Highways are complex systems where the behavior of thousands of individual drivers combines to create the flow of traffic. When too many cars try to use a road at once, the system can jam, causing delays that ripple backward for miles. For decades, transportation agencies have relied on simple measurements to judge how bad these jams are, such as counting the total minutes lost by all drivers or measuring how much slower traffic moves during peak hours. These numbers tell a story about the past: they describe how congested a road was yesterday or last year. However, they offer little guidance for the future. A high number of lost minutes might mean a road is simply overwhelmed by too many cars, a situation that no amount of tweaking can fix. Alternatively, that same high number might mean the road is clogged because drivers are reacting poorly to each other, a situation that could be cleared up with the right management. Without knowing which scenario is true, city planners risk wasting money on expensive projects like adding new lanes, when a cheaper, smarter solution might have worked just as well.

This uncertainty is the central problem addressed by a new study from researchers at the Massachusetts Institute of Technology and the University of Nottingham. The team set out to create a new way of looking at traffic that goes beyond simply measuring how bad things are. Instead of asking "How much delay is there?", they asked a different question: "How much of this delay could we actually fix?" To answer this, they developed a concept they call "controllable congestion." This is not a measure of current traffic conditions, but rather a measure of potential. It calculates the maximum amount of delay that could theoretically be removed from a highway if an intelligent system managed the speed limits perfectly. By using advanced computer simulations that mimic how traffic flows, the researchers tested this idea on both a made-up highway scenario and real data from a stretch of Interstate 24 in Tennessee. Their work reveals that the amount of delay a road suffers is often unrelated to how much of that delay can be fixed. In some cases, a road with heavy traffic is almost impossible to improve, while in others, a road with similar traffic levels could see its delays cut by more than half if managed correctly.

The researchers built their analysis on a well-established computer model of traffic flow known as METANET. This model breaks a highway down into small segments and tracks how the density of cars and their speed change over time. It accounts for how cars enter and exit the road, how they slow down when they get too close to one another, and how a bottleneck, such as a lane reduction, can cause a backup. The team used this model to run a massive optimization exercise. They asked the computer to find the perfect sequence of speed limits for every segment of the highway over a specific period. The goal was to minimize the total time all vehicles spent on the road. The computer was allowed to change the speed limits as often as it wanted, within the physical limits of the road, to see what the absolute best outcome would be. The difference between the actual delay observed in the data and this theoretical best-case scenario is what the researchers define as "controllable congestion."

When the team applied this method to a synthetic highway with a bottleneck, they discovered a surprising pattern. As they increased the number of cars entering the road, the total delay grew steadily, which is expected. However, the amount of that delay that could be removed by controlling speed limits did not grow steadily. Instead, it rose quickly at first, reached a peak where about 56 percent of the delay could be eliminated, and then dropped sharply as the road became even more crowded. This means that once traffic reaches a certain extreme level of density, slowing cars down with speed limits becomes ineffective. The system becomes so saturated that no amount of speed management can prevent the jam. This finding challenges the assumption that more traffic always means more opportunity for improvement. It suggests that there is a tipping point where the road is simply too full to be managed by speed limits alone, and that adding physical capacity, like a new lane, might be the only solution.

The researchers then tested their method on real-world data from the I-24 SMART Corridor in Tennessee, a 5.6-kilometer stretch of highway equipped with sensors and variable speed limit signs. They analyzed ten consecutive weekdays, including two holidays, to see how the potential for improvement changed from day to day. The results were striking. On two different mornings, the total travel time and the average delay per car were nearly identical. By traditional measures, these two days looked exactly the same. Yet, when the researchers calculated the controllable congestion, the difference was enormous. On one day, the system was highly responsive, and the optimal speed limit strategy could have removed 80 percent of the delay. On the other day, with the same amount of traffic, the system was much less responsive, and the best possible strategy could only remove 31 percent of the delay. The reason for this difference lay in the nature of the traffic jam itself. On the day with high potential, the jam was temporary and could be cleared by slowing cars down just enough to prevent a backup from forming. On the day with low potential, the jam was persistent and structural, meaning that even perfect management could not clear it quickly.

This disconnect between total delay and controllable congestion is the most important finding of the study. It shows that looking at delay alone is like looking at a patient's fever without knowing the cause. A high fever could be a minor infection that will pass quickly, or it could be a sign of a serious illness that requires drastic treatment. Similarly, a high delay on a highway could be a sign that the road is simply too busy for any management strategy to help, or it could be a sign that the road is being managed poorly and could be vastly improved with the right tools. The study proves that two roads with the same level of congestion can have completely different potentials for improvement. This means that transportation agencies cannot rely on delay statistics alone to decide where to invest in new technology. They need to know not just how bad the traffic is, but how much of that bad traffic is actually fixable.

The researchers also explored how practical limitations affect this potential. In the real world, speed limits cannot be changed instantly or set to any value. Drivers need time to react, and there are legal minimums for how slow a speed limit can be posted. The team tested how these constraints would change the results. They found that the frequency of updates was critical. If speed limits could only be changed once every two minutes, the amount of delay that could be removed dropped significantly compared to a system that could update every few seconds. However, they found that the minimum speed limit mattered very little. Even if the speed limit could not be set below 90 kilometers per hour, the system could still remove nearly half of the delay. This suggests that the ability to make small, frequent adjustments is far more important than the ability to force traffic to a near standstill. It also highlights that the spacing of the signs matters; having signs close together allows for better control than having them far apart.

The study concludes that "controllable congestion" is a vital new tool for decision-making. It allows planners to distinguish between congestion that is structurally unavoidable and congestion that is highly responsive to management. For a highway where the potential for improvement is low, spending millions of dollars on variable speed limit signs would be a waste of resources. In those cases, the solution might be to expand the road or to accept the congestion as a limit of the infrastructure. For a highway where the potential is high, the same investment could yield massive benefits, cutting travel times and reducing emissions significantly. By quantifying this potential, agencies can stop guessing and start investing in the places where their money will actually work. The framework is not limited to just measuring delay; the same logic could be applied to measure how much pollution or how many accidents could be prevented by better management. While the current study relies on computer simulations and specific data from one corridor, the core insight is clear: knowing how much a system can be improved is just as important as knowing how broken it currently is.