A multi-agent system for optimizing physiological collection based on adaptive strategies

Abstract Wearable devices have emerged from the evolution of communication and information technology, along with the miniaturization of electronic components. These devices monitor the user's physiology on a periodic basis and generally have low battery autonomy. This article proposes the Odin system, a multi-agent system that performs real-time analysis of physiological data in order to adapt the collection according to the user's health condition. Two types of adaptation were proposed in terms of adapting the collection. The first, wait-time adaptation (WTA), consists of collecting physiological data through a sensor defined as the main sensor, where the frequency of the collection is based on the user's condition. Meanwhile, the second, paused sensors adaptation (PSA), manages the activation status of secondary sensors when the main sensor suggests a possible risk to the user's health. An evaluation was implemented through two steps. First, a simulation was performed with requests control to optimize the parameters of the data collection. These results showed an increase of 214% in battery life in the adaptive scenario compared to one with no adaptivity. Second, Odin's adaptivity was evaluated in terms of a real dataset, which allowed for the reduction of the number of requests. This reduction optimized the battery consumption, with a 66% improvement over data collection with no adaptivity.