How AI helps transport

Artificial intelligence is being used for imaging and text message generation in the transport sector, but in a different way than in popular generative AI services.

Published January 30, 2023

You may have already created awesome pictures and some clever paragraphs in generative AI services like ChatGPT and Dall-E, as we have created some fun ideas of what a bicycle airbag or other road safety device would look like. We truly appreciate the mind blowing creativity at the intersection of art and transport, whether it’s done by computers or humans. For the latter, our favorite is the Bull’s Head made by Pablo Picasso.

The capabilities of artificial intelligence can be used in many other areas of transport technology, with less focus on entertainment and more on safety. Here are some examples of areas where the industry prefers to use AI.

  1. Scene recognition
  2. Crash likelihood prediction
  3. Traffic optimization

Scene recognition is a complex process including object detection, tracking and classification. With semantic segmentation smart sensors label each pixel according to the object to which it belongs, such as “sky”, “vehicle” or “pavement”. 

GPUs and TPUs brought massive parallel processing capabilities to road safety sensors and cameras to accelerate the training and inference of neural networks that are built on top is basic computer vision methods. Accurate and rapid detection of a large number of objects is key for the V2X ecosystem. Our Roadside Units need reliable information to formulate V2X messages about what’s happening on the road, so connected cars can display relevant alerts in front of the drivers. This helps in dealing with critical situations.

Roadside and on-board sensors provide data with different levels of confidence. The information therefore needs to be validated using multiple entities to make sure they see the same thing we see.

Crash likelihood detection with AI is a whole new level in V2X. Once we have understood the scene with the help of scene recognition, we need to use this information to predict and prevent potential collisions. One such measure could be to send a V2X alert to the driver about the need to change the speed of the vehicle.

Artificial intelligence algorithms for predictive modeling can build on a wide range of information sources. To predict the probability of a crash, they can analyze data from on-board and roadside sensors, cameras and other sources such as traffic patterns, weather conditions and driver behavior. V2X has the unique ability to provide real-time information about an emergency braking well ahead of us in our lane, or if there has been a collision on our route.

AI also plays an important role at the macroscopic level, in traffic optimization. Only V2X-enabled vehicles and the connected smart infrastructure can provide a realistic traffic environment description for real-time traffic management. AI-algorithms can adjust traffic signals based on actual traffic conditions, road closures, historical traffic data and weather information to reduce congestion and the risk of collisions. 

Besides managing traffic lights, intelligent traffic management systems are able to recommend optimal routes for delivery trucks, taxis and buses to reduce travel time, save fuel, and improve the overall efficiency of public services. V2X-based traffic management can enable cooperation between vehicles over a larger area, which a single self-driving car would not be able to do on its own.

Ultimately, these applications of AI can be just as entertaining as a computer-generated image: we can live in a healthier environment and be safer, spending less time on the road.