Generating Situational Awareness of Pedestrian and Vehicular Movement in Urban Areas Using IoT Data Streams

Abstract

Humans have continuously endeavored to enhance their sensory perception and awareness of the physical surroundings for the betterment of themselves as individuals as well as communities. Advancing this notion to the present day, the Internet of Things (IoT) provides a unique opportunity to attempt the same in an increasingly digital landscape. The prevalence and pervasiveness of IoT data streams gives us this ability to represent a situation with clarity and nuance, leading to a refined awareness of the environment. However, there are several challenges in managing the scale, velocity, and magnitude of IoT data streams, as well as the cohesive representation of these varied sources in a single frame of reference. In this article, we present a new algorithm, the deep growing self-organizing map (deep GSOM) algorithm that addresses these challenges. Deep GSOM incrementally generates a latent representation of situational awareness from high entropy to low entropy IoT data streams. It utilizes an implementation of the fuzzy integral to define a metric that can be moved across spatial and temporal situations to profile the density of congestion. We have also expanded deep GSOM into an IoT platform that can collate, aggregate, and process an entire network of IoT data streams. We demonstrate the workings of deep GSOM on the real-life scenario of profiling vehicular and pedestrian movement using IoT data streams of two highly urbanized cities. The results of these experiments confirm the validity and effectiveness of the proposed approach for generating situational awareness from multiple IoT data streams.