On the Effect on Service Levels of Self Driving Buses

[danieldwilliam]

I am thinking about self-driving vehicles and the impact of them on day to day life. 

I am often on a bus from Chippenham to Calne so I’m using this bus route as the basis for some high-level* thinking about the impact on my life. 

I’m deliberately showing my workings in case** I have made a significant error in my assumptions. 

The service from Chippenham via Calne to Swindon that I use takes about 1 hour 5 minutes one way. It runs every 20 minutes from 6am until 10 pm. A 16 hour service. 

Labour Costs 

I think there will be 5 buses on the route at any one time, one way. Ten buses in total. 

Those buses will need drivers.  Ten drivers.  But they don’t work 16 hours. I reckon they work 8 hour shifts. So two shifts are needed for each day. That’s 15 drivers.   

The drivers will want holidays and will call in sick. I’m guessing about 15% additional headcount to cover absensce.  So that’s 17 and ¼  drivers working this route.  I estimate their salary including on-costs at £30k.  (I’m basing this on adverts I used to see in Swindon for bus drivers.) 

That’s a cost of £520k for direct labour driving the bus. 

Fuel Costs 

The journey is 22 miles one way. 

For three buses an hour each way for 16 hours I make that 96 trips of 22 miles or 2,112 miles per day.  The service doesn’t quite run every day but let’s assume it does.*** 365 days at 2,112 miles per day is 770,880 miles per year. 

Deisel fuel retails at £1.42.  Buses get a rebate of 43p on the fuel duty they pay so the real retail price of the fuel is closer to £0.99.**** 

Fuel costs therefore about £282k per annum. 

Total marginal direct cost for the service £800k  

So of the marginal direct costs 35% (ish) is fuel and 65% (ish) is labour. Shall we say a 3:2 ratio of labour to fuel. 

Sacking all the drivers saves £520k.  

If it helps let me point out that computer driven buses are going to be more fuel efficient than human driven buses.  I’m allowing for a 10% fuel saving, worth £28k. This takes the total cost of the service down to £254k or 32% of its original costs.  Pretty much the third of the cost. 

What conclusions do I draw from this? What is the impact? 

If you were to install a fleet of self-driving buses on the Chippenham to Swindon bus service and sacked all the drivers the savings from driver wages would be about the same the fuel costs for running the service, twice. Therefore you could either triple the frequency or run two equivalent service on a different route***** 

Self-driving buses half or better the cost, double or triple the frequency or double or triple the coverage of bus transport in the UK. 

There are some flaws with the analysis. The most significant is that in order to double the number of bus journeys you need to double or triple the number of buses.  Say £100k per bus. That would be £1.6 million for my Bath to Calne route. Or the savings made in drivers’ wage for 3 years. 

*EDITED*

*To update fuel cost figures for a fuel used per mile more reflective of rural driving conditions* Quite literally your mileage may vary for this post.

 

*Vague and almost certainly wrong but hopefully useful in a Fermi sense. Urban bus labour costs to fuel costs might be significantly different as urban buses are in stop start traffic and the density of buses per route is a bit higher (the routes are shorter and more frequent)

 

**In the expectation 

***The buses will have to be driven back and forwards to the depot and theirs driver training and so on. Let’s call it 365 days. 

**** There are other complexities to do with taxation, especialy VAT but unless you are the Tax manager at my last employer I won’t do VAT for you. 

***** A Bath to Calne route would be my favourite.

Comments

[widgetfox.livejournal.com]

If you're planning to do this on a large scale, I think you might also need to model road capacity.

[danieldwilliam.livejournal.com]

Yes you would - especially in an urban environment.

The Chippenham to Calne to Swindon route could probably easily accommodate a service frequency of a bus every ten minutes. Whether there would be the demand for a service that frequently is another matter. Certainly, there are lots of country areas where the service is very limited or non existant.

In Edinburgh you might be able to up some of the less frequent services from 30 minutes to every 15 minutes without crowding most of the route but I think there are likely to be congestion issues in the centre of Edinburgh (unless you saw a reduction in other traffic as a result of the enhanced service).

What you could add are circular or peripheral bus services. At the moment most Edinburgh services run into the middle of town and then out again. Services that ran from around the centre of town would make the overall service much more useful without adding much to congestion.

In London, I’m not sure you can increase a service frequency much from every 5 minutes.

a

[widgetfox.livejournal.com]

Also need to enquire into maximum possible throughput at known bottlenecks.

Re: a

[danieldwilliam.livejournal.com]

Also that. More of problem in cities I think. Again, a big bottleneck in Edinburgh on Princes St.

The main bottle necks of universal application would be bus stations and stops outside popular stops. Thinking about Kemble Station and replacing the hundreds of cars parked and the hundreds of drop offs by car there with a fleet of buses ferrying people from nearby villages you’d have a bit of crowded bus stop.

Thinking about this, there is a knock on effect of better public transport begetting demand for other public transport. If I can get to the station by bus now instead of car might I stop driving to work in the car and go by a combination of bus and train – requiring more trains?

I think the main boost to the quality of life from self-driving buses is in the country. Talking to my mum who was involved in a study about how people could get to a hospital appointment in her hospital by public transport the

A fleet of small, self-driving buses pottering about the Cornish countryside on a twenty minute cycle would seem to be a great improvement over a service that had a few dozen routes served hourly.

Re: a

[f4f3.livejournal.com]

Also you can get into the travelling salesman problem, and buses which regulate their own route depending on sensing if there are customers waiting at particular stops.

Supply and demand can also be regualted better, in that you don't run empty buses. Will continue this down below...

Re: a

[danieldwilliam.livejournal.com]

You certainly could have a bus call service.

Buses operating along a main road– press a button on a bus stop in your local village and the next bus along diverts to collect you. Most buses won’t stop at most stops. If those buses can avoid actually going into those villages then average journey times and fuel use are both reduced.

In cities there might be more scope for smaller buses that wandered along a corridor (say Morningside to the East End of town via the south of the Meadows) and responded to calls from bus stops spread out all over the place. With a service running at five minute intervals there would be loads of buses going near enough to where you wanted to go that you would ever be waiting long even if the next bus along wasn’t able to divert to collect you.

I have a science fiction short story in draft (very early) about a couple trying to get somewhere using an intergrated public transport system that self-optimises and which as got caught on the cusp of two solutions to its current demand and keeps flipping from one to the other.