I have done some desk-work and put together a few assumptions to see how self-driving transport adds up. What are the problems with self-driving cars or the idea behind them?
Without leaving the kitchen table, I was able to identify some conceptual problems with self-driving cars and ways to improve the efficiency of car use.
If a car drives on average 12,000 miles per year and does an average of 60 miles per hour then that means 200 hours of driving per year. (Cars are spent after just 1400 to 2000 hours of use).
The occupancy rate in relation to hours per year is very low. There are 8760 hours in year. Cars are driven for 200 hours a year, typically. For 8560 hours a year a car is unused. Thus there are 43.8 times more hours of use available than are used each year. If you drove all year, you would still consume around 4 or 5 of cars.
So: let’s get more from the car. In Ireland, for example, there are 2 million cars, give/take. The year’s driving in Ireland could be done by about, roughly, 46,000 cars if they drove 100% of the time and people did not care when they arrived or left.
There are two ways of increasing the efficiency.
A) Drive all the cars all the time
B) Match the cars to the number of hours people need to drive per year.
First A: if we use the existing car fleet to its fullest, that is, drive all the cars all the time then Ireland has a driving capacity of 8760 hours a year times 2 million cars or 17,5 billion hours a year. There is not the demand or road capacity to increase usage of the existing cars to 17.5 billion hours a year in Ireland.
Second, B: let’s assume we could reduce car ownership of cars from 2 million to, say, 50,0000 cars by matching the hours driven to the cars available (fewer cars drive all the time, around the clock). That figure is not practical either.
Car use is not evenly distributed around the day. People drive much less between 18:00 and 6.00. They even drive less around the middle of period 10.00 to 14:00. So, for part of the day we seem to need a million of these cars (which are then not needed much of the rest of time). Whichever way you look at it, we can´t increase the efficiency of the existing fleet by making these cars drive more of the time or with more passenger miles.
Looking at the efficiency another way: the occupancy rate. I recall a promoter of self-driving cars saying hotels and planes would be uneconomical if used as little as cars. So, he imagined that the occupancy rate should be improved, that is when a car is in motion it carries more people.
The occupancy rate of a car in motion is 20% if you allow that most cars can carry five people but only usually car one. It is 25% for four seaters, reasoned the same way.
Twice a day traffic volume approaches the maximum and traffic jams occur (between 06:00 and 18:00 hours). Although the usage rate of cars is really low in terms of hours per year, it is at a maximum in terms of network capacity twice a day. There is considerable underuse of the cars themselves but saturation of use in terms of the volume the road network can carry. Putting more people into cars would reduce the volume of traffic.
What is the effect of this kind of efficiency?
The occupancy rate of the Irish car (or indeed any car is 20 or 25%). If it is increased to 75% (three passengers) then it would appear there are too many cars. Initially one might think that one could reduce the number of cars by two thirds but have them running at 75% efficiency (occupancy rate, not hours per year used).
The consequence of this is that dramatically fewer cars are sold. That’s possibly bad for the current model of car sales. Car manufacturers will have to sell something else.
There are peak times when car demand is high and these times often involve a door to door commute for one person which is reversed in the afternoon. Each time another person is added to the journey, the journey is made longer for all but one of those people (or all of them). See my diagram below.
To increase the occupancy rate to 75% will involve the car collecting three people and delivering three people. Very often those three will live and work in six places (three homes and three workplaces). Three of three passengers will have extended journey times to accommodate the other person.
A possible trade-off is that by reducing the number of cars on the road the journey time (not distance) for the two-to four passengers is shorter – this assumes all other cars are self-driving and fully-occupied.
So, if we reduce the number of cars to increase occupancy, fewer will be sold and self-driving cars that carry extra people will involve longer journey distances/times.
But if people can’t or won’t to share a vehicle with others, then self-driving cars will not change the number of cars on the roads.
If we address the efficiency of journey times by centrally-controlled traffic control systems then the “self-driving car” becomes part of a large, invisible bus-like network of planned journeys or, in theory, an ever-changing, dynamic timetabled system. Conceivably, this network will determine if you drive off immediately at 8.27 a.m. or wait in the car to accommodate traffic volume elsewhere on the route.
The self-driving car will also determine how fast one moves, again to regulate flow to avoid stoppages. So, again, even if drivers own one car and it drives them on one single A-to-B-to-A journey then the speed will be regulated. Journey times may not be very much shorter than with cars driven by humans but there will be a marginal improvement. You might spend less time in motion but more time in the car, waiting.
The other version is a compound of single passenger and multiple passenger journeys as described above which is possibly a sub-optimum solution.
For an undetermined time, there will be human-driven cars on the roads plus bicycles. The autonomous system has not the advantage of the rail network which is effectively a closed system.
From out of this thining I see a few surprises.
One: that the idea of increasing the occupancy rate of cars may not require very many fewer cars as people need them at all around the same time. I imagine that at rush hour, 30% of Ireland’s cars are in motion.
Two: fewer cars are sold but more people ride in them. They will have longer journey times. Even single-passenger cars may find that the optimum journey time is longer so as to make the average journey time across the fleet shorter.
As I said elsewhere, the “self-driving car” is not adding to the attractiveness of private car use. The more effective the system is the more it resembles public transport. I like public transport very much but many people don’t. Should cars become manadatorily self-driving then many people might reason that there is not much point in paying thousands of euros a year for something very like public transport when a much smaller cost gets them 85% of the same mobility.
I can’t say this rough sketch of the issue is much more than problem-spotting but it does show that inside a plausible idea there seem to be some rather serious difficulties for both car makers and car users.
[For this think piece, I have imagined urban and suburban driving. For rural areas, average journey times are more stable as traffic density is very low. Self-driving cars would make very little difference to the overall efficiency of one-person, one-trip driving and increasing occupancy might involve very long and improbable routes for the two/three/four passengers.]
Thanks to Niels Moesgaard-Jörgensen, our correspondent in Sweden, for the conversation that led to this item.