Multi-Layered Safety Architecture of Autonomous Systems: Formalising Coordination Perspective

A pressure to deploy autonomous systems in real-life is increasing. Since exhaustive verification of safety of autonomous systems is unfeasible, the emphasis should be put on safety optimisation and run-time safety-monitoring techniques. In this paper, we propose a multi-layered architecture of autonomous systems. We define the notions of strategic, tactic and active safety -- the complementary mechanisms for achieving safety. We take a swarm of drones as an example and formally define a multi-layered safety architecture and associated coordination mechanisms and underlying communication model to implement the defined complementary safety mechanisms. The derived coordination logic and communication model is formalised in Event-B framework.