Energy conscious online architecture adaptation for varying latency constraints in sensor network applications

Sensor network applications face continuously changing environments, which impose varying processing loads on the sensor node. This paper presents an online control method which adapts the architecture to minimize energy consumption while satisfying varying latency constraints. The method predicts processing load requirements over a finite time window and accordingly adapts the architecture. The behaviour of the hardware modules over time has been approximated with a Continuous Time Markov Process. Adaptive image processing for vehicle tracking was used as a case study for this approach.