Why AI traffic modeling is only as smart as the data we feed it
- Pinnacle Team
- Sep 4
- 2 min read
Accurately forecasting future traffic patterns and reducing congestion are critical challenges that experts around the world face. While advances in AI and virtual reality (VR) are revolutionizing traffic simulation, data acquisition and processing remain significant hurdles to overcome.
Traffic modeling has grown into a massive industry. According to a report by Kings Research, the global traffic modeling market is projected to reach $7.83 billion by 2030. This growth is supported by active research from government agencies, such as the U.S. Federal Highway Administration (FHWA), academia, and the expansion of the commercial market.
The FHWA has been a leader in developing traffic simulation models since the 1970s. By the 1990s, they packaged these models into a single system called the Traffic Software Integrated System/Corridor Simulation (TSIS/CORSIM), which facilitated the transfer of technology between the government and private sectors. Karl K. Andersen, Acting Director of the Office of Safety and Operations Research and Development at the FHWA, stated, "Thanks to the continuous advancement of computing power, traffic modeling can now analyze more complex road networks and consider various modes of transportation."
Data: The Core of Traffic Modeling
In the past, traffic data was centrally managed by public systems. However, today's world allows for data collection from various sources, including cellphone data, infrastructure cameras, and satellite data. Nevertheless, access to OEM (Original Equipment Manufacturer) data remains an international issue, as vehicle manufacturers often do not provide the data they collect.
Claes Rydgren, a professor of transport informatics at Linköping University in Sweden, pointed out, "Connected vehicles have great potential to provide traffic information, but car manufacturers often do not make the data available for use in simulations." In the U.S., the FHWA is collaborating with academia and the private sector to improve access to and security of new data sources.
Data Quality and the Potential of Cellphone Data
Data quality is also a crucial factor in the accuracy of traffic models. Professor Rydgren regularly discusses ways to improve sensor systems and data quality in collaboration with the Swedish Transport Administration. He explained that "the possibility of collecting data directly from cars, in addition to traditional radar or vehicle counting systems, will play a significant role."
Anonymized cellphone data is also emerging as an important resource. Telecommunications companies like Telia have already achieved success by collecting mobility data within Sweden and releasing commercial products. Although still in its early stages, this technology is expected to add a new dimension to future traffic modeling.
The Application of AI in Transportation
Artificial intelligence (AI) has been evolving in the transportation sector for a long time. A 2019 report from the Delaware Department of Transportation discussed applying AI to accident detection, ramp metering, and drone usage. AI is now evolving beyond static statistical models into real-time learning systems, playing a role in optimizing traffic flow in real time.
The continuous advancement of AI and machine learning is changing the paradigm of traffic management. The potential of this technology to more precisely control and predict complex traffic systems raises expectations that future transportation systems will become smarter.
May 24, 2024
