Autonomous Cars

[Big Tea]

Isn't this CANBUS what makes it so some guy can hack into my car and take control?

Thats crowbar

[roon]

The same logic could be applied to a fuel tank. A swappable pre-filled fuel tank would enable a quicker refuel without the consumer handling hoses and nozzles, while avoiding leakage of fumes. There are a number of reasons for and against this idea.

On the subject of autonomy, I can also imagine a stow-able robotic arm being an option on some cars, for operation of gas station fuel pumps. Since major infrastructure changes seem to be off the table most of the time. If the car can be trusted to drive roadways, it could eventually be trusted to pump its own gas as well. This avoids the passenger needing to exit the vehicle, or even stopping at the fuel station at all, if the cars simply make those trips on their own time.

That said, an AV refuel kiosk could be simpler for a gas station to own/operate. The car could plug its refueling arm into a simple universal fitting. No hose or handle to purchase/maintain as with human-operated kiosks. So some minor infrastructure changes could happen, in this sense.

[AJI]

Ahh, just as I feared. The Autonomous Vehicle thread has now become the 'how to re-fuel an EV' thread...

[Big Tea]

Ahh, just as I feared. The Autonomous Vehicle thread has now become the 'how to re-fuel an EV' thread...

[Andres125sx]

They might consider induction plates in conventional car parking slots, BMW are offering them for your drive but they only deliver 3.5 kW and of course there is an efficiency cost.

I think this is the future, 3.5 kW is pretty poor, but it will surely improve in a near future and nothing can be better than self charging with no interaction once you park your car. Perfect match for an AV.

[Zynerji]

They might consider induction plates in conventional car parking slots, BMW are offering them for your drive but they only deliver 3.5 kW and of course there is an efficiency cost.

I think this is the future, 3.5 kW is pretty poor, but it will surely improve in a near future and nothing can be better than self charging with no interaction once you park your car. Perfect match for an AV.

I thought induction charging was super inefficient??? That would need to be perfected before it could be reasonable.

Parking on a rubber mat in your garage might not be a bad idea for overnight charging, but since the air-gap between charger and inductor has a waste equivalence, trucks might cost twice as much to charge as lower cars..

[Big Tea]

They might consider induction plates in conventional car parking slots, BMW are offering them for your drive but they only deliver 3.5 kW and of course there is an efficiency cost.

I think this is the future, 3.5 kW is pretty poor, but it will surely improve in a near future and nothing can be better than self charging with no interaction once you park your car. Perfect match for an AV.

I thought induction charging was super inefficient??? That would need to be perfected before it could be reasonable.

Parking on a rubber mat in your garage might not be a bad idea for overnight charging, but since the air-gap between charger and inductor has a waste equivalence, trucks might cost twice as much to charge as lower cars..

Again, it revolves around having to get your car in a set position overnight.
If you do not have your own spot, you are excluded.
I do not know what the % of people who can not be sure of being able to park on their pad is, but they would not be in favour

[JonoNic]

I think this is the future, 3.5 kW is pretty poor, but it will surely improve in a near future and nothing can be better than self charging with no interaction once you park your car. Perfect match for an AV.

I thought induction charging was super inefficient??? That would need to be perfected before it could be reasonable.

Parking on a rubber mat in your garage might not be a bad idea for overnight charging, but since the air-gap between charger and inductor has a waste equivalence, trucks might cost twice as much to charge as lower cars..

Again, it revolves around having to get your car in a set position overnight.
If you do not have your own spot, you are excluded.
I do not know what the % of people who can not be sure of being able to park on their pad is, but they would not be in favour

If a car is autonomous then surely it can move it self when fully charged? Then the next vehicle moves in

[Big Tea]

I thought induction charging was super inefficient??? That would need to be perfected before it could be reasonable.

Parking on a rubber mat in your garage might not be a bad idea for overnight charging, but since the air-gap between charger and inductor has a waste equivalence, trucks might cost twice as much to charge as lower cars..

Again, it revolves around having to get your car in a set position overnight.
If you do not have your own spot, you are excluded.
I do not know what the % of people who can not be sure of being able to park on their pad is, but they would not be in favour

If a car is autonomous then surely it can move it self when fully charged? Then the next vehicle moves in

Good point. I have fallen back into the old EV charging trap

[Andres125sx]

They might consider induction plates in conventional car parking slots, BMW are offering them for your drive but they only deliver 3.5 kW and of course there is an efficiency cost.

I think this is the future, 3.5 kW is pretty poor, but it will surely improve in a near future and nothing can be better than self charging with no interaction once you park your car. Perfect match for an AV.

