Safety Assurance for Autonomous Driving by University of Waterloo

Autonomous Driving Software Solutions

What a driver sees versus what the UW safety assurance autonomous solution sees

Researchers at the University of Waterloo are constantly pushing the boundaries of autonomous vehicle development in many areas, including:

  • Object detection implemented using the latest in machine learning
  • Simultaneous localization and mapping (SLAM)
  • Behavioral prediction of cars and pedestrians

With a focus on vehicle operation in adverse weather conditions through all-season driving combined with a thorough understanding of real-world safety
needs, Waterloo is conquering the unique challenges of autonomous driving in cold climates along public roads in Canada.

The University of Waterloo’s fault-tolerant and cyber-secure software solution is system agnostic and can be run on various hardware and software platforms, including Renesas Highly Autonomous Drive Kit. This work also includes research into photo-realistic simulations so Waterloo’s autonomous software and hardware can be initially tested in the safe confines of a laboratory before initial deployment on test tracks and subsequent full testing on public roadways.

There are over 130 faculty at the University of Waterloo engaged in automotive
and mobility systems research managed by the University of Waterloo Centre
for Automotive Research (WatCAR), many with unique competencies in secure
mobility software, connected vehicles, and autonomous driving.

KEY FEATURES:

  • Fault-tolerant architecture with safe emergency pullover, modular for rapid benchmarking
  • All-weather SLAM solutions creating detailed maps alongside online localization using multi-sensor fusion
  • Advanced 3D object detection for cars, bicycles, and pedestrians
  • Comprehensive multi-sensor datasets captured through all four seasons and in various weather conditions
  • Custom hardware solutions for both sensor triggering and power management
  • Rule-based behavioral prediction systems for both cars and pedestrians
  • Real-time continuous motion planning technique running at 10Hz on a single ARM® Cortex®-A57 core on Renesas R-Car