HAMMER Experiment Dataset 1 - Fed4FIRE+

In this dataset, we provide the measurement results (DataSet1) conducted in the course of the HAMMER experiment for Fed4FIRE+. The objective of the Experiments is provided in the uploaded reports, the HAMMER experiment presentation and the HAMMER poster. At the report and the documents, the interested researcher may find details about the conducted measurments and their objectives. The specific data set corresponds to the measurements conducted for the Training of the HAMMER deep learning scheme.  More specifically, in HAMMER: a reconfigurable MuPAR antenna array was implemented with 4 RF chains and 8 parasitic elements. the antenna is able to implement 81 different patterns. the project performed extensive measurments inside the w-iLab.t testbed facilities - testing the various patterns, in order to train a Deep Learning model for hybrid beamforming measuments. The measurements were extracted as follows: NR reference signals (DMRS) where transmitted by the HAMMER BS positioned at [9.5, 7.25] coordinates inside the w-iLab.t space. The HAMMER BS is implemented with two X310 USRPs The mobile node measures the signal (USRP B200) and estimates the radio channel. The procedure is repeated continuously sweeping periodically all beam patterns. The radio channel transfer function and the radio channel impulse response are saved. The dataset contains mat files (MATLAB). Meta-data with details for the measurement setup are included in the mat files. Position of the mobile user Position of the HAMMER BS Center Frequency (3.5GHz) Orientation of the mobile user Speed and Waiting time (at each point) for the mobile user. (the mobile user should be considered static when analyzing the measurements). The main dataset contains: The transfer function for 20MHz bandwidth for all 4 active antenna elements. The Channel Impulse Responsefor all 4 active antenna elements. The RSSI in dB (you can consider it as dBm - however, it is not calibrated for the Rx USRP, so it should not be used for propagation analysis vs. distance). The transfer function has dimensions: No Rx (1) No Tx (4) FFT size = 512 (it was downsampled from 2048 to reduce file size) Time index of measurements. It is noted that: Meaurements 1:81:end represent measurements with pattern index #1  (pattern #1 is all OMNI) Meaurements 2:81:end represent measurements with pattern index #2 ... Meaurements 41:81:end represent measurements with pattern index #41 (pattern #41 is all elements looking in-front) ... Meaurements 81:81:end represent measurements with pattern index #41 (pattern #41 is all elements looking back) The channel estimates are raw. In this specific dataset: the training measurements for the points of map1.mat are provided.  the map1 points can be seen in the screenshot of map1.png, in order to exactly understand the geometry of the environment. Additional files provided include: the pattern of the ESPAR antenna simulated at the azimuth. (Four ESPARs are used) pattern indexes (0 is omni, 1 is looking front, -1 looking back). Details can be found in the HAMMER report. For any question, contact kmaliat@unipi.gr Kostas Maliatsos