Integrated Sensing, Computing and Networked Systems: "Mitigating Bottlenecks and Hotspots in Wireless Sensor Systems"

With the recent progress in microelectronics and micromechanics, integrated, small footprint combinations of sensors, wireless transceivers, and energy sources are rapidly becoming reality. We envision that in the foreseeable future, sensor/communication nodes will become cheap enough to be used in huge quantities, and small
enough to be easily integrated into our daily living environment, and that the energy/power consumption of the nodes will be low enough for them to operate continuously for a very long period of time from a self-contained power source.
The latter requirement is actually the essential one. Reliable operation of the sensor network requires that enough nodes are always operational to execute the requested services. "Lifetime" and "survivability" are essential metrics of any sensor network. The limited availability of energy (for battery-operated nodes) or power (for nodes that operate on energy-scavenging) puts an upper limit on the amount of computation and communication that can be performed on a single network node. While average energy (power) consumption per node may seem to be a reasonable measure for the lifetime of the network, this metric is grossly inadequate for virtually any realistic sensor system. The presence of hotspots and bottlenecks causes some parts of the network to consume energy at much faster rates than others, hence causing the network to fail earlier. For example, nodes situated around a data-collection node (called a monitor node) are subject to more traffic than remote sensor nodes, and can fail a lot earlier. These hotspots are the Achilles heel to the potential widespread use of sensor networks.
A number of techniques to address the uneven distribution of activity levels in wireless sensor networks, and consequently extend the lifetime of the network, are advanced in this proposal. While each of these approaches is bound to have a considerable impact, it is their combination that has the most dramatic result. We project that the combined impact of the proposed techniques will increase the longevity and robustness of sensor networks by at least a factor of 10:
> Ad-hoc routing techniques to distribute the traffic over the complete network rather than along a number of bottleneck traffic lanes
> Network topology management to equalize traffic density over the network
> Utilization of "altruistic" nodes
> Aggregation and distributed source-coding to compress the traffic flow in the direction and vicinity of the monitor node
> Adaptive and dynamic relocation of services within the network to avoid exhaustion of specific network regions
The algorithms resulting from this research will not only be analyzed theoretically and by simulation, but will also be tested empirically in a real wireless sensor testbed, as is currently available at the Berkeley Wireless Research Center (BWRC).