Autonomous driving - are we there yet?
In the first part of this user opinion series, written by Jack Creasy, we will find out how far we're already down the road of autonomous driving and if we are even close to large scale production release.
Popular press continues to present the availability of driverless, AI controlled, self-navigating vehicles as - almost ready. All you have to do is read media coverage about organizations such as Waymo or GM/Cruise and you would think that the future is upon us - right now.
It’s my opinion, that we are clearly nowhere near a large scale production release. One example developer, Waymo, has indicated that they will acquire about 80,000 vehicles from FCA and JLR. While there is much discussion about the number of vehicles, I think it’s likely that this acquisition announcement is no more than a ‘Letter of intent’ and that they are more likely to source vehicles over 4 or more years. Even so, this number is impressive, and could potentially mean that Waymo will deploy 1300+ vehicles per month at some future stage. Although this number would represent only 10-15% of the monthly production capability for the Pacifica (FCA) or iPace (Jaguar Land Rover), does anyone really believe that their vehicle deployment could even reach these levels in just a few years? I for one, don’t think it’s even possible.
What are the future service goals for driverless vehicles?
Whether you talk about MAAS, TAAS or Uber/Lyft/Taxi like business models, vehicle requirements are about the same. Driverless vehicles used to supply a service must be able to:
- successfully navigate a specified geo-fenced service area whether it is empty or occupied by passengers.
- respond to all traffic and navigation requirements for a point to point consumer trip.
- identify themselves to the correct consumer since in busy areas there will be many vehicles and consumers mixed together.
- interact with the consumer as trip details may change, or medical or other emergencies may arise.
- be managed and directed from a remote-control center.
- ascertain and have confidence in their roadworthy status.
- understand the complete status of the vehicle and its occupants.
Obviously, most services will adopt a wireless customer or vehicle management system. For example, Uber’s management software is the type of scale required. In this scenario, a customer can request service via a phone application, and the provider allocates the vehicle on demand.
Of course, these requests may be more complex if shared rides are being managed but scaled management software is possible. To date, management software providers have yet to demonstrate a customer or vehicle management product that scales. This may be partly since there are so few driverless vehicles currently on the road and all of these are still in an experimental stage. Uber, Lyft and the Taxi companies give me confidence that scaling management software for driverless vehicles can be resolved.
In my opinion, there are some uncertainties in the current service provider management and support model. For example, Uber and Lyft have an interesting ability to enlarge and shrink the road fleet on demand, and the vehicles come from far and wide - usually where the drivers and their vehicles are located - with no central service, maintenance or storage location.
This human driven vehicle model has no geo-fenced restrictions and provides the maximum flexibility in providing point to point services. If the driverless vehicle provider must increase or decrease the road fleet on demand, this centralized model is much more likely to have most vehicles at a single or at least a very restricted number of locations, and maintenance or storage plus charging if the solution is an EV. This central location must absolutely be within the geo-fenced area serviced by the vehicles. However, this central supply hub may prove less suitable for rapid changes in fleet deployment, and result in more non-productive empty trip miles.
While on the road a human driven vehicle may experience a range of operational problems, from a simple flat tire, to engine light-on problems, or a potentially more serious road accident. Human drivers can easily resolve many of these types of problems. They can change a tire, drive without ADAS functionality, or continue operating the vehicle in the event of a minor crash after assessing the damage.
I expect that driverless vehicles will likely experience similar problems, with the odds being greater with large automated fleets. In a situation where a driverless vehicle has experienced a system failure or minor accident, I believe that the only reasonable choice is for the vehicle to immediately pull off the road and wait for service support to arrive. The idea that driverless vehicles could simply call for a remote driver (human) for support at any time seems highly unlikely to me. For example, a remote driver can’t change a flat tire, or see around the vehicle to assess damage after minor collision.