I thought induction charging was super inefficient??? That would need to be perfected before it could be reasonable.

Parking on a rubber mat in your garage might not be a bad idea for overnight charging, but since the air-gap between charger and inductor has a waste equivalence, trucks might cost twice as much to charge as lower cars..

A self adjusting pad wich moves up until the sensor reads X distance (min possible) is fairly simple and easy

[Big Tea]

I thought induction charging was super inefficient??? That would need to be perfected before it could be reasonable.

Parking on a rubber mat in your garage might not be a bad idea for overnight charging, but since the air-gap between charger and inductor has a waste equivalence, trucks might cost twice as much to charge as lower cars..

Again, it revolves around having to get your car in a set position overnight.
If you do not have your own spot, you are excluded.
I do not know what the % of people who can not be sure of being able to park on their pad is, but they would not be in favour

If a car is autonomous then surely it can move it self when fully charged? Then the next vehicle moves in

You get home, get out, the car goes around the corner to a charging center like a multistory carpark, but without having to have people access, and comes back to your door when its done.

[JonoNic]

That could work very well... but then again.

I guess you will need a priority charge as one may need his car earlier than the other person using the same lot. That means that trips will need to be planned ahead if the charge is to keep the duration of the trip.

[Big Tea]

That could work very well... but then again.

I guess you will need a priority charge as one may need his car earlier than the other person using the same lot. That means that trips will need to be planned ahead if the charge is to keep the duration of the trip.

Or, to return to a topic visited earlier, you just get the next car in the 'done' line.
You would of course not expect to pay the same as you would for owning your own car, but would not have service or storage costs.

This could also be useful if you cover long distances. 'your' car meets you at the railway or airport.

[Greg Locock]

My particular bugbear is path prediction, AFAIK they are using Kalman which is NOT conservative for pedestrians in particular.

Are you talking about object recognition or "which object to hit first" in an avoidance scenario? Something else?

Path prediction. A pedestrian who is stationary or walking can accelerate initially at about 1g (it's a bit less but not much) in ANY direction. Therefore, for a robust system, once you have identified a pedestrian its future path should be represented as an expanding circle. One disadvantage of that is that you have to know it is a pedestrian. So, in the unfortunate case of the lady who was Ubered to death, in 6 seconds she could have been anywhere in a circle roughly 60m in radius. As the car approached it would have been obvious that she hadn't started to accelerate, and so the circle would have got smaller, so 1 second from the collision it would be 5m radius.

[roon]

My particular bugbear is path prediction, AFAIK they are using Kalman which is NOT conservative for pedestrians in particular.

Are you talking about object recognition or "which object to hit first" in an avoidance scenario? Something else?

Path prediction. A pedestrian who is stationary or walking can accelerate initially at about 1g (it's a bit less but not much) in ANY direction. Therefore, for a robust system, once you have identified a pedestrian its future path should be represented as an expanding circle. One disadvantage of that is that you have to know it is a pedestrian. So, in the unfortunate case of the lady who was Ubered to death, in 6 seconds she could have been anywhere in a circle roughly 60m in radius. As the car approached it would have been obvious that she hadn't started to accelerate, and so the circle would have got smaller, so 1 second from the collision it would be 5m radius.

Well put. Randomness presents path prediction development limits. Object recognition in this context is more important.

I see reports of Uber hitting 2 millions autonomous miles in December of 2017. Other companies recording more.

3.2 trillion miles travelled on US roads in 2016 according to Fed highway admin. 68.5% considered urban driving (2013 stat). ~2.2 T urban miles. 6000 US pedestrian deaths in 2016.

If I assume most pedestrian deaths and AV mileage is within urban areas, and combine all AV developer miles* to reflect the diversity of approaches to safety present in both AVs and human minds, (excluding Tesla), I get:

~367 MM mi. per pedestrian death for human drivers on urban roads.
~14 MM mi. per pedestrian death for AVs.

Ignoring urban-rural split, and including Tesla semi-auto (which is likely mostly highway mileage):
~533 MM mi. per pedestrian death for human drivers on all roads.
~1,014 MM mi. per pedestrian death for human assisted AVs on all roads.

Would be good to compare these figures with total collision, injury, and fatality figures.

Can't find a good figure on AV fleet sizes. Uber plan to buy 24k Volvos, but their actual running fleet size is unknown to me.

-----------
*Uber: 2 MM miles
Waymo: 5 MM
Alphabet: 5 MM
Tesla: 1 B (semi-auto)
Cruise: <1 MM