Antoine Omond thesis defense

Abstract : Cyber-physical systems (CPS) deployed in scarce-resource environments like the Arctic Tundra for in-situ and long-term observation face extreme conditions. Nodes deployed on the field are under a limited energy budget. To save energy and increase their lifetimes, nodes are forced to alternate between short uptime periods and long sleeping periods. Uptime schedules depend notably on observation tasks and are not synchronised.
Nodes of the CPS may collaborate to provide services such as data aggregation, analytics, etc. Due to a lack of network infrastructure, nodes have to rely on peer-to-peer connections to communicate. Nodes are able to communicate only during uptime, when their uptime periods overlap. When dealing with short and non-synchronised uptimes, opportunities for communication can be very low.

Due to the Arctic Tundra conditions, nodes of the CPS are forced to be autonomous for long periods (from 6 months to a year). During these periods, nodes of the CPS have to autonomously adapt to external events by reconfiguring their systems. Due to collaboration, reconfiguration of collaborative nodes must be coordinated.

Due to scarce connectivity within the CPS and its isolation from external systems, a central authority is most of the time unreachable. Nodes must be able to coordinate their reconfiguration with other nodes in a decentralised manner. Due to scarce connectivity, coordinating such reconfiguration between nodes can impose a significant energy consumption overhead on nodes and take a long time to converge.

This manuscript proposes to study the decentralised reconfiguration of a Cyber-Physical system with sleeping nodes. It provides a solution to define and coordinate the execution of decentralised reconfiguration programs between sleeping nodes. Based on this solution, it aims at extensively studying the performances in terms of energy consumption and duration of such reconfiguration according to the Arctic Tundra characteristics.