Robotaxis – Half-way house.

Robotaxis are autonomous driving v0.5.

• Since the revelation that autonomous driving was not going to live up to the hype, the focus has moved to robotaxis but these too, face plenty of issues.
• Fiat Chrysler and AutoX are the latest in a long line of companies that have announced that they will be launching a robotaxi offering but like everyone else, the timing and returns of such a launch remain unclear.
• I think that the move to launch a robotaxi service before moving to selling autonomous vehicles makes sense from a technical point of view but I am also concerned that the expectations for the economics will not be fulfilled.
• There are a number of reasons why robotaxis come first which include:
  • First, Deep learning limitations: RFM research (see here) has found that deep learning has substantial limitations in that it only works really well where the data set is both finite and stable.
  • The problem is that the road is neither of these things which have caused considerable issues for the systems that have been developed to address this task.
  • The net result has been a huge push back of expectations of when autonomous vehicles will become a reality.
  • Despite this robotaxis may make it to market much earlier.
  • This is because unlike a private vehicle, a taxi can be limited to a specific area of operation.
  • This makes the task being addressed by the system finite, fulfilling one of the criteria for solving a task with deep learning.
  • This is why all of the prospective robotaxi services will be rigidly geofenced in terms of the service area and may also limit pick and drop-offs to specific locations.
  • Second, remote monitoring: Although there has been a lot of noise about robotaxi trials with no driver, there is, in fact, a driver, it is just that he or she is not in the driving seat.
  • To me, this is not autonomy because a human is still effectively in charge of the vehicle and the economics of offering a taxi service have not improved.
  • There is a half-way house between human drivers and fully robotic drivers which is where the system is good enough that one human can oversee a number of vehicles at the same time.
  • This provides an improvement in the economics that could allow a robotaxi business to become viable at lower fares than exist today.
  • Gradually over time, the number of overseers should decline to zero leaving a fully autonomous feel of taxis operating within a slowly expanding area of geographic coverage.
  • Third, economics: With humans at the wheel, the going rate for a ridesharing service is $2-$3 per mile.
  • Tesla and many others are of the opinion that they can deliver a robotaxi service at $1 per mile which they expect will cost them around $0.2 per mile to operate.
  • This is how many can view the robotaxi opportunity as very lucrative and valuable.
  • However, while I believe the cost estimate, I see no reason why the $1 per mile will hold.
  • This is because there are many players all intending to provide a service meaning that there will be brutal competition.
  • Consequently, I see a stable price being more like $0.2 – $0.3 per mile as these companies all fight tooth and nail for market share.
  • Size is going to be very important and once the economic reality hits home, I see a large round of consolidation with only a few big offerings surviving.
• The net result is that while robotaxis are going to come first, they are unlikely to prove to be the gold mine that everyone seems to think.
• Instead, they represent a migration of the taxi industry into a new format where current taxi drivers are going to need to find something else to do for a living.
• A full roll-out of robotaxis is going to take many years giving drivers time to find another profession, but their ascendancy looks inevitable despite the problems.
• However, with where financial expectations are for this industry, I would not want to get involved until profitability expectations have reached something approaching reality.