Office address: XUEFU BLVD 999, Information Engineering College of Nanchang university, Room A114
Major: Internet of Thing
Current research work: Software defined battery
Battery capacity limits the ubiquitousness of battery powered devices, e.g., smartphones. Being tired of the slow speed of battery evolution, system researchers proposed to build
hybrid batteries to match various power demands from smartphone applications, such that phone
users can enjoy longer battery life and better use experience. However, tuning hybrid battery
is a challenge since software demands different power in different execution stage: Failing to
meet power surge may cause performance degradation, and keeping high power output is a waste.
In this paper, we propose a control loop feedback mechanism to address this challenge,
called adaptive realtime multi-battery performance control, or ARMPC. ARMPC explores the
possible power demand pattern from applications, and learns battery feature ofﬂine. Using
this information, ARMPC models the system and adopts a PI (or maybe MISO) controller design.
We mathematically prove the feasibility (i.e., zero steady-state error, ﬁxed settling time)
and stability (e.g., short rise time and small overshoot) of ARMPC using classic control theory.
We have prototyped ARMPC in modern Android phones and evaluated with several state-ofthe-arts.
Our results show that in practice, ARMPC can prolong the battery life cycle, signiﬁcantly reduce
average power use, and meets all power surge demands, as compared to state-of-the-arts.
We simulate ARMPC with many more different batteries. We illustrate that for hybrid battery management,
classic controller is the good option at the tradeoff between overhead and performance.