An Innovative Dynamic Mobility Sampling Scheme Based on Multi-Resolution Wavelet Analysis in IoT Networks

Mobility is a key aspect of modern networking systems. To determine how to better manage available resources, many architectures aim to a priori know the future positions of mobile nodes. This can be determined, for example, from mobile sensors in a smart city environment or wearable devices carried by pedestrians. If we consider infrastructure networks, frequently changing the coverage cell may lead to service disruptions if a predictive approach is not deployed in the system. All predictive systems are based on the storage of old mobility samples to adequately train the model. Our focus is based on the possibility to determine an approach for adaptively sampling mobility patterns based on the intrinsic features of human/node behaviour. Several works in the literature examine mobility prediction mobile networks, but all of them are dedicated to the study of time features in mobility traces: none took into account the spectral content of historical mobility patterns for predictive purposes. By contrast, we take into account this spectral content in mobility samples. Through a set of wavelet transforms, we adapted the sampling frequency dynamically and obtained a considerable set of advantages (space, energy, accuracy, etc.). In fact, this issue covers an important role in the IoT paradigm, where energy consumption is one of the main variables requiring optimization (frequent and unnecessary mobility samplings can disrupt battery life). We performed several simulations using real-world traces to confirm the merit of our proposal.