Automotive > Infotainment & Telematics Blog



Automated Driving and Potemkin Safety

by Roger Lanctot | Mar 21, 2018

The call to terminate the testing of autonomous driving vehicles on public roads have risen predictably in the wake of the fatal crash between an Uber automated driving test vehicle and a pedestrian walking a bicycle in Tempe, Ariz. Uber halted its testing of such vehicles and, surprisingly or not, Automotive News reports that Toyota has halted its own public road testing.

What is surprising is the editorial assertion of Automotive News Editor-in-Chief Keith Crain that "testing prototypes of vehicles should be done at automotive proving grounds, not on public roads." This viewpoint is nothing less than shocking from an otherwise enlightened reporter of all things automotive - good and bad.

https://tinyurl.com/yasaul5f - "Why Automakers Should User Proving Grounds to Test Unproven Technology" - Autonews.com

What the Defense Advanced Research Projects Agency, and, later Google, demonstrated in their pursuit of the perfection of automated driving is the desperate need to test such systems in real-world circumstances. It is true that a vast assortment of vehicle testing facilities have been created recently across the Midwest generally and most prominently in the states of Ohio and Michigan.

But the impetus behind the creation of these facilities has been to support the now-farcical world of nacent V2V communications using dedicated short range communication (DSRC) technology promoted by the U.S. Department of Transportation. These facilities are indeed vital to the development of this technology for specific use cases such as intersection management and overspeed warnings peculiar to V2V technology. Ironically, these facilities are now being fitted to test cellular-based C-V2X technology alongside or instead of DSRC.

It is necessary to set up so-called proving grounds, away from public roads, for these technologies since they do not exist for the most part "in the wild." An artificial environment is necessary to prove the efficacy of using wireless communications between vehicles (and infrastrucutre and pedestrians) for the purpose of exchanging alerts and avoiding collisions.

Sensor-based automated driving, on the other hand, is an entirely different proposition requiring proof of performance in real world circumstances - a task that Google (now Waymo) embraced with gusto facilitated by California's sometimes supportive Department of Motor Vehicles. Advocating for taking automated vehicle testing off of public roads is at best an effort to drive business to the proprietors of newly minted proving grounds and at worst the equivalent of turning out the lights on automated driving development entirely in the United States.

Thankfully, this proposition is being weighed on an almost-daily basis among experts and engineers at various events around the world - two of which are coming up quickly - Nvidia's GTC AI and Deep Learning event and SAE's Symposium on Automated and Connected Vehicle Testing.

Next week, Nvidia will host its GPU Technology Conference in Silicon Valley where industry experts will take a deep dive into the development of automated driving technology using Nvidia GPUs.

Registration link: http://www.gputechconf.com

For a 25% discount use VIP code NVDASHAPIRO

For those interested parties will a little more patience - 90 days from now, to be exact - the Society of Automotive Engineers is hosting a symposium on June 20-21 in Greenville, S.C. at Clemson University's Center for Automotive Research (CU-CAR). The Symposium on Automated and Connected Vehicle Systems Testing with include stakeholders from across the supply chain and scientific spectrum delving into the requirements for testing and validation inherent in the creation of automated driving systems.

https://www.sae.org/news/press-room/2018/03/sae-international-announces-inaugural-symposium-on-automated-and-connected-vehicle-systems-testing
 
As for where testing of automated vehicle technology ought to take place, there is no doubt that proving grounds can be useful - particularly for testing new concepts and specific theoretical use cases. Proving grounds, away from public roads, are preferred environments for testing systems reliant on technology that is not available within existing infrastructure - such as DSRC or C-V2X technology for inter-vehicle or car-to-infrastructure or car-to-pedestrian communications.

But the progress being made today in refining automated systems, whether it is considered incremental or leaps and bounds, is occurring specifically because testing on public roads is being enabled. The process is even delivering a better understanding of how humans react to the technology - something that cannot be simulated in a Potemkin village.

The crash in Tempe has and will give regulators and automated driving advocates pause to consider the existing regulations or lack of regulations. But it is important to bear in mind that the automated driving horse left the barn as soon as features such as stability and cruise control were first introduced into cars. Whether we realize it or not, automated driving got its start decades ago with the onset of algorithmic vehicle guidance and, more recently, drive by wire technology.

Automated driving is already occurring in millions of cars on public roads. Banishing this technology to the test track is a moot point.

The fatality in the Tempe, Ariz., crash is a tragedy and the parties involved will be held responsible after an appropriate investigation. But when it comes to automated driving development, let's keep it real.

Previous Post: Uber's Monkey in the Wrench | Next Post: The Good, the Bad ... and Tesla, and Uber
Leave a comment