The paper proposes a self-adaptation approach for improving system resiliency in cyber-physical systems (CPS) through the automated coordination of graceful degradation and recovery. The key idea is to treat degradation and recovery as requirement-driven adaptation tasks, where degradation involves temporarily weakening the original system requirements, and recovery involves strengthening the weakened requirements when the environment returns to an expected state.
The paper first provides an overview of the proposed runtime adaptation architecture, which consists of three main components: an event detector, a requirement evaluator, and a degradation and recovery planner. The event detector looks for degradation or restoration events in the environment, the requirement evaluator determines the achievable requirement based on the current environmental conditions, and the planner generates future system actions based on the changing requirements.
The paper then presents an extension to Parametric Signal Temporal Logic (PSTL) to formally capture the concepts of requirement weakening and strengthening. It introduces the notions of minimal, optimal, and current requirements, and defines metrics to quantify the degree of weakening and strengthening between different PSTL instantiations. Finally, the paper formulates the degradation and recovery problems as instances of Mixed-Integer Linear Programming (MILP), where the objectives are to minimize the degree of weakening and maximize the degree of strengthening, respectively.
The proposed approach is implemented and evaluated using a case study involving an unmanned underwater vehicle (UUV) that must maintain a clear line of sight with an underwater pipeline and provide sufficient thrust to complete its mission. The results show that the requirement-driven adaptation framework can achieve a higher level of requirement satisfaction throughout the adaptation process compared to a state-of-the-art approach, while incurring reasonable runtime overhead.
To Another Language
from source content
arxiv.org
Djupare frågor