KRSE3: TDOA-Based Source Localization

1.) Experiment Overview

Experiment Definition and Goals:  The user logs in remotely to open the N6841A Keysight Sensor Management Tool (SMT) and locate AERPAW tower LW1 using a Time-Difference of Arrival (TDOA) algorithm. The AERPAW tower LW1 is continuously transmitting a signal with a center frequency of 3.355 GHz and a bandwidth of 100 MHz. The Software-Defined Radios (SDRs) on the AERPAW UAVs can be programmed to transmit a variety of signals for localization. The main goals of this experiment are to help participants develop a basic understanding of the TDOA localization process and learn how to effectively use the N6841 Keysight SMT and its extensions. KRSE3 is a canonical AERPAW experiment.

Experiment Requirements: It is assumed that the users have AnyDesk, Excel, and MATLAB software (with the Mapping Toolbox) downloaded on their local device. 

2.) AERPAW Tower LW1 Localization Experiment Steps

2.1) First, remotely connect to the AERPAW Windows device for accessing the RF sensors using AnyDesk. A link to the instructions is copied here.

2.2) Click on the Windows Icon in the bottom left corner of the Remote Desktop, open the Keysight RF Sensor folder, and open the SMT software. 

2.3) Open the SMT software and make sure all 5 RF sensors are loaded into the SMT as shown below. If any of the sensors shown are missing, they can be manually added by clicking Tools -> Add Sensors... -> Add a sensor with a known host name or IP address -> and then manually adding the IP addresses listed under the column Sensor alias in the screenshot below for any missing sensors. After ensuring all 5 RF sensors are loaded, select the N6854A Geolocation View button, which is circled in red in the screenshot below.

2.4) OPTIONAL: The default map in the SMT is large enough to show all 5 RF sensors, which encompasses the area from Centennial Campus to Lake Wheeler. If you'd prefer a larger or smaller map, the following instructions show you how to edit the map. If you are happy with the default map, you can continue on to step 2.5. 

To choose a map for the SMT to display results upon, we will use OpenStreetMap which is linked here for your convenience. To export a map, choose export, which is in the top-left corner of OpenStreetMap. Then choose "Manually select a different area", which is in blue under the coordinates shown.

For our experiments, choosing the following coordinates will give a map that encompasses the five RF sensors.

Once the dimensions are set, hit the Share button on the left side of the screen. A share screen should appear with image format and scale options. There should be a check box labeled "Set Custom Dimensions". Choose that option, and a second box will appear on your screen as shown below.

Zoom into two opposite corners and make sure the second box fits over the first bounding box as closely as is possible. Then, as an input to "Scale" on the Share side of the screen, type in 1000. This value should automatically adjust to a larger number around 10,000. This will increase the resolution of the image generated by OpenStreetMap. Then, hit the download button. You should have a file called map.png in your Downloads folder. Make sure you save or write down your map coordinates, as we will need to manually input them into the Sensor Management Tool in a later step.

To import the map into the SMT, choose the Configuration Icon that looks like a gear in the top left corner of the SMT screen, choose the Configure Map tab, and on the bottom of the screen, choose the Browse button near the input for Image File. 

Choose your map from the downloads folder. Make sure you update the coordinates of the upper-left corner and lower-right corner of the map with the coordinates you saved/wrote down from OpenStreetMap earlier in this step. Make sure to select Save Changes when you finish. If you are experimenting high above sea level, you can add a Z Offset in meters to make sure the map is on the same plane as the sensor's GPS readings. 

2.5) Switch back to the N6854A Geolocation View. To localize AERPAW Tower LW1, under Computational parameters, make sure the Show TDOA Hyperbolas box is checked. Under Time series acquisition, change the Center frequency to 3.355 GHz, the bandwidth to 1 MHz, the samples to 32000, and the default antenna to Antenna 1. Under Geolocation sensors, check the boxes for the sensors whose IP addresses end in .13, .15, and .16. Under the attenuation box for each sensor, change the value from 0.0 dB to -19 dB to increase the sensitivity of each sensor to its highest sensitivity. Doppler Compensation can be turned off because all the transmitters and receivers (LW1-5) are stationary. Then, if you are satisfied with the settings and inputs, select Start Acquisition under Acquisition Control. The Geolocation Measurement Setup screen should look like the screen shown below.

