Reproducible experimentation is a cornerstone of the scientific method. But collecting robotics data is hard! As part of our research, we have released several open source robotics datasets for community use, to help drive the field forward.

• 
  Canadian Planetary Emulation Terrain Energy-Aware Rover Navigation Dataset
  This dataset was collected at the Canadian Space Agency’s Mars Emulation Terrain near Montreal, Quebec, using a rover platform equipped with a unique suite of sensors including: power consumption monitors, a pyranometer, an 10-unit omnidirectional stereo camera, a monocular camera, GPS, IMU and more. Aerial maps of the environment including colour, elevation, slope, and aspect views are also available. The complete dataset is more than 1 TB in size and includes six complete traverses. ROSBag format data are available.
• 
  University of Toronto Foot-Mounted Inertial Navigation Dataset
  This dataset consists of inertial measurements from a foot-mounted (IMU), captured in a variety of environments and with test subjects performing various motions. The complete dataset covers more than 7.6 km, with measurements from five different subjects. Along with the IMU data, we provide ground truth positioning information (in varying forms) to enable the evaluation of zero-velocity-aided, foot-mounted inertial navigation systems (INS). Motions include walking, running, crawling, and stair-climbing.
• 
  ALLO: A Photorealistic Dataset for Anomaly Detection During Proximity Operations in Lunar Orbit
  The ALLO dataset (for Anomaly Localization in Lunar Orbit) is a novel synthetic dataset and data generation pipeline designed to enable the evaluation of visual anomaly detection algorithms in the space domain. Our target application is anomaly detection for worksite monitoring by the Canadarm3, Canada’s contribution to the forthcoming Lunar Gateway space station.