Autonomous Vehicle Testing — Where is the Due Diligence?
Is anyone doing or even advocating any minimum level testing?
Governments
· Creating Test Cases — State and Local governments are waiting on the Fed’s in DoT. Who are clearly doing very little. A couple of weeks ago the GAO completed an audit of the DoT. The GAO stated that the DoT should be far more organized and actually have a plan moving forward. They stated they should also test AVs for minimum capabilities. The DoT’s response was that they cannot put a plan in place or even create the tests until the various technologies involved are settled. Why isn’t the government at least mandating the AV makers either create a common core industry test scenario set and make it public or make their test own data public if they are going to put vehicles in the public?
· https://www.salon.com/2017/12/09/warning-keep-your-hands-on-the-wheel-of-driverless-cars_partner/
Autonomous Vehicle and Auto Makers
· Creating Test Cases — There is no effort by anyone in the industry to either create an industry wide core test set or to release information proving that minimal testing has been done when they field their vehicles. As for creating the minimal test set on their own. By and large the industry folks are creating test scenarios by mostly using public shadow driving and stumbling on scenarios and variations of scenarios. This will never lead to the discovery of most test scenarios nor actually lead anywhere near the creation of an L4 vehicle.
· Hyping Capabilities — AV makers are also embellishing if not flat out lying about their capabilities. (Intel actually created a fake commercial wit LeBron James). This practice is leading to false confidence levels in the public, press and governments. The reason for this problem is the fierce completion in the industry to be first. To pump up egos and secure funding.
Press/Public Interest Groups/Publications
· Very few members of the press, transportation special interest groups or industry publications, including Consumer Reports, is calling for a minimal core set of capability testing.
The Myth that technology has to be sorted out first
· Can someone give me specific reasons why this is a valid excuse? How does the tech affect the core scenarios I create? And on the other side. Wouldn’t this test set help the industry with minimal target bars and give them a checklist for AI and engineering? There is almost nothing about the technology involved that in any way prohibits experts from creating the millions of scenarios needed to ensure these systems are as good and then better than humans in the myriad of driving situations they experience or can experience. Does a human live driver’s test care what technology the vehicle has?
For more information on why handover/L2/L3 and public shadow driving are unavoidably reckless and will never lead to L4 please see my article
Autonomous Levels 4 and 5 will never be reached without Simulation vs Public Shadow Driving for AI
https://www.linkedin.com/pulse/autonomous-levels-4-5-never-reached-without-michael-dekort
