The way in which technology appears to change society every few years tends to be sudden and unpredictable. However, most often than not, the core ideas that drive change are not by any means new. Like with most fields of research, the groundwork required for a technological breakthrough tends to have already been laid out over the course of years.
Despite the rise in popularity of the self-diriving car in recent years, the quest to automate vehicles is one that is more well established than people think. A great industry example where high level autonomous control is already being deployed is with large sea vessels such as cruise ships. In the case of navigation at sea, as of recent years, the number of crew members operating the vessel has diminished significantly. No longer is someone needed to perform complex calculations to predict the location and trajectory of the ship, all of that is handled by the onboard computer. The modern job of ship captains has become one of supervising the computer’s decisions rather than piloting.
One might naively think that programming a machine to navigate in open sea is rather trivial. Set a course to your destination, correct course according to weather conditions, avoid running into other ships… the margins for error are probably huge, right? What can possibly be so hard? Let me tell you about the cruise ships that travel through the Finnish archipelago. The amount of little islands in that area are uncountable, and the points at which bedrock hides at shallow depths beneath the sea surface would make a cartographer’s job a nightmare. Yet for decades, people who own summer houses in remote islands have gotten used to seeing 30m tall ships passing by their sauna’s window on the daily . It is crazy enough that these cruise ships were operating back in the day by relying on lighthouses placed on practically every island on the way, but if you think about harsh weather conditions that might occur and the low visibility that the crew have to deal with at night, you might start to wonder how it is even possible for humans to navigate through these waters.
So what if something terrible happens to the crew which prevents them from piloting the ship? Nowadays, not much. Thanks to technologies like GPS, these cruise ship are capable of completing most of the journey without human intervention. I can only assume the ship would eventually show up at the destination and the shipyard would eventually send someone to finish bringing it to shore (if that hasn’t by now been automated as well). The routes these ships take are very well defined, so the onboard computer just has to follow the GPS coordinates with as little deviation as possible.
Now, one might pose the argument that land vehicle traffic is much more complex (specially for a machine) than any environment a cruise ship might ever encounter. Ships don’t have to deal with pedestrians who suddenly decide to cross the road without paying attention, they have much longer periods of time to calculate appropriate courses of action. Furthermore, the surrounding geography is static, you don’t have an island suddenly appear one morning, and GPS is a tool that land vehicles cannot rely on in the same way. Overall, the environment a land vehicle has to deal with is not only highly dynamic, but also very chaotic.
There have already been attempts to make driverless public transport vehicles. These have been in the shape of small busses and shuttles that operate in relatively simple pedestrian environments and have harmless maximum velocities. Despite these examples having mainly worked a ‘proof of concept’, similar to cruise ships, they follow very specific routes and hardly ever deviate from them. Nonetheless, they show that automation technologies are only getting better. I believe it is only a matter of time before fully autonomous vehicles are deployed on roads, but for now advances will come in small incremental steps.
As an example, in order to justify automating trucks, we don’t need to have them be fully able to drive from the factory straight to the doors of a store completely on its own. Just like cruise ships, it is likely that the first version of autonomous trucks will come with a person supervising most of its journey from the driver seat. The vehicle could then require the driver to resume control once it transitions from highways to complex environments such as cities.
Supervision will only be present for so long though. Eventually, requiring a driver to oversee long journeys will be seen as an unnecessary cost. A company will instead prefer to employ local drivers to wait in the outskirts of a city for the truck to arrive and have the driver finish the journey manually.
But wait, why does the driver have to be compensated for the commute of going to pick up the truck? Couldn’t we equip every truck with a camera and have an operator control it through the internet? This would not only reduce the amount of work hours required from human drivers in terms of commuting, but it would also allow the autonomous vehicles to keep driving into populated areas for longer up until the computer feels it isn’t confidently able to keep driving. Controlling trucks from a distance also has further advantages. What if there is a road-block somewhere in the middle of nowhere and the vehicle is not able to improvise correctly? We could have a driver remotely connect to it and drive it manually at a distance. You can always equip the vehicles with satellite connection to the internet if the truck has to drive across remote areas. This process would be equivalent to how telephone customer support works today. Have a room full of trained employees waiting for vehicles to request human support. Once a given teleoperation is over, have the employee proceed onto the next vehicle waiting for human take-over in the queue. On top of all of this, every time an operator drives one of these vehicles, the sequence of actions taken can be recorded, adding to the endless library of existing driving data used to improve the algorithm. Before you know it, these autonomous vehicles will hardly ever need a human’s help.
If you think this idea sounds like science-fiction, think again. Maybe it has not yet been applied to self-driving vehicles, but companies out there already deploy teleoperation for similar purposes. An example of such practices is found in the shipyard crane industry. Crane manufacturing companies sell their products world-wide. They cannot afford to send the few engineers they employ across the world every time something appears to be going wrong. And I do not mean affording price-wise, it literally takes too long of a time to send an employee physically over to where the shipyard might be located. Shipyards lose millions in shipping costs every hour a crane isn’t working. The crane company thus has a room full of engineers in front of computers who connect to the crane through the internet, figure out what might be going wrong, and suggest a solution.
What if the solution happens to be replacing a component manually? These companies have also found a way to avoid having to send parts physically by mail. The cranes come equipped with 3D printers. It is faster to be able to 3D print most basic necessary parts at the crane’s location with the push of a button. The only thing they need is to have someone manually replace it. Imagine if automobile companies also took up the strategy of 3D printing parts. No longer would a specialized mechanic have to keep every possible replacement part in stock or having to wait days after ordering them. Vehicles could be equipped with sensors that allow for predicting if a component is likely to fail soon, and if it decides it will soon need a replacement, have it autonomously drive to the closest repair shop. The mechanic could be automatically notified ahead of time. Similar to how the teleoperators would connect to the vehicles, mechanics could remotely begin running diagnostics and begin 3D printing the replacement parts ahead of the vehicle’s arrival. Those automated trucks could be back in the road in no time.
These are only some of the first steps in the process of removing the human from the loop, but so far I have only mentioned examples of automation for the purpose of reducing risks and costs. The most exciting promise of autonomous vehicles comes from the potential to optimize traffic. The day vehicles are given the ability to communicate with one another, roads as we know them will no longer be the same. On my next entry, I hope to discuss some of the more futuristic applications of autonomous vehicles.