In this work, we
propose a new approach for routing in sensor networks, called opportunistic
routing. The idea behind opportunistic routing is to utilize the spatial
diversity of nodes to combat fading in wireless environments. To achieve this,
both the network and MAC layers work cooperatively to select a suitable
forwarding node from a set of candidate forwarders at the time a packet needs to
be transmitted. The routing scheme is an extension of geographic routing, since
the set of nodes considered for forwarding is based on node location
information. The specification of a set of nodes and not just one next hop node
makes it much more robust and power efficient in the presence of fading channels
and node failures. The MAC layer is an asynchronous signaling scheme that
chooses the best forwarding node from this set of potential forwarders based on
hints from the network layer and the current connectivity within the
neighborhood. This scheme offers both energy savings and reduced per-hop delay
than routing protocols that do not take the spatial diversity of nodes into
account.
Lifetime
issues in sensor networks:
Defining
network lifetime for sensor networks is not a simple task, since this is
completely application dependent. However, there needs to be a metric that is
independent of the application that protocol designers at the network layer and
below can use to quantify the performance of their schemes. The usual definition
used for such purposes is the time it takes for the first node to die (run out
of energy), or the time it takes x% of the nodes to die, with the rationale for
the latter definition being that losing a few nodes does not usually matter in a
dense network. Maybe that will hold true in the future when nodes are so cheap
that massive redundancy is possible, but currently loss of even a single node
starts degrading the network. Hence we focus on the concept of target lifetime,
which is the time a network is expected to remain operational. All nodes try to
achieve this lifetime and failure of even a single node before this time
signifies a failure of the network to achieve its target lifetime. If protocols
are designed such that all nodes try to modify their operation to achieve the
target lifetime and automatically move the load from energy-poor to energy-rich
nodes, then in that case, either the entire network achieves the lifetime, or a
whole set of critical nodes fails, which would definitely construe a failure in
achieving the lifetime.
Duty
cycling mechanisms for topology control in sensor networks:
Energy
usage of nodes is a major constraint in sensor networks. To reduce the
consumption of energy as far as possible, nodes are typically duty cycled, often
times with duty cycles being as low as 1-10%. This leads to another problem –
how to achieve connectivity and packet transfer while satisfying delay
guarantees. The approach we propose is completely asynchronous and decentralized
in nature, with each node making its own decision about when and how long to
sleep without any overhead or exchange of information (even infrequently) among
nodes. This is achieved by just measuring the per-hop delay and rate of packets
that the node itself forwards. The decision is made taking into account the
application delay constraints and the target lifetime of the network. If the
network can satisfy the constraints, it goes further to minimize the energy
consumption of the network as a whole (and not just the node energy consumption)
so that the network lifetime can be maximized above and beyond the target
lifetime. On the other hand, if these constraints cannot be satisfied by the
network, a signal is sent to the network administrator so that additional
resources can be deployed in the network.
Sparse
sensor networks:
The
problem in sparse sensor networks is that nodes do not have sufficient
connectivity to transfer data across the network. To provide this connectivity,
we propose using an intermediate transport layer of mobile nodes to move the
data across the network. Assuming that the mobile nodes, called MULEs, move at
random across the network we defined a system architecture to transfer data from
these sparse sensor nodes to a central access point using the MULEs. We also
studied the system performance in terms of throughput and delay and how these
quantities scale with the system size. This particular scenario is an example of
the broader problem of delay tolerant networking which is another hot research
area currently.
