There are a few ways to teach autonomous cars to drive.
One is to have them drive in the real world and to test different autonomy features by enabling them in different contexts. This requires the presence of a human driver who can intervene whenever necessary.
Another is to have them drive in a simulated virtual world.
Finally, you can have them drive in a fake real world. This is a real world setting that’s built for the specific purpose of teaching autonomous cars to drive. So, for example, the roads, signs, and traffic lights reflect real world conditions but dummies are used rather than pedestrians. Here’s an example from the University of Michigan’s Mcity.
The second and third approaches are low risk ways to prepare for the first approach.
In an earlier post, I wrote about how the future of transportation is one where we access autonomous electric vehicles. However, I didn’t address the question of whether these car access networks will own their fleet of cars or aggregate the cars of individual owners when those cars are not being used. Here’s how I think about this problem.
A car access network that owns its fleet of cars will be able to achieve a lower unit cost per car due to its bulk manufacturing (if it’s an OEM) or bulk purchasing (if it’s not an OEM) volume. In contrast, a car access network which aggregates the unused cars of individual owners will be paying these owners a fee which takes into account the higher per unit costs at which these individual owners purchased their cars. It will also incur the additional cost of transporting the car to the owner’s location when the owner wants to use it.
As a result, the car access network with its own fleet will be able to serve passengers at a lower price point than the network which aggregates the cars of individual owners. And since passengers will flock to the network offering lower prices, as long as the network has the financial capital to fund the up-front cost of its own fleet and enough political capital to receive fair treatment when competing for the right to serve a specific region, it will win over the network aggregator.
I wrote about how the combination of car access, autonomous cars, and electric vehicles are coming together to impact the car industry in an earlier post.
However, the car industry isn’t the only one that’s being impacted by these technological changes. Adjacent industries are also going to be impacted and petrol stations are one of them.
Electric vehicles are going to lower the demand for petrol. That’s clear. What’s less clear is whether existing petrol stations will begin to also serve as electric vehicle charging stations. Early indications suggest that this isn’t going to be the case. Petrol stations don’t feel sufficiently threatened by electric vehicles to serve as their charging points yet, so alternative charging networks are being built. By the time petrol stations recognize the threat and try to transform themselves, it will likely be too late as alternative networks will have already been built.
Car access transforms the upfront cost of car ownership to a variable cost. When transportation becomes a variable cost, passengers try to minimize this variable cost at each stage of their journey. And since public transport is a lower cost alternative than car access (due to public transport’s ability to accommodate more people per ride), car access passengers choose public transport where available. The increased use of public transport reduces car use (independent of the energy form that the car consumes) and this reduces the demand for petrol.
Car access networks (like Uber and Lyft today, or those which will potentially be operated by municipalities in the future) have more negotiating power over petrol prices than individual car owners due to the greater collective petrol consumption that they represent. A more concentrated set of buyers reduces the profitability of petrol stations.
Finally, autonomous cars can be programmed to drive more fuel efficiently than human drivers.
Each of these changes on its own is a threat to petrol stations. The combination of these changes may very well be lethal for many of them.
Tesla’s autonomous driving technology had its first lethal accident in May. The driver’s death is unfortunate and I wish his family and friends my condolences.
Despite the tragedy of the loss, the technology’s first lethal accident provides us with a good opportunity to compare the relative safety of autonomous driving technology and human drivers. In a blog post following the accident, Tesla announced that its autonomous driving technology drove for 130 million miles until its first lethal accident. The same rate for human drivers is 1 lethal accident every 94 miles in the US and 1 accident every 60 miles worldwide.
These figures show that, despite the tragedy of the accident, autonomous driving technology is, even its current initial state, safer than human drivers. It’s important to note that while human drivers are unlikely to become safer, autonomous driving technology is still in its early stages of development and will get safer with time. So autonomous driving technology’s safety advantage will grow larger in the future.
Another important aspect of the accident is that the accident’s description suggests that the party at fault isn’t the Tesla but the tractor trailer with a human driver which “drove across the highway perpendicular to the Model S”. So the accident wasn’t caused by the autonomous driving technology, but by a human.
This doesn’t mean that the autonomous driving technology doesn’t have any responsibility. If the Tesla had been driven by a human, perhaps it could have taken action to avoid the collision. So the autonomous driving technology should eventually be able to do the same. And according to Tesla’s automatic braking technology supplier Mobileye’s statement, together with additional detection capabilities which will be live by 2018, this will indeed be the case.
However, what it does mean is that a human driver caused a human in a self-driving car to lose his life. As stated in my post on accessing autonomous electric vehicles, it’s going to be interesting to see how self-driving vehicles and vehicles with human drivers coexist in the future. The former will be safer than the latter so humans using the former are unlikely to want to travel on the same roads as the latter.