Information-Driven Adaptive Sensing Based on Deep Reinforcement Learning
This addresses energy management for resource-constrained IoT sensors, offering a generalizable solution with reduced human effort, though it appears incremental as it builds on existing deep reinforcement learning methods.
The paper tackled the problem of creating energy-efficient sensing policies for IoT devices by introducing a novel reward function based on Fisher information in deep reinforcement learning, resulting in learned behavior that outperforms uniform sampling and approaches near-optimal performance in a noise monitoring scenario.
In order to make better use of deep reinforcement learning in the creation of sensing policies for resource-constrained IoT devices, we present and study a novel reward function based on the Fisher information value. This reward function enables IoT sensor devices to learn to spend available energy on measurements at otherwise unpredictable moments, while conserving energy at times when measurements would provide little new information. This is a highly general approach, which allows for a wide range of use cases without significant human design effort or hyper-parameter tuning. We illustrate the approach in a scenario of workplace noise monitoring, where results show that the learned behavior outperforms a uniform sampling strategy and comes close to a near-optimal oracle solution.