Given the demand for richer services for drivers and their passengers, the next generation of vehicles will be equipped with extensive communication and computing capabilities and will be outfitted with a wide range of mechanical and environmental sensors. This wealth of data and resources has the potential for enabling new classes of applications that cater to individuals on the road, supporting functions such as real-time (on-demand) traffic reports, fuel-saving and safety tips, lo- cal advisories, local search, and local advertising. The goal of this proposal is to design, develop, and test Locus, a delay-tolerant content-centric vehicular network architecture that geographically caches location-based content of interest to commuters. The architectural contribution of the pro- posed work lies in developing protocols to support this first location-centric content access paradigm for delay-tolerant networks, where content with location affinity is efficiently maintained on top of and served to a fleet of moving vehicles.

In participatory environmental sensing, data is inherently tied to the location where it was generated and not to the user who created it, transforming the primary entities of the network into the data objects rather than individual devices. In a mobile environment with limited resources, the accumulation of such location-specific data raises two important questions of where to store the data and how to access it. At first glance, remote access, where data lives on centralized servers, seems to be a natural choice for supporting these applications (e.g., Cartel). However, the reality is that the load generated from participatory sensing systems will overwhelm existing centralized infrastructures. For example, in Chicago there are on average 2500 vehicles per km2. If only 10% of those vehicles are generating data, the network must be able to support 250 data streams per km2. A typical 3G basestation coverage radius is several hundred meters, or approximately 100 users. Given an upload capacity of around 500 Kbps per basestation, that leaves only 5 Kbps per device. While this data rate can support simple sensor readings, transferring larger data objects like images or mp3s will exceed the capacity of the network. More importantly, whatever capacity the basestations do have is shared with competing applications for mobile Internet access. Given that the demand for mobile Internet access will only increase, centralized approaches for location-based applications likely will not have sufficient bandwidth.

A natural progression is to design and deploy decentralized location-centric solutions. In this context, we propose the design of Locus, a vehicular-based content management network that functions as a distributed data cache built on top of a fleet of moving vehicles to support location- centric applications. The novelty of this approach comes from decoupling of the data from the nodes that carry it. Instead, data in Locus is explicitly tied to the location where it was collected or most needed. Vehicles currently in the local area for that data temporarily store the data and pass it to new nodes as they enter the area or drop it as they leave the area. Support for this dynamic caching is enabled using techniques from routing in intermittently connected or Delay Tolerant Networks (DTNs).

Slides:

Locus presentation at CHANTS 2010 - [PDF] [PPTX]