- Total benefits of autonomous driving technology are estimated at $8300/year.
- This is a very large advantage, but not as large as the car-avoiding technologies discussed in Part 1.
- A long list of challenges must also be overcome to achieve the benefits promised by autonomous driving.
- If these challenges are overcome, it is possible that autonomous technology counter-intuitively reduces city driving demand by making car-free lifestyles more attractive.
Fully autonomous vehicles hold significant fundamental promise to reduce costs related to personal mobility. However, cost reduction potential is substantially smaller and also more uncertain than the car-avoiding technologies of advanced telecommunications and small electric vehicles (SEVs) discussed in Part 1. Achieving full autonomy also poses a massive technical and political challenge.
This article will analyse the benefits and challenges of autonomous vehicles. Subsequently, a number of possible future personal mobility scenarios involving autonomous vehicles will be outlined.
Benefits (low level of certainty)
To my surprise, I could not find a single peer-reviewed study quantifying the economic benefits of autonomous vehicles. The best I could find was a 2013 study (not peer-reviewed as far as I can tell) from Columbia University. Results from the study are shown below.
The study found a clear cost advantage from shared driverless vehicles. Adding autonomous driving technology to a sufficiently large fleet of cars would reduce costs from $0.59/mile to $0.41/mile. If “shared driverless purpose built vehicles” (essentially the SEVs discussed in Part 1) were used as driverless taxis, large cost reductions to $0.15/mile could be achieved.
Reduced costs from autonomous technology alone stem from the much greater capacity utilization in such a fleet of autonomous vehicles. This argument makes sense in a growing economy where the time-value of money is high. However, by the time that fully autonomous vehicles arrive on the global mass market, the time-value of money may well be very low or even negative (e.g. most developed nations today), negating this advantage.
For more clarity, imagine a steady-state developed economy with and without fully autonomous vehicles. If the lifetime of the vehicles is measured in total miles, the rate at which the overall economy needs to replace vehicles will be identical between the two options. In fact, the fully autonomous option might even require more new vehicles due to miles that need to be driven with no passengers on board (see below), imposing a larger cost on the overall economy.
Add to this the fact that capital costs of the fully autonomous technology and overhead costs of managing the fully autonomous fleet will add about 10% to the overall costs, and it is quite possible that an autonomous fleet could be significantly more expensive (to the overall economy) than the current personal ownership model.
The much lower costs of the “shared driverless purpose built vehicles” option result from the much lower capital costs ($6500 vs. $25000) and much lower operating costs ($0.04/mile vs. $0.20/mile) assumed in the study. This option should not be compared against standard cars as in the study, but against personal ownership of SEVs.
If this is done, two main disadvantages with an SEV driverless fleet emerge. Firstly, due to the much lower capital and running costs of the vehicle, the addition of fully autonomous technology causes a substantial relative cost increase (about 35% in the study). Secondly, an autonomous SEV covering 400 miles/day will need to recharge periodically during the day, requiring additional charging infrastructure and increasing peak system load. This will be very costly compared to a self-owned SEV covering 50 miles/day which can charge overnight (off-peak) from a standard socket at home.
For these reasons, I’m fairly sure that a fleet of SEV driverless taxis will be significantly more expensive than personal SEV ownership.
Benefits (higher level of certainty)
Other autonomous vehicle benefits can be identified with a higher degree of certainty. These include potentially large increases in efficiency and decreases in wear, lower costs of the passenger’s time, reduced insurance costs and reduced need for parking spaces.
An interesting peer-reviewed paper summarized the possible energy efficiency advantages of autonomous vehicles. The main results are shown below.
Platooning is the result of drag reduction from vehicles following each other closely (especially at high speeds when drag is an important factor). Eco-driving involves driving patterns to operate in the engine’s optimal range more often. De-emphasized performance implies that passengers would actually prefer a small engine to avoid rapid acceleration while updating their Facebook status. Improved crash avoidance is due to weight reduction by dropping vehicle safety features. Vehicle right sizing is essentially the effect of SEVs discussed earlier.
Taking the central estimates of the top 5 benefits in the graph above and combining them yields a total potential efficiency gain of 85% – almost doubling vehicle energy efficiency. Another potential advantage that is not discussed in the paper is the effect of optimized driving patterns, lower drag resistance and less weight on engine lifetime. It seems reasonable to assume that these factors can extend vehicle lifetime mileage by 85% as well. Total costs per mile should therefore be reduced by about 45%, including half price insurance for reduced accidents.
Savings related to better use of the driver’s time while travelling has been imagined for a long time (below). However, compared to being in an office, productivity in a moving vehicle will probably be low, restricted to tasks such as reading emails and checking the news. Upgrading the vehicle to increase productivity will also increase costs and add weight (the “increased features” effect in the graph above). Overall, a 20% reduction in the cost of travel time sounds reasonable – in the middle of the 0-40% range given for passengers in current cars in the literature review of the aforementioned study.
In total, this amounts to a saving of about $0.35/mile using the same data as in Part 1. 10000 miles per year brings this saving up to $3500/year. In addition, about 80% of the $1000/year parking space saving in Part 1 should be achieved by a fleet of autonomous vehicles. And, if fully autonomous vehicles are everywhere, car ownership becomes unnecessary, saving $2500/year from dropping the need for a garage.
