Dynamic Risk Management for Cooperative Autonomous Medical Cyber-Physical Systems

Medical cyber-physical systems (MCPS) combine independent devices at runtime in order to render new patient monitoring/control functionalities, such as physiological closed loops for controlling drug infusion and optimization of alarms. MCPS and their relevant system contexts are highly variable, which detrimentally affects the application of established safety assurance methodologies. In this paper, we introduce an approach based on dynamic risk assessment and control for MCPS. During runtime, information regarding the safety properties of the constituent systems, relevant information about the patient's characteristics, as well as other relevant context information is utilized to dynamically and continuously optimize the system performance while guaranteeing an acceptable level of safety. We evaluated our approach by means of a patient-controlled analgesia proof-of-concept simulation and sensitivity analysis.