Road trains drive on public roads that they are allowed to use, and share them with other traffic.
2 Likes
i wasn’t intending to imply it was a good thing. it was more a prediction more than a recommendation
my main thought was after that then consumer full self driving might be ten or twenty years in the future. the other way around would be really surprising to me because it’s so much more complicated at every level
which is why it would be cheap (ish). either: lanes get blocked off completely, or are time shared. my point is just that it’s a lot cheaper than building a train, and it’s likely an attractive idea for companies who want to eliminate labor.
the one pressure against i think is that most drivers - in the us anyway - are contract. so the parent company isn’t really in the business of driving or trucking currently. and they, or some new company, would have to be
( a better bulwark would be a union probably )
Especially since software that’s used in mission critical situations has to pass muster with respect to insurance. And most insurance companies aren’t too keen on autonomous vehicles. I think Level 3 vehicles are fine for long haul trucking like on interstate highways. Even that means dedicated infrastructure which really doesn’t justify their usage when we have literal freight train networks that does many times more work for less. I think the US has a technophilia problem. I love my gadgets but I love stuff that works reliably and doesn’t cost me an arm and a leg (which is why I still have an iPhone 8).
And as a programmer, I’ve dealt with folks who think you can magically automate things away like figuring out what a user wants to do beyond their inputs in simple programs. I’ve had that talk too many times in enterprise development where I tell them, “then you need to buy that kind of functionality from Google or some other company” which doesn’t go down so well but it kills their nonsense requests instantly.
5 Likes
Hoo yeah. I’ve learned that quite often, things the customer thinks are hard can be quite simple - the task looks hard because it’s big. But if it’s systematic: perfect fodder for computers.
And I’ve learned that quite often, things the customer thinks should be simple are not even solved problems at a research level. “But that’s just common sense!”, they say. I reply: “Mate, if I could program common sense, I wouldn’t be working. I’d be retired. To my island. Drinking whatever I wanted to drink from jewel-encrusted goblets.”
5 Likes
5 Likes
Photos of birds, you say…? 
I worked a bit with this crew. They keep endangered bird species from getting killed at windfarms. Windfarm turbines almost look like they’re spinning slowly and lazily, but they’re huge - the blade tip can be moving at 300 km/h (200 mph).
They use binocular cameras to identify when a bird is at risk of colliding with wind turbine blades, and they identify the kind of bird. Sounds harsh, but plant owners don’t want to stop for any non-protected species.
So if a protected species is flying away from a turbine, no problem.
If a city pigeon is flying towards a blade, it’s no problem for anyone except the pigeon.
If a protected species of eagle is at risk of colliding with a particular turbine’s blades, that turbine gets halted as quickly as is practical.
I saw this rolled out at a wind farm that is near the migration path or nesting areas for a few endangered species.
5 Likes
I’d be curious to know how much is business and how much is cultural.
It certainly wouldn’t be a surprise if, per unit nominally ‘self-driving’ car, Green Hills sees more money from a non-tesla than a tesla(though it also wouldn’t be a surprise if at least some of their software or dev tools are involved in bits and pieces of at least some teslas); but I also suspect that one of the founders of a company that has been making humorless high-reliability software since 1982 might be, and have surrounded himself with, people who really can’t stand cowboys and empiricists playing at software engineering.
The fact that it’s a profitable niche certainly doesn’t hurt; but I’d imagine that, among software engineers good enough for this to be a choice, there’s a certain amount of self-selection in terms of who goes down the path of writing EAL 6+/SKPP stuff for the feds and assorted similarly touchy use cases in avionics, medical devices; and industrial control systems; and who goes down the path of moving fast and breaking things or training ML black boxes on ‘big data’.
3 Likes
I take it that that’s the plain reading of the below?
" Dynamic data driven strategies are made possible by simulating various strategies to arrive at a balance for the need for avian conservation and revenue generation."
Out of curiosity, given how sophisticated this tech looks, my naive expectation would be that it would have been rendered impressive but more or less irrelevant by comparatively rudimentary mechanisms for actively discouraging birds from flying near turbines(whether that be beamformed sound waves, laser dazzling, landscaping more attractive corridors around the wind farm, etc.) Do you know if discouraging birds is just a much more challenging task than it sounds like off the cuff; or are there legal implications that prevent harassing exactly the birds that you most urgently need to protect from turbine-induced stress?
3 Likes
This project was in Australia, so a fully plain reading would be “fuck the seagulls”.
I don’t know the reason for their choice, but I have an rough idea of how much they spent on this. Let’s just say the other methods would’ve been cheaper, and they weren’t used.
4 Likes
At least in the transportation context technophilia arguably has a really complex relationship with desire to lower costs: sometimes it is indeed just cost-insensitive tech for tech’s sake; but it’s also not uncommon for it to be expensive looking; but of great interest to people looking to avoid certain very large expenses, or certain categories of expenses by buying some comparatively cheap gadgets.
As you noted, truck autonomy substantially overlaps with freight rail and doesn’t necessarily make sense; but for some people that overlap is exactly the charm: just hold out for machine vision and LIDAR a bit longer and we can replace all that scary capex and downright communist infrastructure implied by building a competent rail network with glorious highways like god and the founding fathers intended!
Various ‘autonomous urban mobility’ things (like this one from citroen) occupy basically the same niche with respect to things like subways, trolley lines, actually walkable city planning, etc. Sure; you could move large volumes of people with downright antique technology(and not even all that much human labor, it turns out that you can do a lot of automating if you constrain the problem space with rails) if you want to be some euro-commie and invest in subways or a trolley or cable car system or whatnot; but building one of those means having to make an ongoing commitment to the idea of public goods and dedicate a fair amount of money to ongoing construction.
Hoping real hard at camera arrays and neural networks, by contrast, may currently have not much you can put into production; and may only ever result in “it’s a taxi; but the sensor cluster costs as much as the car”; but it does promise a drop-in solution that doesn’t require any potentially troubling reexamination of where some decades of intensely car-focused postwar city planning has left us; can(if it ever reaches market) be purchased in fairly small quantities and on short timescales whenever you need a dash of ‘smart city’ to spruce the place up, great for low-commitment initiatives whose future beyond your term in office is irrelevant to you.
5 Likes
This topic was automatically closed after 5 days. New replies are no longer allowed.
