$ </tex-math></inline-formula> to control and handle the local reconfiguration scenarios under the instructions of ScA. A token-based protocol is defined in this case for the coordination between the different distributed agents in order to guarantee the whole system’s feasibility under energy constraints.