[hawkinspeter]

From: https://nextcity.org/urbanist-news/bikes-not-self-driving-cars-are-the-technological-gateway-to-progress

[quote]It took a whole lot of noise from activists and campaigners for bicycles to be taken seriously at last year’s COP26 summit in Glasgow, and for active travel to be added to the declaration on accelerating the decarbonisation of road transport.

Beyond the serious lobbying from automotive industries, there seems to be a psychological block that prevents the bicycle from being accepted as a central technology when imagining the future of cities.[/quote]

Topic

[hawkinspeter]

Rich_cb wrote:

Rich_cb wrote:

Those scenarios would only occur with privately owned driverless cars. Once autonomous cars are a reality private car ownership will plummet as it will be far cheaper to summon a driverless taxi. Pick up journeys are a good example of how occupancy could increase. If one person is being taken to a station the car could easily pick up several more who want to travel to the station whilst on the way. Then once they’ve been dropped off the car could fill up with passengers who’ve just arrived at the station and head off again. The cars could easily be partitioned so each customer had their own space. Trips to refuel could be scheduled to coincide with delivering passengers near to the depot.

I think that scenario is more or less re-inventing buses. Autonomous buses could work well as they could add pick-ups and drop-offs dynamically whilst on their journey.

[Rich_cb]

Those scenarios would only
Those scenarios would only occur with privately owned driverless cars.

Once autonomous cars are a reality private car ownership will plummet as it will be far cheaper to summon a driverless taxi.

Pick up journeys are a good example of how occupancy could increase. If one person is being taken to a station the car could easily pick up several more who want to travel to the station whilst on the way. Then once they’ve been dropped off the car could fill up with passengers who’ve just arrived at the station and head off again.

The cars could easily be partitioned so each customer had their own space.

Trips to refuel could be scheduled to coincide with delivering passengers near to the depot.

[hawkinspeter]

Rich_cb wrote:

Rich_cb wrote:

Most people want to travel in the sort of comfort that bicycles can’t always provide. There’s no easy way around that. Driverless cars should increase the occupancy of the average car which should significantly decrease the number of cars needed.

I don’t see why the occupancy would increase. One advantage of autonomous vehicles is that they can go and fill up with fuel or drive themselves to where they’re needed, so I’d guess that there will be some journeys with zero occupancy which would reduce the average occupancy.

There’s also the journeys where someone is driving just to get their passenger somewhere, so with autonomous vehicles, that would be reduced from a journey with two people to a journey with one person which again would reduce occupancy although I’d consider it more efficient as it wouldn’t need the driver to be make the journey at all.

[hawkinspeter]

hirsute wrote:

I think it is a sort of taxi-sharing idea.

I can see that you could offer different rates for exclusive use and shared use which could help with this.

Our car does not move off the drive very much but a hire car for us would need to be able to accommodate 2 bikes or 2 kayaks. No doubt 99% of the population will claim a similar sort of special case !

That should reduce the number of parked vehicles, but would probably result in more vehicles driving on the road due to the journeys with zero occupants. Again, the problem with cars is their excess size, so they’re good for transporting a whole family (with or without bikes and kayaks), but they really suck for shifting a large number of people in densely populated areas. E-bikes don’t suffer from being too big and they wouldn’t need to perform unaccompanied journeys (i.e. without a rider) as they are a reasonable price for commuters to invest in one.

[Hirsute]

I think it is a sort of taxi

I think it is a sort of taxi-sharing idea.

I can see that you could offer different rates for exclusive use and shared use which could help with this.

Our car does not move off the drive very much but a hire car for us would need to be able to accommodate 2 bikes or 2 kayaks. No doubt 99% of the population will claim a similar sort of special case !

[Rich_cb]

New cars now have user
New cars now have user profiles which set the air con, radio and even seat to your exact preference automatically.

That will definitely be a feature of shared use cars.

The lack of a boot for storing stuff will be a negative but the cost differential will likely be enormous between shared car use and private car ownership so I expect many people will be willing to accept it.

[Rich_cb]

They will have to park but
They will have to park but there will be significantly fewer of them.

If you consider that the average car is only in use 5% of the time and carries on average 1.1 person per journey the scope for reducing the actual number of cars is enormous.

Increasing the time in use to just 10% and average occupancy to just 2 could reduce car numbers by almost three quarters.

Most of those could be accommodated in densely packed off road parking locations with a small portion parking on street for easy access when needed.

[Rich_cb]

Most people want to travel in
Most people want to travel in the sort of comfort that bicycles can’t always provide. There’s no easy way around that.

Driverless cars should increase the occupancy of the average car which should significantly decrease the number of cars needed.

[IanMSpencer]

I think that’s the nub of the
I think that’s the nub of the problem, so many exceptions, and UK roads build in a lot of requirement for situational awareness. Things like negotiating a two way street with parked cars either side and effectively a single lane, or simply dealing with vehicles parked forcing you to cross a solid white line, or cyclists doing less than 10mph 😉 The exceptions are endless before you get onto the likes of Abingdon where drivers think heavy traffic means no entry signs no longer apply even when facing an angry queue of drivers!

[chrisonabike]

IanMSpencer wrote:

… I’m really just saying that rule-based systems cannot cope with real-world complexity, so they need more understanding of their environment, and I believe that the early adopters severely underestimated the complexity of interactions with other objects, at a time when it is clear that they are struggling even to interpret the fixed objects of lane markings and roads.

These autonomous vehicles need more situational awareness!  Bad driver with quick reactions reminds me of the comments about AI in chess from back in the day.  Along the lines of: programs play dreadful chess – predictable and unimaginative – but humans make wild and inexplicable mistakes.

