Technology Article

Autonomous Vehicles and the Problem with Speed

Sensible 4 autonomous vehicles have maximum speed of about 30-40 km/h. Yet expressways in many countries have speed limits of 110 to 130 km/h. Would this kind of speed possible for autonomous vehicles?

Sensible 4 autonomous vehicles have a top travel speed of about 30 to 40 km/h. This works well for public transport in the city and is enough to avoid holding back other traffic. Cars can, of course, travel much faster than this. Expressways in many countries have speed limits of 110 to 130 km/h. But is this kind of speed possible for autonomous vehicles?

This is not a simple question to answer, and two factors are particularly important: the Operational Design Domain (ODD) and the level of autonomy

ODD is about the type of events and objects the vehicle may encounter on the road. In an urban environment, a vehicle can encounter anything from cars and kick scooters to pedestrians. On the other hand, you can expect other road users to not travel at excessively high speeds, such as over 100 km/h. 

On an expressway, expectations are different: you expect no slow-moving road users, such as pedestrians, tractors, or bicycles. There is also no oncoming traffic in your lane or in the lane next to you, and there are no traffic lights.

Level of autonomy refers to a classification system with six levels widely used in the industry, with level zero indicating an entirely human-driven vehicle and level five a fully autonomous vehicle requiring no human interaction.

Driver Assisting Technologies Typically Work Well

Various advanced driver-assistance systems (ADAS) available to cars, including lane assist and adaptive cruise control, work well when driving in a limited ODD and there are no unexpected traffic events. Drivers using these systems often feel that their vehicle is practically autonomous, driving by itself on a number of different roads.

The issue is, however, that unexpected events in traffic are inevitable. These technologies also don’t work well in bad weather, such as on a slushy winter road. They also find deviations from regular conditions challenging, such as road work.

With lower-level autonomy (levels 1 to 3), the human driver is always in charge of driving the vehicle. In other words, advanced driver assisting technologies can keep the vehicle in the right lane in favorable conditions and the driver just takes care of special circumstances. The driver, therefore, needs to watch traffic constantly and stay alert and be ready to assume control at all times.

With higher-level autonomy (levels 4 and 5), the vehicle must handle all traffic events autonomously with no human interaction. At level 4, where Sensible 4 vehicles operate, the vehicle has the ability to come to a safe stop when moving forward is no longer safe. Bringing the vehicle safely to a stop is a key issue, especially at higher speeds.

Slippery Roads Extend Braking Distance

Autonomous driving systems try to anticipate traffic 2 to 3 seconds into the future and bring the vehicle to a stop when necessary. For example, in the spring 2020 FABULOS pilot in Helsinki, the maximum driving speed was about 30 km/h. At this speed, it’s fairly easy to bring a vehicle to a stop in 1.5 to 2 seconds on dry asphalt without compromising too much on passenger comfort when braking. Harder braking would be possible if necessary but could pose a safety risk for passengers.

Since opposing traffic is expected to travel at approximately the same speed as the autonomous vehicle, the sensors need to scan traffic at a distance of about double the vehicle braking distance. At 30 km/h, the braking distance is only about six meters. That means that vehicle sensors need to scan other road users in a radius of about 12 meters, drawing the required conclusions and steering the vehicle accordingly. 

In bad weather, such as on a snowy road, braking distance and time needed to stop can easily double. This requires the vehicle to be able to reliably scan at a distance of about 24 meters at 30 km/h, while keeping the braking time within the 2 to 3 second projection.

At higher speeds, braking time exceeds the three-second projection, braking distance increases significantly, and reliable scanning becomes difficult. 

When braking time exceeds three seconds, predicting traffic becomes increasingly important. Whereas sensors typically see better than humans, and a human is no match for a computer’s reaction speed, humans are typically much better at reading traffic and predicting potentially hazardous incidents. Understanding and predicting traffic at high speeds is difficult for computers. 

At High Speeds, On-Board Sensors No Longer Suffice

An extreme example could be braking on a slippery road at 120 km/h. This would amount to a braking distance of about 250 meters, requiring the vehicle to scan its surroundings up to a distance of 500 meters. With current sensor technology, this is practically impossible. For vehicle motion planning, making reliable predictions for a braking time of up to 13 seconds is an insurmountable task.

So, with autonomous driving at high speeds, we are right back at the idea of Operational Design Domain. If we can trust that nothing unexpected will happen over the upcoming 500 meters, we can drop the required three-second projection. This would make autonomous driving at high speeds possible. 

These circumstances could exist, for example, on an expressway with game fences on both sides and some type of fixed sensoring for ensuring that other road users travel in the same direction with no disruptions.

Sensible 4 is involved in one such initiative. The LuxTurrim 5G ecosystem project creates guidelines for future smart city infrastructure. For example, intersections could be scanned with fixed laser scanners and information relayed to autonomous vehicles via high-speed 5G networks. The first of this type of base station equipment has been installed in Espoo’s Kera region in Finland, close to the Nokia headquarters.

Another oft-discussed option is to enhance communications between vehicles by developing so-called vehicle-to-vehicle (V2V) technologies. The vehicle in front would share its sensor data with the vehicles behind it, which would improve traffic predictability at high speeds. But this technology, too, requires developing 5G technologies further.