After having tested an autonomous bus recently, I also got a chance to try a car that gives a taste of autonomous driving. It happened by chance. When I rent, I typically choose a small car, because small cars are big enough for my needs and they are the most economic, both in terms of rent and consumption. This time I booked a category called “mystery vehicle” where you get what is available at the time of pick up. The price corresponded to what I usually pay for a small car, so in the “worst” case I would get what I paid for, while there was a possibility to be surprised with something better. I had no idea how surprised I would be :-).
I don’t remember the last time I got excited about a car I rented. Now I got a brand new Volvo XC40 (it had 700 kilometers on the meter), a compact SUV with top of the line equipment, where the most important in the context of this blog post were adaptive cruise control and pilot assist. The former keeps the desired speed while adjusting it to the speed of the vehicle in front of you. The latter turns the wheel automatically and follows the lane, by scanning the surface lane markings (whether it also takes advantage of the built-in road maps, I don’t know). These two technologies in combination are enough to experience autonomous driving (albeit in a limited scenario – it works best on the highway).
It felt very strange. I could not relax completely, because as a software engineer I have witnessed all kinds of software bugs, even in production-grade software. I also briefly worked with computer vision and know how difficult it is for a computer to recognize and follow a desired object under varying light conditions. To be completely fair, as adaptive cruise control and pilot assist fall under the category of safety-critical applications, they use redundant algorithms, and are tested rigorously and certified, but even if they are to be considered 100 percent trustable, the infrastructure is not – lane markings occasionally disappear. So there I was, not touching the pedals, not holding the wheel but having my hands tremble hovering above it, feeling awed and scared at the same time. I consider that to be a pretty dangerous state, because you are not at full concentration, while at the same time the system is not completely autonomous and as such requires supervision and occasional action (when the car looses track of the lane markings). Commercial pilots are trained to be concentrated while observing, but for drivers this is unknown territory. I felt that my driving education and experience suddenly became insufficient and started to imagine a new kind of driving school where, in addition to the classical driving education, you use the car in autonomous mode while a services engineer employed by the car manufacturer explains the limitations of the system and how to deal with them them. After that, you train how to stay concentrated while not actively driving.
But at the same time, I could definitely see the potential – once we reach the level where we can trust autonomous cars to a sufficient extent, we will suddenly get a lot of time which we can use more productively – for reading, working, learning or just relaxing so we can reach our destination fresh and rested. Not to mention how many lives will be saved by reducing the number of traffic accidents. There is still work left before we get there, but these are exciting times.