At ECR, we know that deploying autonomous systems that exceed human safety performance is one of the biggest technical challenges of our time. ECR’s mission is to enable our customers to overcome these challenges and build products that change our lives for the better.
Our mission demands not only cutting-edge technologies, but also strong partnerships between domain experts across many organizations. Partnering with ECR elevates the state-of-the-art in autonomy safety faster and more effectively than individual developers can achieve on their own.
BUILD SAFER PERCEPTION
Safety isn’t about nominal performance, it’s about how safely you handle edge cases. What edge cases put your autonomous perception software at risk? You can’t know until you understand what’s in your sensor data. Hologram is a powerful testing and analytics platform that unlocks the value of petabytes of recorded road data to accelerate the development of safer perception software.
ARCHITECT YOUR SAFETY CASE
Traditional safety processes can’t keep up with modern software development. Too often, safety analyses are out of date before they’re even complete. MISSA bridges the gap to better safety programs. It empowers customers to build robust autonomy architectures, perform comprehensive hazard
analysis, and develop solid safety cases.
REVEAL THE UNKNOWN
Achieving the promise of autonomy demands the ability to operate everywhere and safely handle unforeseen situations. The real world has a knack for presenting scenarios that you can’t imagine and don’t test for. Switchboard robustness testing helps you prepare your software for unexpected
situations before it hits the field.
Safety usually isn’t about how you handle nominal cases. It’s about how well you handle edge cases.Prof. Philip Koopman
A FULL-STACK APPROACH TO SAFETY
Edge Case Research advocates a holistic, “full stack” approach to safety. This process starts by considering how a new technology will be deployed in the real world, and then analyzes what safety risks that deployment might present. What behaviors must the technology exhibit to stay safe? In which environments and scenarios must it operate reliably?
Risks might be mitigated through procedures such as safety drivers, or by restricting operation to domains in which considerable evidence of safety has already been collected. Deployments that depend on the safety of autonomy count not only on reliable hardware and defect-free software, but also on the suitability of autonomous behaviors and machine learning to real-world environments.
All this complexity can be hard to tackle with manual, ad hoc processes and tools. Our autonomous future depends on reliable automation to multiply the skill of safety experts.
OUR CORE CAPABILITIES
Perception & MACHINE LEARNING Intelligent machines count on unbelievably complex algorithms and machine learning to understand the world around them. While these techniques are beginning to meet and exceed human performance in many areas, they suffer from vey real safety risks, such as fragility in real-world conditions. Our team understands the risks involved and is pioneering ways to make such systems more robust.
Our team has decades of experience building autonomous vehicles and robotics technologies. We understand how these systems should be architected from end-to-end, and know the challenges developers face when deploying autonomy into unstructured environments.
HAZARD & RISK ANALYSIS
System safety begins with a clear, actionable assessment of risks present. Gaps in risk analyses can lead to blind spots in even mature safety processes. Unfortunately, often risk analyses rely too heavily on oversimplifications and analogies to previously deployed technology. We know this stands in the way of innovative, complex products. That’s why we’re developing systematic tools for modeling, tracking, and validating risks throughout development and after deployment.
Software doesn’t “wear out” like hardware can; software safety risks are posed by defects that can arise at a variety of points within a development process. To combat these risks, our team has decades of experience researching and delivering safe software. We are expert in software-safety approaches such as standards-based “V” processes, which begin with high-quality safety requirements that are refined into an implementation. This ultimately produces a chain of evidence that couples final test results back to the safety-relevant system requirements.
Edge Case Research will be participating in NASA’s Autonomy Workshop to be held on Carnegie Mellon University’s campus in Pittsburgh on October 10-11, 2018. The event is closed to the public, however most of the panel discussions will be livestreamed here:...
Edge Case Research and Underwriters Laboratories have initiated work on a new standard, dubbed UL 4600, that covers safety principles and processes for evaluation of autonomous products, specifically their ability to perform the intended function without human...