Predictability of road traffic and congestion in urban areas

Mitigating traffic congestion on urban roads, with paramount importance in urban development and reduction of energy consumption and air pollution, depends on our ability to foresee road usage and traffic conditions pertaining to the collective behavior of drivers, raising a significant question: to what degree is road traffic predictable in urban areas? Here we rely on the precise records of daily vehicle mobility based on GPS positioning device installed in taxis to uncover the potential daily predictability of urban traffic patterns. Using the mapping from the degree of congestion on roads into a time series of symbols and measuring its entropy, we find a relatively high daily predictability of traffic conditions despite the absence of any a priori knowledge of drivers’ origins and destinations and quite different travel patterns between weekdays and weekends. Moreover, we find a counterintuitive dependence of the predictability on travel speed: the road segment associated with intermediate average travel speed is most difficult to be predicted. We also explore the possibility of recovering the traffic condition of an inaccessible segment from its adjacent segments with respect to limited observability. The highly predictable traffic patterns in spite of the heterogeneity of drivers’ behaviors and the variability of their origins and destinations enables development of accurate predictive models for eventually devising practical strategies to mitigate urban road congestion.

💡 Research Summary

The paper investigates how predictable urban road traffic is by analyzing high‑frequency GPS traces from more than 20,000 taxis operating on Beijing’s 2nd, 3rd, and 4th Ring Roads. The authors convert the continuous speed measurements of each road segment into a discrete symbolic time series: each segment is divided into equal‑length cells (ΔL, typically 1 km) and the average speed in each time interval is quantized into speed levels (ΔV, typically 10 km/h). This mapping yields a sequence of symbols that represent the congestion state of a segment over time.

First, transition probability matrices are constructed for each speed level. The matrices reveal that a segment’s speed state tends to stay the same or move to an adjacent level, while jumps to distant levels are rare, indicating a strong inertia in traffic dynamics.

To quantify the uncertainty of these dynamics, three entropy measures are introduced: (i) Random entropy S_rand = log₂ N_i, where N_i is the number of distinct speed levels visited by segment i; (ii) Uncorrelated temporal entropy S_unc = –∑_j p_i(j) log₂ p_i(j), which accounts for the uneven visitation probabilities but ignores order; (iii) Actual entropy S_i, which incorporates both the probability distribution and the temporal ordering of subsequences in the observed symbol stream. Empirically, S_i is far lower than S_rand, demonstrating that traffic states are far from random and possess considerable regularity.

Using Fano’s inequality, the authors derive an upper bound Π_max on the predictability of a segment’s next state as a function of its entropy S and the number of possible states N. For realistic settings (ΔL = 1 km, ΔV = 20 km/h) the average Π_max on the 2nd Ring Road reaches about 0.83, meaning that, in principle, the next speed level can be correctly guessed 83 % of the time. Predictability based solely on S_unc or S_rand (Π_unc, Π_rand) is substantially lower, confirming that temporal ordering carries most of the predictive power.

A sensitivity analysis shows that overly short segments (ΔL < 0.5 km) or overly fine speed discretization (ΔV < 5 km/h) degrade predictability because of insufficient data per state. Within a practical range (ΔL ≈ 1 km, ΔV ≈ 10–20 km/h) predictability remains high across all three rings.

A striking non‑monotonic relationship emerges between average speed and predictability. Segments with very low speeds (near congestion) or very high speeds (free‑flow) are easier to predict because only two transitions are plausible (stay or move toward the opposite extreme). Segments with intermediate speeds have many possible transitions (accelerate, decelerate, maintain, or stop), leading to higher entropy and lower Π_max. Consequently, the 3rd Ring Road, whose average speed lies in the intermediate range, exhibits the lowest predictability among the three rings.

The authors also compare weekdays and weekends. Despite obvious differences in traffic volume and patterns, the distribution of Π_max is virtually identical for both, indicating that each day type possesses its own stable, highly predictable pattern encoded in the historical data.

Finally, the study explores reconstructing the state of an unobserved segment from its neighboring segments. When adjacent segments have high Π_max, the missing segment’s speed level can be inferred with over 70 % accuracy, suggesting that limited sensor coverage or data loss can be mitigated using spatial correlation.

In summary, the paper demonstrates that urban traffic, when viewed at the level of aggregated road segments, exhibits a surprisingly high degree of predictability that can be quantified using information‑theoretic tools. The findings have practical implications for real‑time traffic management, sensor placement, and the design of control strategies (e.g., variable speed limits, adaptive signal timing) that target the most unpredictable, intermediate‑speed segments. The methodology provides a robust, data‑driven framework for assessing and exploiting the inherent regularities of complex urban traffic systems.