Research on Traffic Congestion Area Identification and Traffic Control Based on Multi-source Data Fusion

In the bustling arteries of modern cities, traffic is more than just a collection of cars; it is a complex, living system where a single blockage can ripple outward, turning a smooth commute into a gridlocked nightmare. To manage this, city planners have long relied on counting vehicles or watching cameras, but these methods often see only a fraction of the picture. A newer approach involves gathering many different kinds of information at once, much like a doctor might combine a patient's temperature, blood pressure, and medical history to get a full diagnosis. By merging data from the GPS in ride-hailing cars, the magnetic sensors buried under roadways, and the signals from mobile phones as people move through a city, researchers can build a much clearer, real-time map of how traffic actually behaves. This fusion of data sources allows for a deeper understanding of where traffic jams start, how they spread, and when they will clear, offering a path toward smarter, more responsive city management.

In a recent study focused on the Hongjialou area of Jinan, China, a team of researchers set out to apply this multi-source data approach to solve the persistent problem of urban congestion. Their goal was not just to spot where traffic is slow, but to predict exactly how a jam will grow and to test specific strategies to ease it. The researchers began by dividing the city's road network into a grid of small, square sections. They then fed data from three distinct sources into a computer system: the movement paths of online ride-hailing vehicles, the flow of cars detected by magnetic sensors under the pavement, and the movement patterns of people derived from mobile phone signals. By carefully aligning these different streams of information, they created a unified, high-resolution dataset that captured the speed, volume, and density of traffic with far greater precision than any single source could provide.

With this rich dataset in hand, the team developed a method to identify exactly where traffic jams were forming and how they were likely to spread. They used a clustering technique to group together grid sections that shared similar traffic behaviors, effectively highlighting the specific neighborhoods where congestion was most severe. To understand how a jam in one block might cause a jam in the next, they built a model that calculated the probability of traffic slowing down spreading from one grid to its neighbors. Their analysis of the morning rush hour revealed a clear pattern: congestion did not happen randomly. Instead, it often began in specific grid sections, such as those near major university campuses, and then propagated outward in a chain reaction. In one observed instance, a jam that started in two specific grids spread to cover nearly two-thirds of the study area within fifteen minutes, with a high probability of moving laterally to adjacent streets.

To predict these trends before they happened, the researchers designed a sophisticated computer model that could learn from the past to forecast the future. This model combined two powerful types of artificial intelligence: one that recognized spatial patterns, like the layout of a busy intersection, and another that understood time sequences, like the way traffic builds up over the course of an hour. They added a special feature that allowed the model to focus its attention on the most critical moments and locations, ignoring less important noise. When tested against real-world data, this new model proved significantly more accurate than existing methods. During the chaotic morning and evening rush hours, it reduced prediction errors by roughly a quarter compared to older models, successfully forecasting traffic conditions with about 94% accuracy. This level of precision means that traffic managers could potentially receive reliable warnings about upcoming jams well in advance, rather than reacting only after the roads are already clogged.

The final step of the research was to see if these insights could actually improve traffic flow. The team used a simulation platform to create a digital twin of the road network, allowing them to test different traffic control strategies without disrupting real drivers. They compared the current, unoptimized traffic signals against a new plan generated by an algorithm designed to find the most efficient distribution of traffic. This algorithm worked by constantly testing small adjustments to traffic flow and signal timing, gradually refining the solution until it found the best possible arrangement. The results were tangible: the optimized plan reduced the average time a vehicle spent traveling through the area by 12.5% and shortened the length of vehicle queues by more than 22%. Furthermore, the total time cars spent waiting at red lights dropped by nearly 18%, and the number of vehicles passing through the network increased by 7.6%.

While the study demonstrated that fusing multiple data sources with advanced prediction models can significantly improve traffic management, the researchers acknowledged that their work is a starting point rather than a final solution. The data used covered only one week in the spring, and the tests were conducted in a specific area with a grid-like road structure. The team noted that the system would need to be tested over longer periods, across different seasons, and in cities with different road layouts to ensure it works everywhere. They also pointed out that the complex computer models act somewhat like a "black box," making it difficult for human managers to understand exactly why a prediction was made. Despite these limitations, the study provides a compelling proof of concept. It shows that by listening to the many different signals a city emits, planners can move from guessing at traffic patterns to understanding them with clarity, paving the way for cities that manage their own flow with greater intelligence and efficiency.