Shopping for an autonomous vehicle? Don’t put the cart before the horse
May 04, 2021
May 04, 2021
What the standard approach to choosing AV technology often gets wrong―and how to get it right
Stantec GenerationAV™ External Consultant Chase Williams contributed to this blog.
Successfully rolling out automated vehicle (AV) technology demands a deep commitment to safety, one that must start with a deep dive into the hazards and risks of the environment where an AV operates. Understanding that environment plays a critical role in reducing risk and enhancing performance.
Even if you haven’t helped deploy an automated driving system (ADS), you can easily follow the logic of the “standard” approach to assessing the operational design domain (ODD)—the engineering term for where the technology will operate. That approach calls for the project manager and the ADS developer to carry out a preliminary assessment where they identify operational risks, ways to mitigate them, and actions the team needs to complete before deployment starts.
First, like any standard approach, the apparent simplicity of this step can lead to a “we’ve got this covered” attitude, which is problematic. Team members without deep ADS experience may not ask enough questions, ask the right questions, or take seriously enough issues unique to a particular situation.
A second, less obvious problem, can also hide here. Most teams choose their ADS technology before they assess the ODD where it will operate. That means that performance characteristics play no role in determining the best application, and the team ends up with a “patchwork” approach, addressing risks it finds in the ODD assessment but in ways limited by the ADS. Put another way, would you buy a refrigerator before you measured the space where you planned to install it? Of course not.
Our advice? Never choose the technology first. Safe deployment begins with an in-depth review of the ODD and its hazards. The team needs a robust understanding of the environment and its variables before considering which AV technology can meet (and exceed) project goals.
We start with one of the best tools for environmental safety assessment, the British Standards Institute’s (BSI) Publicly Available Specification 1883 (“PAS 1883”). The BSI explains the tool’s function as a “minimum hierarchical taxonomy for specifying an ODD to enable the safe deployment of an automated driving system.”
We start with PAS 1883 because it provides a detailed inventory of the elements any comprehensive assessment needs to review. It doesn’t spell out specific inherent hazards, of course. Rather, it outlines where to start looking for operational risks but puts the onus on the project team to identify them—along with any unique hazards it might find.
During an ODD assessment we build on this framework. We expand each subcategory into unique defining traits to build a robust picture of the environment where we plan to deploy. (We recommend downloading PAS 1883 for details and guidance.)
Deployment teams should focus first on fully understanding the ODD and only then shift their focus to an ADS solution.
Some characteristics, such as speed limit, can turn out less clear-cut than they look at first glance. For example, it might seem reasonable to operate an ADS that supports a top speed of 25 mph on a road with a posted limit of 25 mph. Actual speeds, however, could easily exceed 25 mph, and the ADS may have a safe operating range of 12-15 mph. That would create a speed differential too great for operation in mixed traffic. Generally, the ADS should operate at the maximum speed that all other vehicles operate.
This should be a nonnegotiable best practice for ensuring the safety of passengers, other road users, and property. That cuts out the need to add engineering controls after the fact to mitigate risk because the ADS performance limits don’t match the ODD. In other words, we don’t have to worry about rebuilding our refrigerator space. Why? Because we measured before we went shopping.
Another foundational principle in running our ODD assessments: Always plan for the unexpected. Returning to an ADS with a listed maximum of 25 mph, what guarantees that no other vehicles will exceed 25 mph? What about drivers who speed intentionally—how fast might they go? How badly could they hurt our ADS if they hit it? We can’t overlook any possibilities, and we must build them into our initial analysis.
The complexity of the ODD’s variables underscores the need to tackle that assessment first. Once we complete it, we can begin to look at ADS candidates and zero in on the platform that best fits the operating environment as we now understand it. At this point, a project team might consider:
Our Stantec GenerationAV™ team always works to build a complete and detailed understanding of how any AV will react to the environment and the obstructions within it. Further, we won’t proceed without a solid grasp of how variability within environmental factors could affect performance. For example, does the ADS disengage from automated mode when it rains? What operational constraints should we plan if we know it works fine in light rain but can’t handle a downpour?
Bottom line: deployment teams should focus first on fully understanding the ODD and only then shift their focus to an ADS solution. Each ADS has advantages and disadvantages, but without a detailed grasp of ODD hazards, choosing the ADS first looks a lot like buying a refrigerator before measuring the kitchen space.
We’ve compiled a brief primer describing where both ODD and ADS reviews should focus. Download it for free at Stantec’s GenerationAV™ Learning Center.