Đang chuẩn bị nút TẢI XUỐNG, xin hãy chờ
Tải xuống
Tham khảo tài liệu 'communications and networking part 15', kỹ thuật - công nghệ, cơ khí - chế tạo máy phục vụ nhu cầu học tập, nghiên cứu và làm việc hiệu quả | Data-Processing and Optimization Methods for Localization-Tracking Systems 409 Kil xl - xi yi - Vi e eE aa d2 Kil xl - xj Vl - Vj 2 j2 Oil djl j l j l and jl e E 66 where ej indicates the set of links connected to the j-th node. The first case-of-study is a network with NA 4 anchors and one target deployed in a square area of size -10 10 X -10 10 . The target location is generated as a random variable with uniform distribution within the size of the square while anchors are located at the locations X1 -10 -10 x2 10 -10 x3 10 10 and X4 -10 10 . We assume that all nodes are connected and the distance of each link is measured Kij times with Kij e 2 7 . We use the ranging model given in equation 2 to generate distance measurements and we consider Oij e 1e-4 7max . In figure 9 we show the RMSE obtained with different localization algorithms and unitary weight unweighted strategy . In this particular study all algorithms have very similar performance and the reason is due to the convexity property of the WLS-ML objective function. Indeed if the target is inside the convex-hull formed by the anchors and the noise is not sufficiently large then the objective function in typically convex. However all algorithms do not attain the CRLB because under the assumption that ơij s are all different the unitary weight is not optimal. In figure 10 we show the RMSE obtained with the L-GDC algorithm using different weighing strategy namely the optimal the unweighted the exponential and the dispersion weighing strategy given in equations12 14 17 and20 respectively. The results show that the L-GDC algorithm using w j is able to achieve the CRLB whereas the others stay above. Performance of the WLS-ML Algorithms Comparison of different optimization techniques MDS 1 Nystrom sMacof -o- L-GDC CRLB 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 Ơ noise standard deviation Fig. 9. Comparison of different optimization techniques and using binary weight unweighted strategy for a localization problem with