In the current version, the testbed is mainly used for implementing Localization algorithms based on Received Signal Strength (RSS). The following is a list of currently supported algorithms. The algorithms marked in bold were developed at NSL as a part of this project:
1. Centroid |
This is a range-free algorithm and comprises of the localizing node simply computing the centroid (arithmetic mean) of the co-ordinates of the anchors/beacons it can hear. |
2. Weighted Centroid |
An improvement over the basic centroid method, this algorithm involves a weighted average of the co-ordinates of the anchors. The weights are assigned based on the distance from the anchor, which in turn, is calculated using RSS. |
3. Grid Scoring |
Liu, Ning, and Du proposed the use of a majority voting protocol in which the network area is partitioned into an evenly-spaced grid. Each node uses RSS based techniques to estimate the distance to each neighboring beacon and constructs a virtual annulus, or ring, around the beacon’s coordinates, where the inner and outer radius of the annulus represent a fixed error bound on the estimated distance. Each grid point receives a score equal to the number of annuli that contain it. The sensor then estimates its location as a point inside the region with maximum score, such as the center-of-mass of the region. |
4. Gradient Descent |
Kwon et al. proposed the use of a minimum mean-square error (MMSE) estimation for range-dependent localization in WSNs. In a distributed implementation, each sensor node estimates the distances to neighboring nodes and iteratively adjusts its current location using a gradient descent algorithm which converges to the MMSE estimate. The distributed implementation yields location estimates in a virtual coordinate system, which can be mapped into a global coordinate system using the reference coordinates of fixed-location beacon nodes. |
5. Weighted Grid Scoring |
This is a weighted version of the Grid Scoring algorithm. The anchors/beacons that are farther away are found to give noisier measurements and are hence assigned a lower weight/trust value than nearer nodes. This modification shows significant improvement over the basic version. |
6. Weighted Gradient Descent |
This is a weighted version of the Gradient Descent algorithm. The anchors that are nearer to the localizing node have a greater effect on the minima calculation than the ones that are farther away. |
|