Sometimes moving into the future brings us back to the past. We have been tinkering with some thoughts about autonomous cars and the future of automobile ownership for a few months, and over the weekend a Twitter conversation with Farooq Butt and Ben Evans prompted us to post this.
Not so long ago there was a dramatic shift in the way networks got built. For the first century of electronic communications, all signals had to go over dedicated connections. In theory, you could trace a single copper wire from your phone all the way to the person you were calling. There had to be this single line between two points for communications to work. And when your signal got to a transfer point the signal had to be switched (connected to outgoing lines) with this in mind. As a result, we call this ‘circuit switching’. Telecom networks had to be designed with this in mind. As we started adding computers to these networks, and giving computers control of those networks, a new paradigm emerged. Messages between two points could be broken up into constituent parts with each ‘packet’ allowed to traverse the distance from sender to receiver over whatever path was deemed optimal by the intermediate transfer junctions. We called this ‘packet switching’ and it led to a massive increase in efficiency of the underlying networks. There would be no modern Internet without this. The transition from circuit to packet switching is mostly complete today, although there are still some significant hold-outs for circuit switching in mobile networks.
And while this debate is largely finished, the underlying principles are still relevant albeit in new areas, like automobiles.
Twenty years ago, to explain the difference between packet and circuit switches, people often compared cars to trains. Trains could only go along one, pre-defined route like a telecom circuit, while cars could move along like packets. Today, the car is the ultimate example of packet switched technology. You can take your car wherever you want to go along whatever route you choose.
However, as we move to autonomous vehicles it is possible that we will see a big change in this method of operations. Many people think that consumers will not need to own their autonomous cars. Instead, autos will be owned by ride-sharing fleets, and consumers will call them with their mobile apps. The thinking is that most cars are idle 90% of the day – basically taking people to/from work, school and shopping. Thus vehicles are greatly underutilized. A fleet could optimize the usage of this vehicles so that they are driving at close to capacity.
This will effectively convert car driving into a circuit switched activity. That autonomous car you hail on Uber has to be used only for you for the length of your trip (with the minor caveat of things like Uber pooling, which may add a stop). This has important implications for how these fleets will operate.
In the days of circuit switching, telecom networks were built to handle peak capacity. Telcos had to install enough capacity so that at the busiest moment of the day there was sufficient capacity to handle the load, plus some buffer. The operators spent immense effort calculating this capacity, giving rise to some important planning concepts. Those who worked around these networks will remember the key measure of BHCA or Busy Hour Call Attempts, which, as it sounds, is the number of calls attempted at the busiest hour of the day. Network capacity was measured in ‘Erlangs’, which is the capacity of a single wire over one hour. In the telecom world, these concepts were seen as something an obstacle. Everyone was a slave to the BHCA, it drove capex planning. And as you can imagine this metric resulted in a lot of necessary but underutilized capacity. Telecom networks were usually only at BHCA capacity an hour or two a day, and operated at 10%-20% capacity the rest of the time.
This concept is coming back with ride-sharing of automobiles. While we have yet to hear anyone in those circles talk about these terms, the principles are the same. Our understanding is that car availability is the most important competitive metric for the ride sharing apps. If Uber says a car is available in 5 minutes, but Lyft says 6 minutes, Uber wins the ride. This concept, not reputation issues, has reportedly been a big part of share shifts we have in seen in some markets this year. The ability to get cars on the network is almost a founding principle behind the ride-sharing companies. But this whole concept is just a physical manifestation of BHCA. We can rename it Rush Hour Commute Attempts (RHCA), the planning principles are the same.
We suspect that ride sharing companies are deep into this sort of networking planning. Hopefully, they have a few telco grey-hairs around so they do not have to reinvent the wheel on this because RHCA is going to be big part of the lifecycle of ride-sharing and fleet-base autonomous driving. It will drive capex dollars in a major way.
There is also a corollary here for what this means for highway and transportation planning. Today, road building is largely a circuit switched process. We build roads where we can and give them enough capacity to withstand rush hour driving. And while the advent of autonomous vehicles will not fundamentally alter the tyranny of location and real estate, it will at least give transportation agencies some important new tools in controlling traffic flows. The conversion of cars to circuit switching will conversely allow traffic routing to become somewhat packet switched. Today, when people start a car journey they may check traffic conditions at the start of the trip, but usually only for one or two major routing decisions (e.g. 101 or 280, Midtown Tunnel or RFK). With autonomous vehicles connected to traffic monitoring systems, routing decisions can be much more finely tuned. To belabor our historical analogy, traffic systems will become statistically multiplexed much as modern data networks are, with cars routed in real-time.
For those of us who remember the old packet vs circuit debate, there is some irony in this swapping. Viewed a different way it is important to remember that technology is not about right and wrong, it is driven by what is optimal. All systems have their trade-offs, and sometimes what was old is new again.