CrowdWiFi: Efficient Crowdsensing of Roadside WiFi Networks


Roadside WiFi networks are increasingly being tapped into by end users with WiFi interfaces in vehicular networks opportunistically for a broad range of applications including ad hoc data dissemination and low-cost Internet access. These networks use fixed access points (APs) that provide improved higher bandwidth connectivity due to better signal propagation characteristics and their ability to exploit spare spectrum. This is especially the case in locations with limited cellular coverage or in environments vulnerable to the obstruction of satellite signals by buildings and is typical in both urban environments (with significant built infrastructure) and in rural areas (where cellular connectivity may be sparse). To support smooth continuous Internet operation in the presence of dynamics caused by vehicle mobility, the design of a middleware that supports accurate real-time identification of roadside APs presents unique opportunities and challenges.

To address the above challenges, CrowdWiFi, a crowdsensing middleware specifically designed for vehicular networks, has been proposed.

It consists of two major components to enable efficient lookup on roadside WiFi networks: an online compressive sensing component and an offline crowdsourcing component. The online compressive sensing component running at the vehicle end coarsely counts and localizes nearby APs in real-time while driving, using sparse signal collection capabilities. The offline crowdsourcing component running at the server end assigns online compressive sensing tasks to some mobile vehicles, then aggregates the online sensing results uploaded by these vehicles, and produces a fine-grained estimation of AP distribution. We exploit the use of compressive sensing (CS) techniques to reduce complexity for in-network localization algorithms which require a large number of RSS (Received Signal Strength) readings. The AP lookup tasks are assigned to crowd-vehicles using geographical participation to gain information from aggregated answers.

The aggregation problem is then transformed into an iterative inference problem on the graphical model to obtain the reliability of each crowdvehicle. 

As the first mobile middleware using the concept of crowdsensing to localize roadside APs in vehicular networks, CrowdWiFi provides multiple benefits in roadside WiFi networks. For example, in conditions where a popular AP is congested, the mobile vehicle can switch to other candidate APs in its communication range. It also helps understand the topologies and network characteristics of large scale WiFi networks, e.g. network density, connectivity, interference properties, etc, in urban areas. Furthermore, the lookup of APs may reveal interesting social aspects of vehicular networks so that mobile vehicles can be involved in location based services and mobile cloud support.