Don’t miss out on the latest CyclingTips updates.
Whether we like it or not, driverless cars seem to be on their way. It might still be many years until the technology is ubiquitous on our roads, but when automated vehicles (AVs) do arrive, they’re going to shake things up in a huge way.
One of the biggest benefits AVs are expected to bring is improvements to road safety. Human error is the biggest cause of road trauma — remove human drivers from the equation and the number of those injured or killed on our roads should decrease significantly.
As vulnerable road users, cyclists can rightly feel a sense of excitement at the prospect of driverless vehicles. Such cars won’t be controlled by smartphone-distracted drivers, drivers who actively hate cyclists and drive accordingly, or those who simply aren’t confident driving around cyclists.
Indeed, the AI behind driverless cars tends to be designed from the ground up with vulnerable road users in mind (see video below). It could be that the rise of AVs is what we need for cycling safety to progress in a meaningful way.
Of course, speculation like this is good in theory but will AVs actually make things safer for cyclists? We won’t know that for sure until such vehicles start to become the norm, but thanks to some research out of Belgium we do have a bit of an idea of how things might turn out. And the signs are good.
To test whether AVs make things safer for cyclists, a team of Belgian and Dutch researchers created a simulated version of a real-world urban area full of cyclists and cars. More specifically, they used traffic flow simulation software called PTV Vissim 11 and a software application called the Surrogate Safety Assessment Model (SSAM) to evaluate the safety implications of their simulated environment.
For their sample space, the researchers created a virtual replica of central Hasselt, a city of roughly 70,000 in north east Belgium. As the researchers write, “Due to its size, the flat terrain, and an extensive cycle network that runs throughout the city, cycling is a popular mode of transport among [Hasselt’s] citizens.”
To keep the simulation manageable, the researchers reconstructed only a small section of the city centre: a region containing five straight sections of road and two intersections without traffic lights.
The researchers were determined to have cars and cyclists in their simulation behave as realistically as possible. To that end they took local observation data and combined it with data from a Danish project called “Microsimulation of Cyclists in Peak Hour Traffic” to help provide baseline data for their simulation.
From their observations they were able to input average cyclist speeds and program the way that drivers and cyclists behave around one another in Hasselt. They found that cyclists tend to leave less space when following other cyclists compared to when they’re following cars. They found that cyclists give more space when overtaking cyclists than they do when overtaking cars (given a lack of available space on the narrow, inner-city roads). And they found that drivers left more distance between themselves and a cyclist ahead than they did when a car was ahead.
In setting up the simulation the researchers also had to decide how they wanted the AVs to behave. PTV Vissim offers three different behaviour models, all based on real-world data: Cautious, Normal, and Aggressive. The researchers opted for the most cautious behaviour, instructing AVs to maintain large gaps between other vehicles and always adopt safe behaviours. They felt this was the most realistic option for Hasselt given cars and cyclists in the city centre share the same lanes, there are no traffic lights at the intersections, and the simulated region was an urban area with a high volume of traffic.
The AVs were set to obey the local speed limit of 30 kph on straight sections, with their speed dropping by 5 kph in turns.
With their baseline parameters sorted, the researchers then set about creating two simulations. The first: a replica of the current situation in Hasselt, where human-driven vehicles and cyclists co-exist. The second: the same system but with human-driven vehicles replaced entirely with AVs. By comparing the two, the researchers were able to simulate the impact of AVs on road safety, both in terms of the number of conflicts, and the severity of those conflicts.
So how did the researchers define a conflict? In line with previous work in this space, they used a variable called time-to-collision (TTC): the time required for two vehicles to collide should they continue at their present speed and on the same path. The lower the TTC, the more severe the crash. As per industry standards the researchers took any incident with a TTC of less than 1.5 seconds to be a conflict — a likely collision.
The researchers then ran the traffic simulation 10 times to generate a range of outcomes, using the safety assessment software to determine how many conflicts occurred and how serious those conflicts were.
The researchers found that in the AV-only simulation, the average speed of cars increased slightly and delays across the system were down slightly.
“The modeling of car traffic flow consisting of only AVs improved the network performance by reducing the average delay from 3.83 s to 3.78 s and increasing the average vehicle’s speeds from 18.37 km/h to 18.88 km/h,” the researchers wrote in their paper. “Cyclists performance measures also presented minor improvements in that case as the average delay was decreased from 1.1 s to 1.04 s while the average speed was increased from 16.7 km/h to 16.73 km/h.”
These aren’t significant improvements for either vehicle type, but the more important focus was on safety. In that regard, the researchers found some more promising numbers.
“The results in the second scenario [AVs and cyclists] show a reduction in the total number of conflicts for the entire network from 120 to 106 but also in the number of conflicts between cyclists and cars, from 72 to 67,” they wrote. “The main reduction of conflicts in the network was noticed in rear-end collisions which were the most frequently observed type of conflict.”
More specifically, the researchers noticed a reduction in the number of conflicts at intersections — the place where most conflicts occurred. There “the total number of conflicts [was] reduced from 62 to 46, while the number of conflicts involving a cyclist and a car from 34 to 25.”
In terms of the severity of conflicts, the researchers noted a significant reduction in incidents with a time-to-collision of less than or equal to 1.5 s, both overall (from 88 incidents down to 72) and specifically at intersections (from 42 down to 30).
Where that leaves us
We’re many years away from AVs finding their way onto our roads in any serious numbers and several generations until they become the dominant vehicle type (if that even happens). Whether they ever come to represent 100% of all cars on the road — as simulated in this study — is another question entirely.
Regardless, when AVs do start to increase in number, it appears they have the potential to bring along safety improvements for vulnerable road users, even if those improvements are modest.
“The findings confirmed that AVs have the ability to improve road safety and network performance,” the researchers wrote. “From the analysis of the results, we can directly observe that the introduction of AVs in the road network reduced the total number and severity level of conflicts.
“The total number and severity level of conflicts between cyclists and cars were also decreased while minor improvements were observed in their traffic performance.”
Of course, despite the researchers’ best efforts to replicate real-world Hasselt as accurately as possible, this simulation is far from perfect.
Perhaps the cautious logic applied to AVs in this simulation isn’t realistic — perhaps it comes at the detriment of traffic speed and as such, a slightly more aggressive behaviour would be adopted by manufacturers. At the very least, there are certain to be differences in the AI of different AV makers, making the road networks of the future quite different to the homogeneous system created here. Might the owners or passengers of AVs also have the ability to somehow adjust the behavioural characteristics of the car they’re in?
There’s also the behaviour of cyclists to consider. In a future where AVs are the dominant form of vehicle, will cyclists feel safer and therefore ride less cautiously than they currently do? It certainly seems possible.
And as mentioned above, this research assumes 100% adoption of driverless technology. At best, this is many decades away; at worst, it will never happen. In a system featuring even some driver-operated vehicles, any benefits of an AV-dominant system are sure to decrease.
The researchers admit that more work is required to find the optimal configuration of parameters for their simulations, “as their values can highly influence the results of the simulations and the investigation of cyclists’ behavior when they encounter AVs.”
More broadly, riders can only hope that improvements in cycling infrastructure keep pace (or even outpace) the development of AVs, thereby increasing cycling safety above and beyond what might be possible with driverless vehicles. There are big gains to be made through clever use of cyclist-minded infrastructure, so too through better driver education, harsher penalties for distracted driving, and a general appreciation for the rights and safety of cyclists. This is where cyclists should be looking to ensure their safety.
If the introduction of driverless vehicles helps to make things a little safer as well, well, that’s a bonus.