Wireless mobile devices such as cell phones, PDAs, and Wi-Fi laptops become ubiquitous in our daily lives and guide us into the era of pervasive computing. For instance, clothes and cars equipped with such devices are going to seamlessly give us helpful information when we are traveling a new city or shopping. Not only do such devices enrich our daily activities, but also create an environment such that epidemics of cooperation can thrive, e.g., among rescue workers or pedestrians, and thus people can cooperate in ways that were never possible before. Futurist Howard Rheingold first named these kinds of activities as "smart mobs," where people with shared interests/goals can pervasively cooperate using wireless mobile devices.
Reflecting on tragedies such as 9/11 and London Bombing, we envision that such smart mobs may actually help relieve losses or investigate accidents if they were properly organized beforehand. For example, in the London Bombing, the police were able to track some of the suspects in the underground using closed-circuit TV cameras in the subway, but they had a hard time finding helpful evidence from double-decker bus, in spite of the abundance of pictures taken by shutterbugs using their camera phones. This incident convinced the British police to install more cameras that can read license plates to track vehicles. Yet in such a scenario, a smart mob approach is more desirable than an extensive deployment of cameras/sensors: in fact, the vision of smart mobs that exploit completely distributed epidemic cooperation makes it hard for potential attackers to disable surveillance. Let us briefly note that people are willing to sacrifice privacy and to accept a reasonable level of surveillance when monitored data can be collected and processed only by recognized authorities, e.g., for forensic purposes, to counteract terrorism and common crime.
The reconstruction of a crime and, more generally, the a posteriori investigation of an event potentially monitored by distributed mobile sensors, e.g., the nodes of a Vehicular Sensor Network (VSN), require the collection, storage, and retrieval of massive amounts of sensed data. This is a major departure from conventional sensor network operations where data are collected, examined, and dispatched to a "sink" under predefined conditions such as alarm thresholds. In this scenario, it is impossible to deliver all the data detected to the sink (for instance, the police authority) because of size of streaming data (e.g., from video sensors). Moreover, sensing nodes usually cannot determine a priori whether their data will be of any use for future investigations. Then, the problem becomes that of searching for sensed data in a massive, mobile, and completely decentralized storage, by ensuring low intrusiveness to other services, good scalability up to thousands of nodes, and disruption tolerance against mobility/terrorist attacks by promoting completely decentralized cooperation networks.
To tackle such a problem, we have developed MobEyes, a system to form smart mobs for VSN-based proactive urban monitoring. MobEyes exploits wireless-enabled vehicles equipped with video cameras and a variety of sensors to perform event sensing, processing/filtering of sensed data, and ad hoc message routing to other vehicles. Since the sheer amount of data will be generated from those sensors, directly reporting raw data to the authority is infeasible. Thus, MobEyes proposes that: sensed data stay with monitoring mobile nodes (i.e., mobile storage); vehicle-local processing capabilities are used to extract features of interest, e.g., license plates from traffic monitoring images; mobile nodes periodically generate data summaries with extracted features and context information such as timestamps and positioning coordinates; mobile agents, such as police patrolling cars, move and opportunistically harvest summaries from neighbor vehicles. The harvesting agents are interested in the following data: where was a certain vehicle at a certain time; which vehicles were at a given time in a given place, and: what data/video did the vehicle(s) collect? To access the data later, one needs to get to the actual vehicles (based on summary reports) and pump out the data.
The original protocols exploited by MobEyes nodes for summary diffusion/harvesting exploit intrinsic vehicle mobility and single-hop communications among nodes. As thoroughly demonstrated by the reported experimental results, in common deployment scenarios, opportunistic summary diffusion/harvesting permits to build complete indexing in feasible time, with good scalability, and with limited overhead, by maintaining a completely decentralized disruption-tolerant organization. Note that the use of MobEyes is not confined to forensic data collection, yet it can be used in a wide spectrum of applications: for instance, vehicles can measure pollution levels and collect traffic-related data, such as congestion and road conditions. The MobEyes modular architecture facilitates its exploitation in different fields, by clearly confining and limiting the support modifications depending on the specific application logic.
Page updated on Sunday, 25-Mar-2007 23:00:50 CET
In case of problems, or if you find any bug, please contact us.