Debates on AI are amusing but I’d be out of my depth.  I’ll simply agree current systems are still too simple.  The roads and rules are mostly simple (by design) but occasionally weird things do happen.  However current drivers are also prone to awful errors and themselves frequently get overloaded by the unfamilliar *.  So in some senses it’s a low bar to clear.

* Since we’ve an increasing understading of the ways humans will go wrong I’m more for the very simple than the complex at this point.  However complex attracts more money.

[hawkinspeter]

“A bad driver with quick

“A bad driver with quick reactions” – that nicely sums up how I think about autonomous vehicles.

[IanMSpencer]

Given that vehicles have

Given that vehicles have limitations on manouverability, we, as humans, dismiss certain hazards, and place the responsibility on evading a collision on the other objects, as it is physically impossible to avoid. Where we suspect that the other object is not going to cooperate with that, we take preemptive action. So for example, when passing a horse, we do not predict that it will continue on its path, we are told we cannot – and the results of being sat on by a scared horse make it not worth the risk – so we alter our speed and path accordingly, but we have to do that before there is any assessment of there being any obvious intersection of paths.

I guess what I am exploring is the risk factor – motorists are supposed to give cyclists 1.5m+ to mitigate the risk of unexpected manouvres, and motorists are supposed to assess the landscape as a whole for potential hazards, the farm entrance should trigger caution even though nothing is visible.

That last point reminded me of Ashley Neal complaining about the Tesla intelligent cruise control which happily maintained its speed even when there was a car waiting to emerge from a side road – humans are taught to prepare for other drivers’ pontential mistakes, which is all about assessing the pitential for collision even when one is not currently a possibility based purely on the detection of objects. Ashley described it as a bad driver with quick reactions, and it seems that this is the model you are suggesting is acceptable – I’m saying that for AI to approach an acceptable standard, which is to say that it needs to avoid avoidable collisions, I don’t believe that is possible without it having a good understanding of the nature of the enemy. I’ll give you another example, on driving down the M40 I noted a car ahead which initially seemed to be driving predictably and cautiously, but before I was committing to pass it, for no discernable reason wandered out of its lane slightly across onto mine. I watched, couldn’t see a reason and then continued past, noting that as I was about to pass the car was again moving towards the edge of its lane, I gave a “friendly toot” in warning and moved to the other extreme of my lane. I had a look at the driver and concluded that their inability to hold a lane seemed to be incompetence rather than any obvious distraction, there was no other event happening and the vehicle they were planning on passing was still a couple of seconds ahead. So for AI to avoid a collision with non-AI drivers, it needs to have a sense of caution around drivers who display a lack of ability, whereas at the moment, I get the sense that they assume all is well until it isn’t, that as long as they follow the rules, everyone else will. In America, an example would be the Four Way Stop which relies on each driver noting who arrived so they advance in turn, but not everyone will abide by the rules, so AI cars can get stuck due to a cautious approach as some drivers will sniff out the hesitation.

I’m really just saying that rule-based systems cannot cope with real-world complexity, so they need more understanding of their environment, and I believe that the early adopters severely underestimated the complexity of interactions with other objects, at a time when it is clear that they are struggling even to interpret the fixed objects of lane markings and roads.

[hawkinspeter]

I don’t think it’s worthwhile

I don’t think it’s worthwhile to base heuristics on dogs being on leads or not. Surely any object moving along a pavement can be safely ignored until their direction changes. AI should be able to respond almost instantly, so if a dog on a lead suddenly spots a squirrel and manages to slip its lead and run into the road, the car can be hitting the brakes the moment the dog is off the pavement. Humans would believe a dog on a lead to be under control and would be unable to react in time.

However, things suddenly running into the road is an edge case and certainly human drivers will be unable to avoid collisions in those circumstances if they’re moving quickly.

[IanMSpencer]

I think though that

I think though that identifying an object is key to deciding whether it is feasible to predict the path on an object, so we will drive past a dog on a lead, which we probably identify by the juxtaposition of dog to owner as much as observing the lead. without a worry, but a dog that is on the loose, or, as I was suggesting, other dog sized objects, cannot be dismissed. If you don’t dismiss the dog on the lead, then you end up driving to a near stop in case it randomly leaps out. The point being that wildlife and ill-behaved humans of whatever variety may not have predictable paths so need an aggressive defensive response, whereas well-behaved objects tend to be dismissed.

If you have a fast moving object like an over-excited free-running dog, you cannot really predict where they are going to move – the extreme example from my cycling experience is pheasants, which seem to cause a significant number of injuries to cyclists as they burst out at you and you haven’t got time to assess that you need to ruin them over in preference to taking out other riders in the group, or the surprise that causes people to over-react and fall off. Drivers tend to drive through pheasants because they can’t react at all – but what would an AI driver do when detecting an incoming pheasant?

Without cars being segregated from other road users (and I deliberately phrase it in that priority) automation has a far harder task because it needs a far greater understanding of what the object is and its current mode of behaviour to determine how it needs to be handled, and of course, this analysis needs to be near instantaneous from point of detection, or else, in a mixed environment you get AI-based behaviours performing apparently random acts where human drivers have dismissed objects. Now humans, may dismiss some objects incorrectly due to lack of detection, but I’d guess that my dog example would be one where a safety first protocol would be sticking the brakes on because a dignificant number of dog-like objects have an unpredictable behaviour so need extreme caution – like a current Tesla braking in response to ghost objects – jumping at its own shadow – leading to poor interaction with meat-based computers.

[wycombewheeler]

Bungle_52 wrote:

If there are any collisions though I would hope it would delay the introduction of autonomous vehicles, which could do a lot more damage, until they have had a couple of trouble free years but with the current government hell bent on deregulation I fear it won’t.

1800 people a year killed by human driven cars, roll out hasn’t been paused yet.

[Author]

Replies