Autonomous vehicle & robotics
Functional Safety

The safety of autonomous vehicles relies on bleeding-edge machine learning algorithms. These algorithms involve staggering levels of complexity, and they can suffer from surprising gaps in their knowledge. Edge Case Research uses our expertise in autonomous functional safety as well as a set of proprietary, scalable software robustness testing tools to help you deal with this complexity.

 Can your autonomy system address the everyday extraordinary event?

Can your autonomy system address the everyday extraordinary event?

Robustness testing can find the bugs that you weren't looking for, and could not look for using real-world testing. We have found significant, system-level vulnerabilities in early and mature builds alike, including:

  • Loss of control authority
  • Out-of-bounds control signals (e.g., speed-limit violations)
  • Unacceptably poor responsiveness
  • Inappropriate state transitions
  • Collisions (in simulation, fortunately!)
  • Behavior that traces to violations of safety requirements

ECR's robustness testing can efficiently detect and present failure modes to developers for further diagnosis, not only saving significant work down the road, but ensuring that you will be able to deploy confidently, safely, and more quickly.

Example of the Challenge: PEDESTRIAN DETECTION

Driverless cars and autonomous robots can detect pedestrians using computer vision software trained with labeled video. However, what is a “pedestrian”? Some use wheelchairs or crutches, some are walking a dog, and bicyclists may look like some combination of pedestrian and vehicle. Add in environmental conditions, sensor failures, backgrounds that might camouflage pedestrians or confuse sensors, and you have the potential for millions of combinations that the algorithm may not have learned yet. Each represents a gap in the knowledge of the pedestrian detector, a software error, and a potential safety hazard.

Edge Case Research's proprietary robustness testing methodologies and scalable software tools prioritize tests that are most likely to find safety hazards. Scalable testing tools give developers the feedback they need early in development, so that they can get on the road more quickly with safer, more robust vehicles.

Get there safely

Autonomous vehicle and robotics manufacturers are in a race. You need to get to market faster than your competitors at a cost that won't blow your budget. And above all, if you don't release a product that is safe and secure; even if you are first to market, you won't last long.

So: how do you get your vehicle out on the market quickly when you need as much as a billion hours of operational testing to validate the catastrophic failure rate of a vehicle fleet? Even less variable environments such as autonomous robotics in a warehouse can have surprising problems. Events that human walkers, drivers, and pilots handle intuitively become increasingly intractable problems when dealing with autonomous systems.

Edge Case Research has executed multiple functional safety deployments of autonomous vehicles and robotics, and can help you do so as well.  Our team has a deep background in this area, with multiple members having over a decade of experience in developing and testing in the autonomous robotics and vehicle field.

We can help you to develop your preliminary and refined hazard analyses, understand your risks better, and help you address those safety concerns in a cost-effective manner. In addition, our robustness testing methodologies can help you find the issues you weren't even looking for in a fast, cost-effective manner.

With our experience in the autonomous vehicle functional safety space, we can move quickly to help you develop a safe system that gets out of the high bay and into the field more quickly than you otherwise could.

3485 Butler Street

Copyright 2015, Edge Case Research, LLC. All rights reserved.