Sleep science informed algorithms for driving extreme runtime in wearables

We propose to develop sleep-science informed algorithms for optimizing the transmission of collected data from wearables to data collection hubs, aimed at minimizing power usage in the wearable devices. Specific techniques will include (i) summarization of collected data into sleep parameters relevant to fatigue assessment, which is a sort of ‘domain-specific semantic data compression’, (ii) adaptive data transmission frequency from wearable to hub, based on sleep stage of the person, and (iii) lazy data send from wearable to the hub, and thus improving payload to the overhead ratio of data packets, based on sleep assessment needs. In addition, we can also reduce the energy consumption of data collection by (iv) adaptive sampling of the data time series with highly predictable values, and (v) dynamic sensing to collect data during time-periods of interest.