DeepUWB

A dataset for UWB ranging error mitigation in indoor environments, built using Decawave EVB1000 devices and the firmware contiki-uwb. Additional information can be found in the attached file "readme.txt" or in the paper Robust Ultra-wideband Range Error Mitigation with Deep Learning at the Edge.

[ readme.txt ]

* Every sample has the following structure:

  || CIR (157 float values) ||
  || Error [m] ||
  || Room (int) ||
  || Obstacle (10 bool values) ||
  || Measured Range (UWB) [m] ||

* Room encoding:

  0 -> cross-room measurements
  1 -> big room
  2 -> medium room
  3 -> small room
  4 -> outdoor

* Obstacle encoding: (1-hot encoding)

  1000000000 -> wall
  0100000000 -> polystyrene plate
  0010000000 -> plastic (trash bin and chair)
  0001000000 -> plywood plate
  0000100000 -> cardboard box
  0000010000 -> LCD TV
  0000001000 -> metal plate
  0000000100 -> wood door
  0000000010 -> glass plate
  0000000001 -> metal window

* Reading Code:

  # Import libraries
  import pandas as pd
  import numpy as np

  # Extract dataset
  dataset = pd.read_pickle('dataset.pkl')

  # Select specific obstacle configurations
  ds = np.asarray(dataset.loc[dataset['Objects']=='011111111'][['CIR','Error']])

  # Select specific rooms
  ds = np.asarray(dataset.loc[dataset['Room']==1][['CIR','Error']])

  # Select all samples
  ds = np.asarray(dataset[['CIR','Error']])

  # Get X,y for training  
  X = np.vstack(ds[:,0])
  y = np.array(ds[:,1])