Learning Approach for Smart Self-Adaptive Cyber-Physical Systems

Abstract

Modern Cyber-Physical Systems (CPSs) need to be able to operate efficiently and reliably within continually changing, uncertain, and unanticipated environments. Namely, these systems should be capable of learning, automatically reconfiguring themselves, and be able to cooperate and collaborate with other CPSs. In a nutshell, exhibit human-like, smart capabilities in an autonomic manner. However, engineering such systems is all but trivial, primarily because we need to develop systems at design-time that are capable of autonomously coping with the uncertainty and change at runtime. Therefore, not only the importance of self-adaptivity as a system's feature increases, but it becomes a fundamental approach for the systems to continue meeting their functional specifications, fulfilling their business objectives while preserving the performance-despite all the runtime changes that the system may encounter. To tackle these challenges, this paper proposes the initial research vision and agenda with the envisioned contributions towards an approach for self-adaptation of cooperative, smart CPSs through shared knowledge and learning.

Function Modeling for Collaborative Embedded Systems

“…Trigger events Adaptivity to the internal and external events in order for the systems to continue meeting their functional specifications while preserving the performance (or another quality objective) -despite all the changes that the system may encounter during runtime [Petrovska and Pretschner 2019].…”

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

“…As modern society increasingly relies on CPSs, these systems must act reliably even when faced with different run-time uncertainties. For instance, internal system uncertainties, i. e., different software or hardware faults and failures, as well as external uncertainties, e. g., uncertain and changing environments or execution contexts [21,24,28] in which the systems operate. Further, due to the limited range of their sensors, CPSs can only observe their context partially, which introduces another uncertainty dimension in the systems.…”

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

“…Their close connection to the physical world means that they are exposed to high uncertainty during their operation. Namely, modern CPSs need to be able to operate efficiently and reliably within a continually changing, uncertain, and unanticipated environment (execution context) [25,27,29]. The context is the relevant part of the environment for a particular system.…”