| Mobile ad hoc networking (MANET) enables networked
communications, and hence novel applications, in situations
and venues that do not have fixed communications
infrastructures. Examples of such venues and applications
include military training while deployed to developing
countries, first responder voice and data communications at
disaster sites, and even tactical military and police
communications. In MANET, mobile nodes themselves relay each
others' packets instead of each having to send its traffic
first to a possibly-distant infrastructure point. This
relaying, however, can result in unacceptable latency and
connectivity gaps when the area of operations is large
compared to the typical short range of mobile nodes' radios.
Geocast is a network protocol that replaces standard IP
addressing with addressing by geographic region: a packet is
to be delivered to all nodes occupying the designated
geocast region. This type of addressing is useful in many
MANET applications, such as those listed above, as long as
nodes can sense their geographic locations, e.g. via the
Global Positioning System. However, geocasting over large
areas can suffer from the problems mentioned above when
implemented in a flat network.
This paper describes a new *tiered geocast protocol* that
combines a novel mix of geocast heuristics for short range
transmissions with decision rules for bridging packets onto
a long range subnetwork for long haul transport. In it, a
packet is relayed from node to node until a long range
capable node is reached; that node then transmits the packet
over the long range network to one or more nodes close to
the geocast region; these nodes then retransmit the packet
throughout the geocast region via the short range network.
This paper describes the tiered geocast algorithm, the
design insights necessary to implementing it in the real
world, and an extensive simulation study of the protocol
within two applications from the OneTESS military training
system.
|