Autonomous Vehicle (AV) Taxis - Yea, Nay or Maybe? - An ARRSEr Poll

Autonomous Vehicle (AV) Taxis - Yea, Nay or Maybe? - An ARRSEr Poll

  • Hell yes! Where do I sign?

    Votes: 17 30.9%
  • Umm…maybe let others be the guinea pigs first, will take a rain check.

    Votes: 14 25.5%
  • Hell no! This is the start of Skynet. Now, where is my Model T taxi?

    Votes: 24 43.6%

  • Total voters
    55
FYI for cars. Right now, Tesla is at Level 3. The Waymo one is Level 4 (mostly).



This is for trucks. Slightly out of date(~2yrs), so most of the pending features can be moved into the "today" part.

Predictive power train controller is here now, most of our drivers disable it as when loaded it just messes up maintaining momentum whilst loaded, ( and the fuel difference according to our telematics is negligible.
Platooning they have trialled on our section of the M6, again on a very long very quiet interstate ( with few exit slips) it may make a difference to fuel burn. In the congested U.K. unless they have dedicated lanes they are a big moving obstruction.
Stand by for a funny from the Autard !
 
That will work for some operators but not all.
We simply don't have time in our network for that kind of operation. As a rule of thumb our customers order up to 14.00, we start dispatching from 20.00 and delivering from 01.00
In between order and delivery 10's of thousands of products run through production, picking, dispatch and on to the vehicles. In some cases delivery can take place around 12 hours after the order is placed in excess of 200 miles away from the factory and delivered by HGV.
Average is around 16-18 hours from order placed to delivery.
There just isn't time to run that through an extra depot or two for onward distribution. One of our larger customers does that (we deliver in to their RDC's) and it adds 24 hours to the delivery cycle with the goods we deliver in to them not arriving in their stores until 24 hours after that. That delivery cycle is three times as long as delivery to end user direct from the factory.
Shelf life on our products is 2 to 6 days depending on what it is.
Our product is typically being consumed less than 20-22 hours after the order is placed.

We supply what is known in the industry as FMCG (Fast Moving Consumer Goods) and the cycle is incredibly quick.
Much of the distribution operation takes place during the night or in the early hours. In most cases the product is comes out of production, picked as it does so and often loaded directly on to the vehicles. Those vehicles are dispatched as loading completes and the Whacky Races start. Vehicles are running with anything from 3000 to 15000 products on. Typically delivered nationwide before 09.00 the following morning (most a lot earlier).

We are far from alone in this practice and most manufacturers of short shelf life products follow a similar model.

When we specify lorries and vans they are ordered with the specific criteria that they must be able to stay at the speed limit up hill and down dale. Slow lorries are no use to us at all. They must be capable of staying on the limiter on any motorway climb they come across.

I had the Renault Truck rep in here last week, he was dumbstruck when told that economy comes a very distant second to performance. Even our Chief Exec and Financial Director understand and are happy that we specify the biggest available engine on lorries.

The distribution model envisaged by most people works fine for your kitchen sink ordered from B&Q. Doesn't work at all for FMCG and that is pretty much every fresh food you eat....

Apologies for the long winded explanation. RDC's and onward distribution only works for medium and slow speed distribution. Fast distribution networks don't work so well that way.
Agree wholeheartedly , I ignore it but the autard lobster will now tell you you are wrong as it knows everything about our job.
 
FYI for cars. Right now, Tesla is at Level 3. The Waymo one is Level 4 (mostly).



