I’ll check in and contribute, but this is going to be one long messy bunch of posts, so I’ll write my bit and run.
Tesla has been running its research and testing in the middle of the general population rather than only in the closed world of a research facility. That choice is disturbing. However, there seems to be no support for the idea that it has been any more dangerous to the public than our history of trusting the general population to operate powerful machines in a very crowded and fast environment. The mistakes Teslas make seem very similar to those drivers make, and don’t seem to be more frequent.
That’s all true @wentwest . I wonder how companies that make self driving cars will treated in court with the fact that their cars cause accidents, but at a lower rate than the population as a whole.
But this is a different issue - Tesla has constantly, knowingly overstated what their cars can do. That’s a problem.
This is the reason for the SAE and IEEE standards on autonomous vehicles. Manufacturers MUST be held to these standards. IMHO there should be a FEDERAL oversite committee that will certify each vehicle to ensure it meets these standards. We can’t leave it up to the manufacturers to do their own mandating. No company should make a claim about it’s vehicles autonomous capabilities unless it’s federally certified.
There already is a Federal oversight mechanism in place for automobiles. The NHTSA defines and administers the FMVSS to which all automobile manufacturers must comply. All that has to be done is to apply any new standards for autonomous operation to the existing standard. Probably has already been done. They wouldn’t be allowed to operate on public ways without meeting some level of testing and compliance certification by some body of the NHTSA or perhaps a 3rd party NRTL operating under the umbrella of the NHTSA.
Maybe. I think R&D requirements are less stringent than general requirements for uncontrolled use of self driving vehicles. It’s probably something like the NHTSA will allow limited use of self driving technology in public so that the manufacturers can test against real world problems, like unpredictable human drivers and pedestrians if mutually agreed to rules are obeyed. This would include time of day, limited geographic locations, and maximum speeds. Maybe we’re saying the same thing.
Remember, you can’t make an omelet without breaking eggs.
The automobile business has always generated a big supply of BS, about almost everything. Musk has a very large number of spiritual grandparents. In a perverse way it’s a pleasure after a generation of squinty-eyed financial professionals squeezed every drop of joy out of cars.
Yes, we are talking the same thing however, I feel the R&D requirements are much more stringent, especially early on as the device is unproven. Those controls are gradually loosened as successful testing increases confidence in the product performance.
The process of achieving certification involves many steps along the way. First of which is design verification. This is done by the quality group at any company I have ever worked for that does end product development. It has to be done by a separate and autonomous group from the design group. This is the part of the product evaluation that proves the product adheres to the design specifications (did we build the product right). Prior to any product validation (testing in application) the regulatory compliance folks assess the design verification definition, process and results.
Only then is a validation plan defined. This will include gradual steps of introducing the product to the real world environment (did we build the right product). Initially, that is under heavy supervision by people and equipment that can disable the product under test if any significant anomalous behavior is witnessed. It takes a huge amount of accumulated testing before these safety gates are gradually lessened and removed entirely.
The validation plan will include the limitations and test conditions for each stage as you pointed out. They don’t progress to the next stage without a comprehensive review by all parties.
Sometimes, due to various circumstances, the manufacturer is requested/allowed to perform some of the V&V testing. However, when done this way, this is always guided and witnessed by the 3rd party regulatory agency to ensure compliance and accurate results.
In the end, nothing compares to the real world. Unanticipated results are likely to surface. The plans and testing protocols are designed to mitigate as much risk as feasible.
Anybody here recall the first public demonstration of the advanced-technology fly-by-wire Airbus?