Improving sleep through closed-loop autotuning of a robotic bed using Gaussian processes

Abstract

Rocking beds that provide vestibular stimulation may be a promising alternative to conventional pharmaceutical treatments that show many side-effects. Past studies have demonstrated that the effectiveness of the vestibular stimulation is influenced by the selected rocking acceleration. Moreover, the movement must be smooth and comfortable to avoid disturbing the user’s sleep. Previously, the tuning of the control parameters was done manually, which was time-consuming and did not guarantee an optimal movement of the bed. In this paper we show an efficient and effective way to automatically tune the control parameters of the bed using Gaussian processes while achieving the desired acceleration trajectory and providing a comfortable movement for the user.