AI + Transportation

AI + Transportation

Traffic pollution contributes to a large part of environment pollution, such as air pollution (greenhouse gas and road dust emission) and noise pollution. In urban aeras, traffic is one of the major sources of air pollution, which significantly affects our life. However, with the unprecedented development of artificial intelligence (AI), transportation is revolutionized, pollution can also be reduced. Signal control, navigation, and autonomous vehicle (AV) are several fields in transportation that are most related to our life and have already been cooperated with AI.
Traditional signal control uses fixed signal plan, which means that the duration and offsets of lights will be constant. Although fixed timing plan is stable and works for the most scenarios, it fails in some scenarios like congestion because it cannot change based on various traffic states. However, as AI can be involved, real-time signal control is possible. Real-time signal control changes light duration and offsets based on current traffic state and thus can be adaptive and flexible to resolve congestion. Google in 2021 launched a project that uses AI in signal control to make traffic lights more efficient. Apart from AI controlled signal that reduces congestion, AI also helps to navigate drivers (ie. Google map navigation) and then reduce congestion in one region. When detecting high traffic volume or density in one region, AI can propose different routine plan for nearby drivers such that traffic flow can be divided and balanced to prevent congestion. Faster traffic flow and less congestion reduces fuel usage and air pollution.
With the deployment of intelligent signal and navigation, the development of AVs can be accelerated. Since AV is controlled by computer, it accelerates and brakes smoother, which makes fuel uses more efficient and reduces energy consumption. Ideally, after all vehicles become AVs, vehicles together with signal and navigation can be fully controlled by AI, efficiency will be optimized, and congestion will be minimized.
With all the benefits of AI being said, the application is still challenging. Unlike classifying an image where making mistake is tolerable, any mistake in real-world application may be fatal, such as turning on the wrong signal light or failing to recognize a pedestrian in AV. Therefore, computer scientists and traffic engineers are still seeking better AI tools to accomplish cleaner and faster traffic.

CHENHAO ZHANG