Handset/Tablet Battery Characterization, Modeling, and Runtime Estimation
Andrew Chemistruck and Ahmad Bahai
Key figures of merit need to be tracked for proper operation of handset and tablet systems. This is especially important when the battery is nearing end of discharge and the user would like to use power-intensive features such as GSM or camera flash. The goal of this project is to examine the performance of traditional state of charge algorithms against optimal estimators and machine learning techniques. The data that is collected over many months will be used to examine adaptive estimation techniques to maintain good accuracy on state of charge estimates. Usage profiling will be examined to determine if there are any machine learning techniques that improve the fidelity of the runtime estimation and perhaps add error bounds using interval calculus techniques.