Low Cost

A location sensing system for Smart Dust imposes certain space, capital, and time costs [16]. These are due to software and hardware required for location sensing on the dust nodes and in the infrastructure (i.e., the base station). Space costs involve the amount of installed infrastructure and the node hardware's size and form factor. Capital costs include factors such as the additional price per Dust node and base station. Time costs include the overhead for system installation, calibration, and administration.

The envisioned application areas for Smart Dust impose certain limits on these costs. The intended low capital cost and small size, for example, require that the location sensing hardware overhead needed on the nodes is minimal. Ideally, a location system would reuse the existing optical receiver instead of adding additional hardware to the nodes. On the other hand, adding additional hardware to the base station is not so critical, because there will be very few base stations when compared to the number of deployed dust nodes. However, introducing additional infrastructure components is not a good idea, because installation and administration of the latter contradicts the ad hoc nature of sensor networks.

Figure 1: Errors in estimated node location depend on whether or not the points of reference for multilateration form a wide baseline.

Note that the limitation to a single piece of infrastructure (i.e., the base station) is a challenging one. In the Smart Dust single-hop network, where nodes cannot communicate directly with each other, node localization requires an external infrastructure. In multilateration-based systems, for example, the distances $d_1, ..., d_N$ to multiple points of reference $1, ..., N$ provided by the infrastructure are measured and used to compute the node's location. In order to achieve high accuracy, the reference points should form a wide baseline, that is, the distances among the reference points should be in the order of the distance of the node to the reference points. Figure 1 illustrates this situation in 2D. There, the distances $d_1$ and $d_2$ of the node to the two reference points 1 and 2 are measured. The node's location is computed as the intersection point of two circles with radius $d_1$ and $d_2$ centered at the reference points. If the two reference points form a wide baseline, an error $e_2$ in the distance measurement $d_2$ causes only a small error in the estimated node location. If the two reference points are close together, the same error $e_2$ causes a large error in the estimated node position.

Implementing a wide baseline typically requires multiple geographically distinct infrastructure components in order to provide the reference points (beacons). Moreover, placement of the beacons is often a non-trivial problem [5]. Usually, the exact locations of the reference points have to be known in order to compute node locations [4,15,24,30]. In some systems, the beacons even need accurately synchronized clocks [18]. In order to avoid these problems, we developed a new localization approach based on cylindrical lateration, which does not have a wide baseline requirement.

Another important overhead involved in setting up a localization system is node calibration [32] in order to enforce a correct mapping of sensor readings to location estimates. In systems based on RF received signal strength (RSSI), for example, the received signal strength is mapped to a range estimate. Variations in transmit power and frequency among the nodes can cause significant inaccuracies in the range estimates when used without calibration [17]. Since the cheap low-power hardware used in WSN typically introduces a high variability between nodes, sensor nodes have to be individually calibrated. This, however, may not be feasible in Smart Dust installations due to their expected large scale. The Lighthouse location system does not require node calibration and thus completely eliminates the overhead of the latter.