An edge computing-based monitoring framework for situation-aware embedded real-time systems

An embedded real-time system (ERTS) needs to provide continuous service in various dynamic situations. Such service requirements in different situations create the need for the ERTS to monitor its environment at run time, gather knowledge of its situations, and guarantee its timing and operational behavior through self-adaptation. Recent advances in sensor technologies have introduced cameras, lidar, and radar as powerful monitoring tools. However, processing and storing raw sensor streams require significant storage and computational ability. An ERTS is embedded in nature, and therefore, it has limited storage and processing capacity. This paper considers that the ERTS contains an analytics endpoint (edge node). We present an edge computing-based monitoring framework that characterizes environmental situations at run time by identifying events and their properties. We enable the framework to store and process from a significantly reduced dataset by creating a knowledge base. The framework also allows the ERTS to identify resource, performance, and safety constraints in the edge node for each situation. The framework assists the ERTS in adapting to the situations (if the constraints are satisfied) by determining adaptive tasks that need to be triggered with respect to the environmental events. The experimental analysis shows that the framework present in the edge node assists in situation characterization in terms of the identified events and admission of adaptive tasks. The monitoring framework also allows improvement regarding the probability of failure and average response time. We use the earliest deadline first (EDF) scheduling algorithm with and without considering the edge node and perform a comparative schedulability analysis. We demonstrate that overall demand due to the admission of adaptive tasks and situation-driven analytics exceeds available supply, which can be addressed using the proposed edge computing-based framework.