The Oldest EVs Had An Excellent Vision Driving System We Can Learn From Today
While it isn’t mandatory or required by the technology, it seems like electric vehicles and vehicles with driver assist (or upcoming autonomous vehicles) go hand-in-hand. This is probably in part due to Tesla’s deep involvement in it, and in part caused by the historical coincidence of both technologies becoming popular around the same time. It may surprise readers to know that in the 1910s, a similar thing occurred.
After doing some research, it turns out that several EV manufacturers created a fully autonomous system for their vehicles where the owners or people paying a fare didn’t have to do a thing once they sat down in the vehicle. A quick voice command, and they were off and away on clean electric power.
A Complicated & Expensive System
As you could imagine, a system that could do all that over 100 years ago would be quite complicated, and only affordable by the wealthy. Even they couldn’t purchase one in most countries, and had to make do with renting them.
The mounting hardware for the system was, like today’s Waymo vans, on the top of the vehicle. This allowed the system’s sensors to get maximum visibility of the road while keeping it out of sight of the passengers. How charming or ugly the system appeared differed from installation to installation, but in general, they wanted the system out of sight. To ensure longevity of the system, a vibration-absorbing material and bracing was used to keep road vibrations from causing problems.
The system itself was rather ingenious, but some think its development was not only long in the making, but also an accident of sorts. Over a considerable time period, the different sensors were developed. They included vision sensors, sonic sensors, steering feedback and vibration sensors, inertia sensors, and even weather detection and prediction. The system used an ancient form of neural networks (the oldest kind, in fact) to process all of this data and ensure a safe ride.
Unlike today’s under-development autonomous vehicles, it only had two optical sensors. However, they were quite advanced. Mounted on a very flexible gimbal and to a limited extent individually aimable, the system used a clever hydraulic arrangement that could help the neural net cope with balancing and even inertia. This allowed the system’s visual sensors to rotate and see different objects of interest, often when the sonic sensors detected something.
There was even an advanced vehicle-to-vehicle communication system. Using sonic vibrations or optical signaling, the system in one vehicle could send short messages from one vehicle to another, even among different brands of vehicle. This aided in the navigation of difficult intersections, arranging for parking, and even a “summon” feature.
The system could even interact with telephone and telegraphy networks of the day to arrange rides, and extended exchanges of information often occurred when two of the vehicles were parked near each other waiting for the passengers to return. This, of course, led to better route planning and even better internal mapping of the road environment. Information about what routes were impassible among many other things likely was transferred.
An owner could even program the system to leave them behind and come back later when they were ready to be picked up. It was that advanced. During these off-times, the system could navigate to, arrange for, and park itself at a charging station. If necessary, the system could plug the vehicle in using hardware not that different from your own arm and hand.
Now that we’ve described the advanced capabilities of this system, let’s talk about who made it. It turns out that women were heavily involved in the manufacture of these systems, and they were each built one-by-one. Each one took months to assemble, and then years to program and prepare for service. These systems were so complex, that many didn’t make it past the early programming phase. About 13% of attempts failed within the first year.
Given that each was built individually, and so much time was needed, getting one to install on one’s vehicle was expensive.
There was another downside: these systems were very good, but would sometimes fail. In fact, legend has it that the first two installations in one state somehow managed to crash into each other due to some sort of miscommunication. The accident rate was, in fact, similar to human drivers today.
A Photo of the System
Now that we’ve talked about this system, let’s look at a rare photograph of one from the German Federal Archives:
This Shows Us Why It’s So Hard To Develop an Autonomous Vehicle
While my description of a human chauffeur was somewhat deceptive (the driver’s position atop the vehicle was a great way to throw you guys off), there’s an important point to be made here.
This shows us just how complex human driving is. We have a bunch of very advanced components, some of which evade human understanding. We don’t know what consciousness really even is, or how it develops in the brain. We know a lot about how small structures of the brain works, but we don’t know exactly how it all works together.
Outside of the brain, our eyes, ears (including parts that help our sense of balance), neck, mouth, and even our skin all play important roles in the driving task. Each of these systems took billions of years to develop, even if there wasn’t a developer (that’s a topic we’ll leave for debate elsewhere).
Even the most well-funded projects, like those at Tesla, are finding that it’s a lot harder to achieve even the functionality of a car in 1904 with a hired chauffeur than anticipated. Many AV fans think of human driving as primitive and in dire need of replacement, but it’s actually a very complex system that we don’t even fully understand.
Expecting developers to make a better system quickly isn’t realistic. We’re going to have to be patient.
Featured image: a crop of a 1912 Detroit Electric advertisement (public domain).