Changes between Version 12 and Version 13 of Other/Summer/2024/ml5G


Ignore:
Timestamp:
Jul 2, 2024, 3:22:34 PM (5 months ago)
Author:
chrisbovolis
Comment:

—

Legend:

Unmodified
Added
Removed
Modified

  • Other/Summer/2024/ml5G

    v12 v13  
    6262- 4 sensors that compute the FFT of the received signal
    6363
    64 Designed the neural network's input/output formatting.\\\
     64Designed the neural network's input/output formatting.\\
     65
     66\\
     67\\
    6568
    6669
    6770**Week 5 (6/24 - 6/27)**
    6871
    69 This week, we made significant progress in developing and validating our SNR Sensor. The sensor effectively differentiates the original signal from noise, allowing for accurate SNR value calculations. To ensure the sensor's accuracy, we also used FOSPHOR to verify its performance, which alleviated our initial skepticism about the readings. Now we have to test it for our 3 experiments.
     72Developed and validated our SNR Sensor. The sensor effectively differentiates the original signal from noise, allowing for accurate SNR value calculations. To ensure the sensor's accuracy, we also used FOSPHOR to verify its performance, which alleviated our initial skepticism about the readings. Now we have to test it for our 3 experiments.
    7073
    71 We successfully implemented a Satellite Transmitter (DVBT) and a Satellite Receiver in GNU Radio. For the receiver, we developed and integrated a new block that converts stream data into vectors, uploading them as text files within the node. These files serve as collected data to train our machine learning model to predict SINR values, which are our target labels.
     74We successfully implemented a Satellite Transmitter (DVBT) and a Satellite Receiver in GNU Radio for the ORBIT Sandbox 2. For the receiver, we developed and integrated a new block that converts stream FFT data into vectors, storing them as text files within the node. These files will serve as collected data to train our machine learning model to predict SINR values, which are our target labels.
    7275
    7376During our first experiment, we encountered an issue where the connected UHD devices did not correspond to the physical location of the nodes. To address this, we are planning to connect to nodes using SDRs at different grid locations.
    7477
    75 We also working in a stage for automating the entire process to eliminate the need for the GNU Radio GUI. The following steps were taken:
     78Also worked on automating the entire process to eliminate the need for the GNU Radio GUI. The following steps were taken:
    7679
    7780- Extracted Python files and parameterized them, varying the transmitter gain.
    7881- Created functions to log into different nodes.
    7982- Streamlined the emulation process, including running time and data storage.
    80 - Conducted one experiment every three seconds.
     83- Accelerated the process (one experiment every three seconds).
    8184
    82 This automation would enhance our efficiency and consistency in data collection and analysis in the future.
     85\\
     86\\
     87\\
    8388
     89Week 6 (7/1 - 7/3)
     90Tested the automation process on the ORBIT Sandbox 2. Collected about 1000 measurements and then created a simple neural network that computes the transmitter's gain based on the receiver's FFT measurements, in order to test the data saving and data loading mechanisms.
     91
     92Worked on scaling the SB2 Satellite Experiment up to the full Network Coexistence experiment on the ORBIT Grid.  Matched the transmitter/receiver components to specific nodes. Debugged so that the nodes communicate with the on-site SDR devices.
     93
     94