Next revision | Previous revision |
postuloca:bayesian [2013-03-20 18:06] – created pietro | postuloca:bayesian [2013-04-24 08:20] (current) – pietro |
---|
The figures in the following show the mass probability density of the localizzation error, committed through the bayesian approach described in the paper Savazzi & all; "Radio Imaging by Cooperative Wireless Network: Localizzation Algorithms and Experiments"; IEEE Int. Conf. WCNC 2012. | The figures in the following show the mass probability density of the localizzation error, when the bayesian approach is used. This method is described in the pape Savazzi & all; "Radio Imaging by Cooperative Wireless Network: Localizzation Algorithms and Experiments"; IEEE Int. Conf. WCNC 2012. |
| |
Scenario C40 | ==== Scenario C40 ==== |
| |
| |
The figures below show the results of the scenario represented in the figure 'Scenario C40', where the surface localization has dimensions 6.4 m X 4.5 m, the sensors used to collect the measures are 25, the target move along a path at the rate of one step every two seconds and the RSSI values are collected on a single radio channel. The calibration Phase required by the algorithm has been performed on 22 cells. | In the following table, the Localizzation Error is misured by the bayesian approach by the a-posteriori probability density, based on a gaussian model. |
| The results in the the first row of the table are about the schenario in which we assume that also the cells in the middle in the room, where the sensors are placed, can be crossed. The results in the the second row show the result about the scenario where we assume that the cells can not be crossed. |
| |
| ^ Scenario ^ Parameters ^ Anchor positions ^ Error Distribution ^ Metrics ^ Notes and video ^ |
{{:postuloca:c40_anchors.jpg?200|}} | ^ C 40 | //area//: 6.4 X 4.5 [m]\\ //anchors//: 25\\ //speed//: 0.5[step/s]\\ //channels//: 1\\ //origin//: WnLab | {{:postuloca:c40_anchors.jpg?200|}}|{{:postuloca:err_dist_bayesian_c40.jpg?200|}} | RMSE: 0.867 [m]\\ 75<sup>th</sup> pc: 0.881 [m] \\ 90<sup>th</sup> pc: 1.225 [m] | Error distribution in [0.8, 1] m.\\ {{:postuloca:c40_25sen_1p_1sec.avi|}} | |
| ^:::|:::|:::|{{:postuloca:err_dist_bayesian_c40_prob_loc2.jpg?200|}}| RMSE: 0.77 [m]\\ 75<sup>th</sup> pc: 0.904 [m] \\ 90<sup>th</sup> pc: 1.1 [m] | Error distribution in [0.5, 0.75] m.\\ {{:postuloca:c40_25sen_1p_1sec.avi|}} | |
| ==== Scenario Pat_2 ==== |
| |
{{:postuloca:error_dist_c40.jpg?200|}} | In the following table, the Localizzation Error is misured by the bayesian approach by the a-posteriori probability density, based on a gaussian model. |
This figure shows the localizzation error distribution when the only image radio tomography information is taken into account. | |
| |
{{:postuloca:c40_error_kalman_mdf.jpg?200|}} | ^ Scenario ^ Parameters ^ Anchor positions ^ Error Distribution ^ Metrics ^ Notes and video ^ |
This figure shows the localizzation error distribution when the image radio tomography information is filtered out by the Kalman's Filter. | ^ Pat_2 | //area//: 6.5 X 4.5 [m]\\ //anchors//: 28\\ //speed//: 1[step/s]\\ //channels//: 1\\ //origin//: SPAN Lab | {{:postuloca:pat_2_anchors.jpg?200|}}|{{:postuloca:err_dist_bayesian_pat_2.jpg?200|}}| RMSE: 3.88 [m]\\ 75<sup>th</sup> pc: 4.700 [m] \\ 90<sup>th</sup> pc: 5.800 [m] | Error distribution in [1.2, 5.5] m.\\ {{:postuloca:pat_2_28sen_1p_1sec.avi|}} | |
| |
Through the results shown into the TAB. 1 arises that the Kalman filter does not give significant advantages, in fact, the error in both cases, "with and without" Kalman, is distributed within the interval [0.5, 1.75] meters. | |
| |
| |
The figures below show the results of the scenario represented in the figure 'Scenario Pat_1', where the surface localization has dimensions 8.4 m X 6.5 m, the sensors used to collect the measures are 33, the target move along a path at the rate of one step every two seconds and the RSSI values are collected on five radio channels, but only the two channels with the best signal strength are taken into account. The data for this scenario are granted by Dr. Patwari & all. | |