4 ms·
This might sound a bit retroactive, but why not require that systems which allow for even temporary autonomous mode have road-embedded systems and real-time com
by mc32 9y ago
This might sound a bit retroactive, but why not require that systems which allow for even temporary autonomous mode have road-embedded systems and real-time comms with neighbors in order to be activational?
That is, do not allow autonomous mode, unless there is an old-school road-embedded autonav system and if there are other vehicles on the road, it be able to realtime comm with at least xx% of its directional and radial neighbors (to add predictability and reduce misinformation.) This in addition to all current guidance systems.
Obvs road-embedded systems do not account for debris, accidents, etc., but know the ideal path and would be backup-of last resort (incapacitated operator, other systems down). Real-time comms with neighbors surfaces good information and buries stale or inaccurate info.
- 83457 9y agoBecause then we would never have autonomous cars
- mc32 9y agoI would contend that. Dissemination (even distribution) might be slower, but it could evolve in specialized local areas (large campuses, special transit corridors, etc.) and slowly but safely expand outside those areas.
- greglindahl 9y agoGiven that you don't have any idea if autopilot was involved in this accident, why are you calling for anything to change? Shouldn't you wait until a few more facts are known?
- mc32 9y agoThat's true with regard to this incident. But there are the incidents by other developers of the technologies where the systems have proven at fault. What I'm saying is that given that the technology has not yet proven itself in demanding scenarios, we may ask for more rigorous requirements if this is to become the default navigation mode of the future (no human intervention). Do we have data on how many times people have had to avert an incident that wasn't reported (my autonav almost crashed me, but I took over)? If we had public data, we could make some kind of assessment but even then, if they proved good, we could make them even better yet with the additional support/failsafe systems.
- CamperBob2 9y agoActually there's some merit behind the suggestion, based on recent work with ground-penetrating radar tech: http://news.mit.edu/2017/lincoln-laboratory-enters-licensing-agreement-to-produce-localizing-ground-penetrating-radar-0718 http://news.mit.edu/2017/lincoln-laboratory-enters-licensing... What absolutely cannot work safely is any sort of system that relies on lane markings. Anyone who's driven around the Seattle area can tell you that much. State DOTs simply DGAF about compliance with road-marking practices, especially with respect to visibility in bad weather. And then there's the question of snow. Additional fail-safe mechanisms of one type or another are definitely going to be necessary.
- Bromskloss 9y agoIs it obvious that the people who would be in power to decide about such a requirement want autonomous cars to exist?
- Piskvorrr 9y agoWhich sort of helps if everybody is honest and at worst feeds data that's stale or imprecise. Enter Mallory. (cf security of Internet @ 1990 and Internet @ 2018)