Knowledge aggregation with subjective logic in multi-agent self-adaptive cyber-physical systems

Abstract

Modern software systems, such as cyber-physical systems (CPSs), operate in complex and dynamic environments. With the continuous and unanticipated change in the operational environment, these systems are subjected to a variety of uncertainties. Self-adaptive CPSs (SACPSs) can adjust their behavior or structure at run-time as a response to the changes in their perceived environment. Namely, self-adaptation is commonly realized through a MAPE-K feedback loop incorporating newly derived knowledge obtained by the sensed data from the run-time monitoring, during the operation of decentralized SACPSs. However, to build the knowledge, the need for run-time observations' aggregation and reasoning emerges, since the observations made by the decentralized systems might be conflicting. In this paper, we propose an approach for observations aggregation and knowledge modeling in SACPSs that is domain-independent and can deal with inaccurate, partial, and conflicting observations, based on the formalisms of Subjective Logic.

Reliability-Based Aggregation of Heterogeneous Knowledge to Assist Operators in Manufacturing

“…According to Ramirez et al [11] the most relevant sources of run-time uncertainties are related to interactions of the system and its context. To address these, the application of subjective logic to aggregate run-time observations into actionable knowledge has been proposed by Petrovska et al [12]. However, since their methodology is centred on cyber-physical systems and ours on human operators it is not directly transferable.…”

Run-time Reasoning from Uncertain Observations with Subjective Logic in Multi-Agent Self-Adaptive Cyber-Physical Systems

“…The complex and heterogeneous nature of CPSs often requires the SACPS to consist of multiple MAPE-K loops [25,27,41]. The use of multiple, interconnected MAPE-K loops leads to numerous challenges, one of which is the coordination of the control loops [39].…”

“…Contribution. Building on our previous work in which we introduced the idea of knowledge aggregation via Subjective Logic [25], in this paper we detail and extensively evaluate our fully developed approach. Succinctly, we make the following contributions:…”

Towards a Self-Adaptive Architecture for Federated Learning of Industrial Automation Systems

“…In line with this, our research targets multi-agent industrial systems, where the self-adaption of agents leads to collaboration and utilizes decentralized optimization to solve con- strained decision-making. A related approach was presented in [17], where the decentralized aggregation process for shared knowledge resolves decision conflicts through subjective logic.…”