1. Introduction
Continuous physiological monitoring in hospital settings generates a mix of latency-tolerant logging tasks and latency-critical alert tasks, and a one-size-fits-all offloading policy, whether always-local or always-cloud, fails to jointly optimise for both energy and responsiveness.
2. Methodology
A mixed-integer programming formulation was developed to assign each processing task from 50 simulated wearable devices to either a ward-level fog node or a remote cloud server, minimising a weighted objective of device energy consumption subject to per-task latency deadlines, solved using a genetic algorithm with a population of 80 over 200 generations per scheduling window.
3. Results
The genetic-algorithm-based offloading policy reduced average device energy consumption by 38 percent relative to an always-cloud baseline, while meeting the 200-millisecond deadline for latency-critical alert tasks in 99.2 percent of simulated cases, compared with 91.4 percent for a static threshold-based offloading rule.
4. Conclusion
Dynamic, optimisation-based task offloading between fog and cloud tiers can substantially extend wearable device battery life without compromising the responsiveness of safety-critical alerts. Future work will validate the approach against real hospital network traces.
References
[1] Bonomi F. et al., Fog computing and its role in the internet of things, MCC Workshop, 2012. [2] Mahmud R. et al., Fog computing: A taxonomy, survey and future directions, Springer, 2018.