MILUV Dataset

Introducing MILUV, the first multi-robot dataset with ultra-wide band (UWB) measurements. This dataset features

  • Pose data
    • On-board inertial measurement unit (IMU) data
    • Camera IMU data from Intel RealSense d435i
    • Height measurements from an onboard downward-facing laser rangefinder
    • Ground truth pose data from motion capture system
  • UWB data from Decawave DWM1000
    • Range measurements
    • Passive listening measurements
    • Channel impulse response (CIR) data
    • Low-level data (raw ROS timestamps, received signal power)
  • Vision data with AprilTag as features
    • Downward-facing camera measurements
    • Monocular and colour forward-facing camera measurements from onboard camera
    • Stereo and infrared forward-facing camera measurements from RealSense d435i
  • Magnetometer data

MILUV Devkit

MILUV is accompanied by a Python devkit to facilitate the development of applications that make use of the dataset. This devkit features

  • Parsing and data loading scripts
  • Three localization benchmarks with which users can compare their localization algorithms
    • $SE(3)$ extended Kalman filter
    • $SE_{2}(3)$ extended Kalman filter
    • Visual inertial odometry using VINS-Fusion
  • Examples of alternative uses for the dataset
    • Line-of-sight/non line-of-sight classification from CIR data
    • AprilTag detection from camera measurements