A recent article on nationalacademies.org (which certainly sounds like a solid reference) has as its title “Driverless Motor Vehicles: Not Yet Ready for Prime Time” (link below). Since an abundance of media space too often is given to press releases from technology companies, which are then echoed by techno-optimistic bloggers, this was a title that caught my eye. While it would be nonsensical to deny the achievements of technological progress and a mistake to minimise them, there are still reasons to remain cautious about some of the promises made by visionaries of technological progress. If they all would have come through as envisioned, I would already in teenage years have been flying to school in a jetpack and friendly robots would have guided me to fully immersive learning experiences (instead of having to bicycle to school and watch the teacher write on a blackboard – and yes, I am that old, blackboards and overhead projectors were the prime means for display of information back then).
The article acknowledges that the current death toll in traffic could be significantly reduced with increased use of automation for road transportation. However, it does pose two questions often asked in regards to automation: 1) What happens if the automation fails? and 2) What if the automation has to handle a situation that has not been considered by its designers? These questions has been asked throughout the history of automation, robotics and AI and it has been explored by researchers and others thinking about the potential and limitations of technology. In the article, these questions are focused on fully automated vehicles, i.e. self-driving cars but it brings up many aspects which are equally relevant for aviation.
The article goes on to discuss the challenges to achieve the vision of a fully automated road transportation system and references the experiences of automation in aviation. The point made is that in aviation, automation has been an ongoing process for decades, that this has been in a less complex environment than road transport and that still today pilots are not expected to be replaced any time soon. I am not sure that I agree with either of these statements, given the complexity of any airspace environment, especially close to a major airport, and the current developments towards single pilot operations. But perhaps complexity can be a matter of perspective.
Developing autonomous vehicles is, according to the article, more challenging than previously envisioned and will take longer than many of its proponents advocates have anticipated. Also, the many Human Factors challenges experienced in aviation are mentioned, with implicit reference to accidents and experienced related to automation. Special emphasis is placed upon the difficulties of mixing autonomous and human operators and the potential failures of automation. The recent Boeing 737MAX accidents are brought up, so is the “Miracle on the Hudson”, as a counterpoint. While providing a less positive perspective than often encountered in texts about automated transport, it is useful to read this perspective and consider the remaining obstacles to this vision.
The point of the article may be supported by the experience in daily life of interacting with the “chatbots” often used these days for customer service. Even if it is not made clear to us that we are speaking to a bot, it does not take long (a non-standard question is usually enough) until we realise that we are not chatting with another human being. For a range of simple questions (location, opening hours, ticket prices etc.) we would probably not have known the difference. But for anything beyond this they usually create nothing else but frustration. Even as these chatbots gets far more advanced, those last five percent of difference to a human may still be challenging to resolve. The same may prove to be true for both driverless cars and pilotless planes.
Link to article:
Driverless Motor Vehicles: Not Yet Ready for Prime Time
There were autopilots in aircraft before WWII. All the large aircraft had an autopilot during the war. The RAF operated single pilot so it was a necessity. And yet, it was dispensed with by the pilot except for physical relief ; ie draining a distended bladder. Under attack definitely and certainly closing on their target, the pilots handflew.
Even up until the 1980s, the A310 and B767 were the only civil aircraft with sophisticated autopilots. It was sold to airlines as a way to save fuel by reducing trim drag ( from constant control inputs by human pilots). But still when the demand arises, pilots prefer to disconnect and handfly.
Thanks for the comment, Peter. Sorry for late response here. Further advances in automation will probably also be sold as cost savings but it is difficult to not worry that it sometimes is some form of pure technological ambition and adventurism behind it, i.e. it is seen as a technology challenge.
Thanks for an interesting post Nicklas. I totally agree – it is not as easy as some seem to believe.
I have been in involved in projects that are rather complex. As you near implementation, you typically see a design that is not perfect. There are a few problems and deficiencies. Still, you know you can deploy and be fairly confident it will work well. Why is that so? Because, there will be people in the system that will `finish the design`. People are so good at making non-perfect systems work with intended results.
I can´t see automation `finishing the design`, and for me that means you cannot deploy a fully automated system until it´s perfect.
Agree on that. But it is probably enough to be more perfect than a human operator. If that is the criteria autonomous operation may look more reachable.
But I´m not talking about perfect automation. I am talking about a perfect system where there is never a need for adaption, for trade-offs or for ´finishing the design´.
Such systems may exist today, but then they are probably not complex and perhaps not even too complicated…