Improving Our Road Infrastructure for a Self-Driving Future and a Safer Now

The Gap in Autonomous Self-Driving Vehicle Development, Part 2

In the first part of this three part essay series I proposed a reason why there has not been more progress in the development of self-driving vehicles, despite a few dozen companies throwing billions of dollars at the problem. I likened it to trying to force a square peg into a round hole. There are two obvious ways of solving that problem, make the peg round, which is what the aforementioned companies are trying to do; or make the hole square. This essay will focus on the latter solution, and in this scenario the round hole is our road and highway infrastructure.

“The reality is, today’s infrastructure is not well equipped for autonomy. In essence, potholes, poor lane markings, and all the other crumbling aspects of our nation’s infrastructure aren’t going to support high-tech well.” (Business Insider)

Accidents generally occur because something unexpected happens. Autonomous vehicle manufacturers are trying to fit self-driving cars into an existing road infrastructure system that is poorly maintained and has too much unpredictable behavior. In Part 1, I introduced an equation to calculate unique combinations, or scenarios, based on the total number of variables, n, and how many variables can interact at a given moment, r. If you put a self-driving vehicle into a controlled environment, with low n and r values, they can perform safely and reliably. Instead of adding more cameras and sensors onto vehicles and spending more time teaching the AI, we should shift some focus onto improving the road infrastructure. This will create an environment more suitable for self-driving vehicles and safer for everyone on the road.

Figure 1: Restricting pedestrian access to roads will reduce accidents like this.

There are three main areas where infrastructure improvements should be focused: road conditions, road access and bicycle lanes. Poor road conditions can cause unpredictable behavior such as cars slowing for uneven surfaces or swerving to avoid a pothole. This also includes not having adequate lanes for merging, or suitable notification for road construction or lane closures. States and cities already have agencies responsible for the roads, but they need to do a better job of maintaining them. We should not have to worry about the condition of the roads when driving. If we can ensure the roads are properly maintained and marked, it is one less subset of variables to worry about.

The next area is road access. In 2018, 6,823 pedestrians were killed in traffic accidents. Additionally, the CDC “estimated 137,000 pedestrians were treated in emergency departments for nonfatal crash-related injuries” in 2017. It is essential that we limit road access to only vehicles, except at designated locations like crosswalks. By doing this we can greatly reduce the possibility of pedestrians, animals and other objects like shopping carts or balls from entering the road unexpectedly. Doing this can practically eliminate accidents involving pedestrians, such as what is depicted in figure 1 above. Taking another look at the equation for combinations, removing pedestrians from the equation will reduce the number of variables that a self-driving car will encounter. This is the previous equation we looked at,

The population size for this equation, n, or the number of variables, was 55. If we remove the 33 people we counted, our new n will be 22. We can also reduce the r value, which is the number of variables that can act unpredictably at any given moment. Since pedestrians can be very unpredictable, let’s reduce the r value to 6. The new equation looks like this:

That is a huge reduction (0.00025%) in the number of potential scenarios a self-driving vehicle can encounter. An environment that is more predictable will allow an AI to perform more reliably.

To limit road access, cities will need to construct barriers or fences along all roads. Electronic gates can be placed at intersections to control foot traffic and limit it to designated areas only. For busy streets, or large intersections, bridges and walkways can be constructed, similar to what cities like Shanghai have done, see figure 2 below. Alternately, tunnels and subterranean walkways can be used where suitable, like what currently exists in subway systems. To handle curbside parking, a gate can be installed at each parking space, or on each block, which will control access and handle payment. This should not require major changes to the existing infrastructure, and it will make the roads much safer right away.

Figure 2: Large elevated pedestrian walkways, like in Pudong, Shanghai, China, help pedestrians safely navigate busy intersections

The last area for improvement is to provide dedicated bicycle lanes. Bicycles typically share the road with vehicles, they can also be unpredictable. It will be challenging to create dedicated bicycle lines throughout a city without the ability to widen roads. A short term solution may be to restrict bicycles only to roads that already have dedicated bicycle lanes. Longer term, separate bike paths will need to be constructed. It will require planning and creativity but it is something we must work towards to increase bicycle safety, and make the roads more conducive to self-driving vehicles.

A major factor in realizing these improvements is how they will be funded, since it will not be cheap. A possible solution is to increase the cost of vehicle registration. According to Statista, in 2018 there were 273.6 million vehicles registered in the US. If we add a $20 fee to each registration renewal, that is almost 5.5 billion dollars for infrastructure improvements each year. We can also add a fee onto car rentals, taxis and ride-sharing, since they typically drive the most miles. Another option, although potentially less popular, is to raise the cost of public transportation. We can also require any bicycle ridden on public roads to be registered and licensed, similar to other vehicles. This can help fund the construction of bike paths. These fees don’t have to be a lot, but they will add up to billions of dollars in funding. If that is not enough, states can increase the cost of permits to test autonomous vehicle, since, in the long run, these infrastructure improvements will save the companies involved both time and money.

The improvements discussed here are key to creating a predictable environment where self-driving vehicles can operate safely and reliably. This it is how we make the round hole more square, and more suitable for self-driving vehicles. While a completely autonomous, self-driving future may still be many years away, the benefits of making these improvements will have an immediate impact on society with safer roads and a large reduction in pedestrian and bicycle related vehicular accidents.

An additional benefit is the potential to create many new jobs, think of it as a modern version of Franklin Roosevelt’s New Deal. Making these improvements will create jobs across many different industries and will require a wide range of skill sets and experience levels. There can also be incentives to use American-made products and equipment. This is something we should start working towards right away, even more so now to help us get out of this economic recession and combat the growing rate of unemployment, which Fortune claims is likely the highest since, you guessed it, Franklin Roosevelt was President.

The next, and final, essay will focus on the need for a standard communication protocol and operational framework for how self-driving vehicles will operate and interact together. This is the final piece to get the square peg to fit in the round hole.

Click here to read Part 1.