[ABSTRACT] Sensor networks are made of nodes that have a processor, memory, wireless transceivers, sensor(s) and a onboard battery. Sensor networks are expected to revolutionize information gathering, processing and dissemination in many diverse environments.The most important factor in the design of sensor network protocols is the conservation of battery capacity. Surprisingly, almost all existing work uses unrealistic battery models. These models assume that a battery is a ``bucket'' filled with a number of units of energy and that each packet removes a fixed number of units. Real batteries are much more complex, dynamic systems that have several different phenomena existing simultaneously. The objective of this paper is to demonstrate that algorithms designed using more complex battery models perform better in real networks.
In this paper, we model arguably the most important non-ideal property
of real batteries, viz., the charge recovery effect -- battery capacity
actually increases when it is allowed to rest for some time. We develop
a simple model for the charge recovery effect, and then propose a routing
algorithm based on it. Our algorithm makes use of two simple, intuitive
objectives: each battery should be allowed to rest between uses if possible,
so that its capacity regenerates, and that the communication load between
two nodes should be distributed over multiple paths between them. While it
would be nice to balance the communication load between multiple paths, generating
edge or node disjoint paths is a computationally hard problem. We address
this problem by generating multiple paths using existing algorithms for
braided multipaths. This can be done efficiently and produces paths that
are mostly disjoint. We compare (using simulation) our algorithm with the
well known Directed Diffusion algorithm which uses a single path for all
packets. Our experiments show that our algorithm significantly outperforms
Directed Diffusion for our battery model.