Fault-tolerant real-time architecture for elderly care

The ongoing transition from traditional elderly care to the use of modern technologies from the field of Cyber Physical Systems (CPSs) results in new challenges for both industry and research. This shift is mainly motivated by the increasing share of elderly people in the population which is causing a notable shortage of nursing staff. With the availability of new technologies, the CPSs for elderly care are also enabling new fields of applications in the area of biomedicine and robotics. Use cases like the automatic injection of insulin and robotic assistance are prominent examples for these application fields. These new application fields impose new requirements on architectures in the field of elderly care, such as deterministic real-time behavior and dependability along with an open-world assumption in which dynamic changes within the composition of the system can occur at run-time. Likewise, the application of robotic systems in the field of elderly care introduces stringent real-time requirements to the whole CPS, affecting the integration of complex and heterogeneous sensors, the control of actuators and the communication network. Moreover, the application of fault-tolerance and mixed-criticality techniques is required to establish a dependable CPS that is able to tolerate faults in order to prevent dangerous situations for human life. Furthermore, CPSs have to encompass different integration levels like the local network and the Internet in order to support services from professional stakeholders like medical services from caregivers or a doctor.
The proposed architecture for elderly care takes into account the new emerging application fields in elderly care as well as the associated challenges, which are (1) real-time support, (2) dependability and (3) support of an open-world assumption while taking into account multiple integration levels and the heterogeneity of the underlying technologies. A review of state-of-the-art architectures for elderly care shows that there is no architecture available at present that meets all these challenges. The proposed architecture addresses this gap by taking advantage of a broad range of well-known technologies and standards from the state-of-the-art like ISO/IEEE 11073 and Time Sensitive Networking (TSN) while further introducing new concepts and technologies, such as fault containment among containers for high-critical applications as well as real-time container-to-container communication with latencies and jitter in the low microsecond range. A huge challenge is further to address the open-world assumption while providing real-time guarantees and fault-tolerance. In particular, this puts further requirements to the real-time system like the capability of a dynamic rescheduling of real-time resources like the real-time network. This is addressed by the introduction of a service for the dynamic rescheduling of real-time communication resources that takes care about topology and service management, scheduling, configuration building and distribution of communication schedules. By this way, changes within the physical model (e.g. a new network switch or end system) and the logical model (e.g. a new service) are supported at run-time of the system.
In the field of software architectures, the use of microservices has reached a strong technical maturity in recent years. The proposed architecture is embracing this trend and introduces platform services as microservices. Finally, several proof-of-concept implementations are presented and evaluated in different experiments ranging from a real scenario to experiments in a laboratory in order to show that the proposed architecture for elderly care is able to address the shift in traditional elderly care to the use of modern technologies from the field of CPSs.