Optimisation of battery operating life considering software tasks and their timing behaviour

Today, batteries serve as a power supply for almost every electronic device which people carry around or install in locations that have no access to the mains. The longer a battery powered system can operate without need for recharge or battery replacement, the more will maintenance cost and the number of faults due to insufficient power supply decrease. This work considers multiple independent software tasks that are assigned to the same processor. The tasks are supposed to have hard deadlines and are to be scheduled earliest deadline first (EDF). Energy consumption is optimised using processor voltage and frequency scaling, per task or per task set, and processor shutdown, periodically or task-dependent. An event stream based hard realtime test ensures the feasibility of the solutions, and it provides low computational complexity for the optimisation problem. Optimal congurations in the sense of a long operating life are determined with the help of a battery model which is developed in this work. The model is evaluated with measurements on real batteries. A positive effect of pulsed discharge was falsified for pulses in the range of 50 µs to 10 ms. Based on the results, a battery model is derived which is as simple as the Peukert-formula and which provides an error of less than 7% for the battery samples which were used in the experiments.