Intersections are the most dangerous place on the road. Half of all crashes occur at intersections. Traffic lights attempt to meter traffic to help prevent the types of higher speed crashes that occur at rural intersections, but there are still a number of crashes that happen such as nose-to-tail, vehicle vs pedestrian and crashes caused by drivers running red lights. The causes of these accidents are almost always driver error and inattention but a number of factors was identified in this Monash University report that increase the likelihood of an accident.
The first traffic light was installed in London in 1868 but traffic lights in the form that we know them have been around since the late 1800s in America and since 1928 in Australia when the first one was installed in Melbourne. While they’ve developed in their artificial intelligence with the use of induction loops in the road and systems such as SCATS, there’s still huge room for improvement in both traffic throughput (i.e. the number of vehicles that can get through the intersection in a given time period) and safety.
In some parts of the world there are huge intersections that exist without traffic lights; big, multi-lane roads with a lot of traffic where drivers seem to find a way through, creating their own space.
As we can see in this intersection, traffic arriving simply turns into the intersection if there’s a gap, with the drive expecting that oncoming traffic will stop for them. At around 50 seconds a small jam appears but it quickly dissipates.
But those kinds of intersections have huge risks of accidents and they can’t cope with an enormous volume of traffic. At the University of Texas, Peter Stone (Professor of Computer Science) proposed a system called Autonomous Intersection Management. Vehicles approaching the intersection communicate with the intersection and one another and their speed and trajectory is adjusted so that they arrive just at the right time to fit in a gap in the traffic.
Recently a team from Massachusetts Institute of Technology (MIT) Senseable City Lab, the Italian National Research Council (CNR) and the Swiss Federal Institute of Technology (ETHZ) developed a slot-based system called Light Traffic.
Again, it’s a slot-based system which relies on communication between vehicles. The glaring problem is that the average age of a vehicle in Australia is 10.1 years! While the technology to accomplish what is proposed (i.e. fully autonomous vehicles) already exists, it hasn’t been implemented in a commercially available car or truck, let alone a motorbike. In order for it to work, all vehicles must be autonomous. But if the system was to operate, what kinds of data would need to be transmitted, and how would it work?
- Pedestrians and motorcyclists would need to be eliminated from the intersection – they are not autonomous and will always be directly controlled by a human that doesn’t know the bigger picture
- Each vehicle will need to transmit how fast it is approaching the intersection so that its speed can be adjusted, and what its final destination will be. The final destination is fairly simple: if the vehicle is autonomous it will be using a GPS and the entire route can be shared with the network to completely optimise traffic flow.
- The system will need to understand which lane the vehicle is approaching in. If it’s turning left or right, that’s simple. But if there are two lanes going straight ahead, which lane has the vehicle chosen? Its lane position will affect its position in the intersection where it will need to miss other turning traffic.
- Each vehicle will need to broadcast its characteristics in terms of length, acceleration and braking ability and cornering ability. The longer the vehicle the longer it takes to clear the intersection. The heavier the vehicle, the more likely it is to need to turn slowly, and should something unexpected happen, the vehicle’s braking characteristics need to be known in order for other traffic to be able to avoid it.
- As the system relies on microgaps, current automatic braking technology will need to be modified otherwise it will consistently think that an accident is about to occur and will slam on the brakes.
- The whole network will need to have awareness of the weather and road conditions. Wet or icy roads have less grip and this affects braking distance and cornering ability.
- Each vehicle will be scanning the road and intersection for something unexpected that might enter like a stray dog or a pedestrian. It will need to communicate this to the network.
- If a vehicle is having a mechanical problem which affects its performance, e.g .a flat tyre, the vehicle will need to know how to deal with that emergency, whether to carry on through the intersection or try to pull over. In rare cases the vehicle could break down completely.
- Finally, with all data-based transmissions, there are people that will want to hack the system. Each vehicle will need to be sure that any data received from other vehicles or from the intersection itself is genuine data and not spoofed data which could cause a serious crash.
So, in conclusion, having an automated intersection is a brilliant theoretical idea which could (if implemented across the entire network) almost completely eliminate traffic congestion, but there are a huge number of technical and social considerations that could render this idea unattainable for two to three decades.
Darren is an expert on driving and transport, and is a member of the Institute of Advanced Motorists
