C-ResPECT

Collective Resilient Unattended Smart Things (CRUST) meet Resilient Power-Constrained Embedded ommunication Terminals (ResPECT)

Titel des Gesamtprojektes: SPP 2378 Resilient Worlds Projektleitung: Dr. Peter Wägemann
Projektbeteiligte: Thomas Preisner
Projektstart: 1. Januar 2026
Projektende: 31. Dezember 2028
Akronym: C-ResPECT
Mittelgeber: Deutsche Forschungsgemeinschaft (DFG)

Abstract

Resilient Worlds; the topic of the focus program pays respect to the fact that resilience in the era of ubiquitous connectivity of all kinds of devices has many facets and forms many worlds. One very important of such worlds is the world of sensor/actuator networks composed of resource-constrained, embedded devices that need to operate and communicate in an unattended and unsupervised manner and might even need to harvest their energy from external energy sources like light or electro-magnetic waves. C-ResPECT brings together two projects from phase I and therewith two facets of resilience, implying significant evolution in the holistic system design and operation: The CRUST distributed application managers will embrace the transactional paradigm developed in ResPECT, decomposing tasks into transactions and transaction chains. Those transactions will then be weighted so that devices based on the current and historical set of weights they see can learn to decide which transaction or transactions to fire at each point in time. The devices themselves will, on the one hand, make their energy-budget handling more mature by complementing the worst-case energy consumption with factual energy consumption observations and, on the other hand, will offer a set of decision criteria. The application management will guide the devices following these decision criteria (the number of transactions that can be carried out could be maximized, the completion of a task could be prioritized or the age of information could have the highest preference). Since resilience basically stems from (exploitable) redundancy, C-ResPECT will additionally develop a holistic anomaly detection: The devices will observe the behavior of components with respect to their timing and energy consumption and will detect anomalies and faults in their own parts, while the collection-based view can detect operational anomalies (e.g., an occluded lens of a camera turns it useless, even though it is physically working well). The set of detected anomalies can then be used to guide the decision process by modifying the weights for the transactions. As a side effect, also the absence of anomalies is important information, since it can be used to guide devices to put some of their sensors into standby and hence enable energy optimization without undermining the overall system functionality. C-ResPECT will hence develop resilient collection-based task management based on the model of transactions and tasks and the shared view onto a set of current and historical weights of such transactions and tasks. Transactions and weights have to be stored in a format that enables highly efficient, incremental storage in non-volatile memory as well as incremental communication.

Publikationen