Modelling and designing dependable autonomous systems
Time: Tue 2020-04-21 13.15
Participating: Elena Troubitsyna, Division of Theoretical Computer Science
Autonomous systems are increasingly integrated in our everyday life. Autonomous vehicles of all kinds, variety of servo-, industrial and medical robots perform a wide spectrum of tasks alongside humans. It raises a serious concern regarding the risks associated with blending the autonomous technologies into the safety-critical activities and dependability of the resulting systems.
Dependability is an ability of a system to deliver its services in a trustworthy way. To achieve dependability, we should ensure that the likelihood of occurrence and severity of all hazards associated with the system have been sufficiently reduced, i.e., the risk associated with the system is acceptable. Currently, risk assessment and mitigation are typically performed at the design time. They focus on identifying hazards associated with the system and verifying that the system design incorporates the adequate means for preventing or mitigating hazards. However, such an approach assumes that the hazards are known, i.e., the system operates in a predictable environment according to the rules and constraints defined at the design time. We argue that such an approach is inadequate for ensuring dependability of future autonomous systems. Indeed, firstly, an autonomy of a system implies that it will evolve and change its behaviour according to the emerging goals and operating conditions. Secondly, the richness of behaviour of autonomous systems means that the system will operate in the environment that is characterised by a high degree of uncertainty and unpredictability. Therefore, endowing the autonomous systems with the capability to recognise, assess and mitigate unpredictable dynamically emerging risks constitutes an important research goal.
In my talk, I will present an integrated modelling and design framework for the development of dependable autonomous systems capable of detecting and mitigating risks at run-time. The framework combines formal modelling and verification of system architecture and its properties as well as the algorithms enabling run-time adaptation to the dynamically changing operating conditions.