“Safety Kernel for Cooperative Sensor-Based Systems”
in Safecomp 2013 Workshop on Architecting Safety in Collaborative Mobile Systems (ASCoMS), Toulouse, France, Sept. 2013.
Abstract: Developing smart vehicles, either automobile or aerial, to realise cooperative functionality in open and inherently uncertain environments is a difficult task. One fundamental challenge is to make cooperation predictable and safe, despite the uncertainties affecting the operation. Traditional approaches for the design of safe control systems rely on the possibility of defining safe operational bounds, both in the value and in the temporal domain. Unfortunately, when considering wireless communication networks and varying sources of sensor data, it becomes very hard, or even impossible, to define safe and small enough bounds. To deal with this problem, a possible approach is to consider a hybrid system architecture in which some components may execute with uncertain timeliness, but which also includes some predictable components. In addition, a Safety Kernel implemented in the predictable part of the system will be instrumental to manage the system behaviour and ensure safety. In this paper, we describe the architecture and role of such Safety Kernel in the context of a hybrid system architecture. The Safety Kernel is responsible for monitoring and managing the run time configuration of the system, as needed to avoid hazardous situations. We specify the individual components of the Safety Kernel and how they interact with other components in the system architecture, including the functional components of the control system. Finally we present a high-level description of a concrete implementation based on time and space partitioning.
Research line(s): Timeliness and Adaptation in Dependable Systems (TADS)