Once a measurement is acquired, there are three types of displays that the SMT will generate in the measurement pane (bottom-right corner of the GUI). Clicking the Time Series tab displays a time-domain display of the signals received by the RF Sensors selected for this geolocation measurement. Clicking the Spectra tab displays a spectrum display of the signals received by the RF Sensors selected for this geolocation measurement. The Correlations tab displays the cross correlation between the signals received at pairs of RF sensors. The correlation coefficient, rho, is plotted versus time offset between signals. Each correlation display is labeled at its top left with the names of the sensor pairs whose signals' correlation is displayed. A composite of all correlations is displayed in the Measurement Status area as "Composite rho". The Time Series Detail tab displays measurement information for the traces shown in the "Time Series" tab, as well as settings from the rest of the Time Series Acquisition and the Trigger selection areas used to make the measurement. Representative figures from the Time Series, Spectra, and Correlations tabs are shown below.

For better visualization than what the SMT map provides, once a measurement is acquired, as shown in the below image, choose the Keysight N6854 to KML tool. The software can be found by clicking the Windows icon in the bottom left corner of the screen, scrolling down to open the 'Keysight RF Sensor' folder, and clicking on the N6854 to KML icon.  Enter the measurement ID number that you would like to visualize in the 'N6854A Measurement IDs' input box, check 'Mobile sensors' and 'Estimated position average' under the KML Output heading, and then hit the refresh button. The sensors used for your specific localization measurement should automatically populate into the 'Sensors' box and Google Earth Pro should open and show you the estimated position. 

To generate multiple location estimates without repeatedly pressing the "Start Acquisition" button, you can check the Auto repeat box beside the "Start Acquisition button.  This will attempt to localize the signal source about once every 3-4 seconds. 

2.6) After generating a series of localization estimates, you can export your results using the "Keysight N6854 Export" software, which generates .csv files based on the N6854 Measurement IDs. To use the software, open Keysight N6854 Export while the SMT is open. The software can be found by clicking the Windows icon in the bottom left corner of the screen, scrolling down to open the 'Keysight RF Sensor' folder, and clicking on the N6854 Export icon. 

It can be helpful to write down the Measurement IDs for the first and last measurements of any particular experiment. These Measurement IDs can be found as shown in the following screenshot and need to be input into the following box in the Keysight N6854 Export tool. Once the Measurement IDs are included in the Measurement ID box, you can hit export, and a new .csv file will be created and written to the Documents folder of the remote desktop.

It is recommended to save your .csv file on your personal computer. If you need to save your .csv file on the remote desktop, save it to the ThawSpace (D:). Otherwise, if the remote desktop is turned off, your measurements will be permanently deleted.

2.7) Once the data is downloaded in .csv file form, you are almost ready to post-process and analyze the data. A MATLAB script is available for download to help analyze the data here. The script loads in all the data from the .csv file, uses the Mapping Toolbox to plot the estimated transmitter coordinates and the ground-truth locations of AERPAW towers LW1-4, plots localization error vs. time, and prints average localization error. To ensure your .csv file is compatible with the MATLAB script, open the .csv file in Excel. Highlight all entries in the Time column as shown below.

Right-click on the highlighted cells, and select the Format Cells option. Change the format of the time entries to h:mm:ss as shown in the screenshot below and select OK. Then save and exit the .csv file. It is important to make sure that your column headings are all in the first row of the .csv file, as in the image above. Also, make sure that your file is saved as a .csv file instead of .xlsx or any other file formats.

2.8) Then, add your newly saved and re-formatted .csv file to your working directory in MATLAB.  Open the KeysightRFTDOALocalization.m file downloaded from Github. Be sure to change the filename variable in line 2 to your filename. After that, you can hit run on the MATLAB script to generate the aforementioned plots and find the average localization error. Feel free to make any changes to your individual download of the MATLAB script.

The KeysightRFTDOALocalization.m MATLAB script will output two figures. The first figure shows the estimated transmitter coordinates in reference to AERPAW towers LW1-4. The second figure plots localization error in meters against normalized time (with time = 0 corresponding to the beginning of data collection). Two example plots are included below for reference.

As you can see, there is a high amount of variation and error in the localization estimates. One reason for the high amount of error and variability between localization estimates is that AERPAW Tower LW1 has line of sight transmission to AERPAW Towers LW2 and LW3, but does not have a line of sight transmission path to AERPAW Tower LW4. 2-D localization requires 3 reliable and spatially-separated reference sensors and the lack of line of sight transmission between LW1 and LW4 introduces positive bias into the localization estimates. To improve performance, we could use a UAV to fly above the trees blocking the line of sight between LW1 and LW4. 

Congratulations! You should have successfully localized AERPAW Tower LW1 using the Keysight RF Sensor Management Tool and downloaded the localization data for post-processing and analysis in MATLAB.