This is for trucks. Slightly out of date(~2yrs), so most of the pending features can be moved into the "today" part.
Problems from my perspective on your guide. Level 4 states "under certain conditions" would like to know what they are. Level 3 "the vehicle will inform the driver when he/she must take control" is a massive Human Factors issue in my opinion. As I briefly mentioned previously I participated in some trials on a similar subject. The long winded explanation:
Due to Reagan firing all the Air Traffic Controllers in the USA some years ago it was a big problem for NASA (they don't only do space, they're in charge of airspace as well) to get volunteer ATCOs for any testing they wanted to do. They came to Europe and based themselves in the Dutch testing facility (NLR) in Amsterdam and were trawling to get ATCOs to participate. I got involved and eventually persuaded about 10 other ATCOs to participate as well. The intention was to try a form of 'free flight' where aircraft would be allowed to fly the most direct route they wanted to. There were 3 different samples:
a) ATCO controlled as normal and the aircraft followed a route structure.
b) 'The system' would detect a conflict and instruct aircraft to avoid.
c) 'The system' would detect a conflict and alert the controller who had to jump in and sort it out.
During the trial you were wired up to a heart monitor and eye tracking device, at certain times you pressed a button to indicate how busy you felt.
The results:
a) Same old same old and everything went OK.
b) 'The system' turned aircraft so far off route that it was extremely difficult to get them back to anywhere near where they should be or onto a track taking them back to their route. The loss of time was huge.
c) The controller had to take in the situation that was developing, work out where all the other aircraft were and what they were doing, formulate a plan of action and then implement it. It took a lot of time to do that and even alerting a long way in advance still didn't always work, not to mention the large number of false positives.

There is my issue with level 3, the driver is sitting fat, dumb and happy until suddenly the vehicle needs them to do something. I very much doubt that the average driver would be able to respond in time. There have already been several instances with trial drivers, I would assume specially chosen people, who fail to control the vehicle when instructed, how Joe Bloggs would fare.......
 
And I'm guessing you also don't know the legal ins and outs of traffic signing coupled with the maintenance liabilities thereof, let alone the financial side of it...
In fairness, no. But the point about a dynamic capability still stands. It's also why I can see the advantages of taking such as VMS information inside the vehicle, rather than have it at the roadside.

Yes they can be seen from beyond the headlight range of your motor. That's their only plus in comparison to normal studs. Overall their disbenefits far outweigh that.
Which is why you don't see them abso-bloody-lutely everywhere.
 
Agree wholeheartedly , I ignore it but the autard lobster will now tell you you are wrong as it knows everything about our job.

What a strange post. I laugh at you when you write supposed SME posts explaining the difficulty with creating automous driver “logarithms” as you put it, but I make no claims to know about supply chain.
 
Last edited:
And they can't get the go ahead until they prove its safe, a catch 22 ;-) There has to be a very large issue here in testing under realistic conditions with all of the idiot non AV drivers on the road. Also the realistic situation which TESLA are trying to guard against where you're following a car which does an avoid manoeuvre right smart and your AV doesn't. As per this video:
Tesla Model S on autopilot crashes into van on highway | Daily Mail Online
I dont disagree with you but my point is still not making it through, they have solved that issue but cannot release the code.

Teslas code can detect stationary objects, its a simple calculation based on closing speed.

The incident you linked to was A) usig old code and most importantly B) using a 3rd party device to overide basic safety protocols
 
I dont disagree with you but my point is still not making it through, they have solved that issue but cannot release the code.
I suspect one reason that the road transport authorities are reluctant to allow the 'new code' to be used on the roads is once bitten, twice shy. They allowed live trials on the roads and don't particularly want to be caught out again using untested software in a live environment.
 
I suspect one reason that the road transport authorities are reluctant to allow the 'new code' to be used on the roads is once bitten, twice shy. They allowed live trials on the roads and don't particularly want to be caught out again using untested software in a live environment.
Yes, precisely.
 
This is three years old and shows what was the currrent state. What I find amazing is the 360 sensor coverage, and each cars ability to learn from every journey and share this with all other vehicles. At 11m10 onwards it shows how the AV deals with a wheelchair chasing a duck in the middle of the road. Probably an arrser.
Currently they have 10 billion simulated miles and around 5 million actual in the bag.

 
Currently they have 10 billion simulated miles and around 5 million actual in the bag.
The difficulty being complexity. 5 million miles through the desert or on motorways is vastly less challenging than in a city or town. For me simulators are always a bit meh.
 
The difficulty being complexity. 5 million miles through the desert or on motorways is vastly less challenging than in a city or town. For me simulators are always a bit meh.
Don’t disagree, I think the exponential nature of the AI involved will scale quickly. The video shows it interpreting hand signals, what flashing lights are, cyclists going the wrong way down the street etc.
 

Latest Threads

Top