Improvements in traffic flow and pollution are more debatable. To substantially improve traffic flow in town with many intersections and pedestrian crossings, the autonomous driving software will need to make optimized decisions based on data from many of the vehicles around it. In essence, the autonomous fleet will have to behave like a single coordinated organism instead of isolated units only responding to information from their immediate surroundings. This will be very complex and also more expensive.
Highway traffic flow, on the other hand, should be significantly smoother even if the autonomous vehicle is only aware of its immediate surroundings. It should also be considered that a reduction in driving costs will generally create more demand, adding to traffic and pollution. In total therefore, it appears reasonable to assume half of the cost saving of removing cars from the road assumed in Part 1, i.e. $1500/year.
The total estimate of potential savings therefore amounts to $8300/year.
The main challenge with autonomous driving is that there is nearly zero tolerance for error in this highly complex undertaking. When Microsoft Windows does something strange, we can simply curse softly to ourselves and click the button to send a crash report to Microsoft. When an autonomous driving software does something strange, on the other hand, many people can die. Deployment of autonomous technology can therefore be expected to be slow due to very tight regulation.
In a controlled environment like a highway that is closed off from the outside environment and contains only identical autonomous cars, it is not hard to imagine this technology working just fine. However, in a city with roads and intersections of all shapes and sizes filled with pedestrians, cyclists, children and pets, the number of variables and possible scenarios to which the software must respond correctly increases tremendously.
Any variation in the autonomous vehicle fleet will also cause problems. For example, if the front car in a closely packed platoon of autonomous cars travelling at high speed on a highway must execute an emergency stop (maybe because of wildlife on the road), it is essential that all the cars behind it have the same stopping distance. Otherwise, we can easily end up with something a lot worse than the bicycle peloton pileup below caused by a single pedestrian. This can limit the utility of the autonomous vehicle fleet, complicate the implementation of incremental vehicle upgrades, and remove all the soul from the car industry.
Then we also have the issue of hackers/viruses. When you get into a car that is fully controlled by a computer, anyone who can access that computer automatically gets full control over your life. In the best case, extremely tight cybersecurity can be expected to add significant costs to autonomous driving technology. In the worst case, this threat can prevent us from realizing the great benefits full autonomy.
Another significant problem is that very few of the substantial advantages discussed in the previous section really kick in before full autonomy (the point where the steering wheel can essentially be dropped to save weight) is achieved in all cars on the road. Partial autonomy can be nice, but it is certainly no game-changer. More importantly, as long as there are human drivers on the roads, tapping the benefits of platooning and weight reduction will be too dangerous, while eco-driving and de-emphasized performance will annoy the human drivers.
Thus, the decades of awkward transition where human and autonomous drivers share the roads will need to be subsidized or mandated because the economic benefits will most likely be smaller than the costs. Given enough time, full autonomy should happen. But getting there will be neither smooth nor fast.
The aforementioned paper investigated four interesting scenarios, the effects of which are summarized below.
Scenarios a) and c) are similar scenarios where complete autonomy is successfully achieved and all efficiency benefits, including vehicle right-sizing (e.g. SEVs), are achieved to their full potential. The difference is that scenario c) also assumes increased energy intensity from higher highway speeds and increased features.
Scenario b) shows the case where full autonomy is not achieved due to technical and/or political reasons. Efficiency gains are much smaller and it is clear that this scenario will not lead to any sort of fundamental shift in personal mobility.
Scenario d) is a rather scary one where a fully autonomous fleet is achieved, but efficiency gains are not realized. Platooning and reduced safety features are not allowed due to the perceived risks of rare-but-catastrophic high speed pile-ups of unsafe cars. In exchange, people opt for higher speeds and stronger (less efficient) engines. As a result, the increased travel demand causes significant congestion, but people are OK with this in the comfortable and feature-rich interiors of their cars. The result is a doubling in energy demand of the transport sector.
My personal view differs from these scenarios in that I actually see autonomous vehicles reducing travel demand. The reason for this is that autonomous vehicles will enhance the attractiveness of the car-removing technology trends discussed in Part 1 of this article. People working from home in a car-free neighbourhood or commuting via e-bike from a centrally located apartment may well need a car from time to time. Being able to summon any type of vehicle from their smartphones at any time will ensure that all personal mobility needs are conveniently covered, thus making a car-free life not only highly economical, but also highly practical for the majority of the population.
In addition, even though the large economic benefits of telecommuting and SEVs are clearly achievable, significant uncertainty remains about whether autonomous driving economic benefits can really be realized. Option d) described above definitely does not sound too far-fetched. For example, the dangers of high-speed platooning in vehicles with very limited safety features will be further enhanced if the wide range of different vehicles needed by society form part of the same platoon. However, even if the economic advantages calculated in the previous section are not fully realized (e.g. no platooning or reduced weight from dropping safety features), the practicality of not owning a car will remain, thus promoting the highly economic car-free lifestyle options described in Part 1 even more strongly.
So, the only discussion point left now is the the types of cars likely to form part of a future autonomous fleet catering to a population with a large proportion of telecommuters and SEV owners. This will be the topic of the third and final part of this article.