Enhancing Road Safety and Efficiency with Speed Recommendations
25th March, 2025
Nuria Toribio
Data Scientist, Aimsun
Managing modern traffic efficiently requires advanced tools to ensure safety and minimize congestion. High vehicle density increases the risk of sudden braking, which can cause chain reactions leading to even more congestion and hazardous situations. These risks become more severe on high-speed roads (100-120 km/h) as drivers have less time to avoid collisions.
Speed recommendations are a proactive way to address these challenges. By detecting when speed levels drop below a critical threshold and identifying the origin of the issue, the system can offer progressive speed recommendations to upstream sections. This gradual adjustment helps vehicles approach congestion smoothly, reducing sudden stops and improving overall road safety.
The primary goal of this approach is to reduce the risk of incidents by preventing congestion buildup, ensuring smoother traffic flow, and giving drivers enough time to react and avoid rear-end collisions. This is particularly important during incidents or accidents that create significant bottlenecks, as these situations increase the likelihood of sudden stops and reduced reaction time, heightening the risk of collisions. Furthermore, another key objective of this strategy is to lower vehicle emissions by reducing speed limits, which minimizes inefficient acceleration and deceleration cycles, addressing both traffic efficiency and environmental concerns.
The algorithm processes real-time speed data to identify congestion and recommend appropriate speed adjustments. It detects congestion based on speed data, marking sections as congested when speeds drop below a threshold (e.g., 80 km/h on a 120 km/h road) and using recovery criteria to refine these identifications.
This recovery threshold is applied to determine the congestion status of downstream sections regarding its upstream. If a recovery condition is met, the downstream section is marked as non-congested as it is downstream of the bottleneck, even if the speed data is lower than the threshold. For example, consider the sequence A → B → C, with all the sections labeled congested since speed remains below 80 km/h. B is the bottleneck, and C has a speed higher than B (indicating vehicles are accelerating after congestion).
We apply this filter to avoid erroneously marking section C as congested since it’s the downstream of the bottleneck and we do not recommend speed for that section, as we want traffic behavior to return to free flow.
After identifying the congested sections, we recommend a speed for these by adding a recovery speed to the measured speed. The system then propagates speed recommendations upstream, ensuring smooth traffic transitions. We calculate the recommended speed for an upstream section using its length and the recommended speed in its downstream sections:
rec_speedupstream = rec_speeddownstream + lengthupstream * relative_speed_increment
The process adjusts progressively, and when the recommended speed exceed the usual speed of the road, the propagation of speed recommendations to further upstream sections is terminated.
Speed recommendations are propagated only to sections with a functional class greater than or equal to the current section. Such ensures that congestion on a primary road may affect secondary roads while not the other way around, meaning that an accident on a highway may impact an entrance ramp, but congestion on the ramp would not influence the recommended speed on the main lane. As additional considerations, we eliminate self-loops to avoid erroneous propagation and impute missing values to ensure the algorithm operates on complete data.
Moreover, the speed adjustment algorithm incorporates receiving external incidents, such as accidents. When an incident is detected, the system receives validated data on its location and severity, including the number of blocked lanes. Affected sections are assigned a fixed recommended speed, with recommendations propagating upstream using the previous congestion-based approach. For example, if we receive information that the severity of an accident is high with three out of four lanes blocked, a lower speed, such as 40 km/h, will be recommended for the affected section. However, if a vehicle is partially occupying the right lane, the recommended speed could be 60 km/h. Therefore, the speed propagated to upstream sections will vary from one case to the other. Adapting to particularities allows traffic to be regulated effectively, preventing further disruptions and reducing secondary collisions.
The algorithm identifies the source section of congestion through propagation, allowing the detection of external incidents that trigger speed recommendations across multiple sections and pinpointing the specific incident causing each recommendation.
Pollution-based speed recommendations
As a result, the protocol is activated for the affected segments to ensure compliance with environmental standards based on air quality zones.
A real-world case study: the C-32 highway in Barcelona
A real-world case on the C-32 highway between Castelldefels and Viladecans demonstrates the effectiveness of this system by applying progressive speed recommendations. Speed limits range from 100 km/h in the western part to 60 km/h near Viladecans, with gradual transitions through 90, 80, and 70 km/h. The stepwise reduction pattern allows drivers to adjust their speed smoothly, minimizing sudden braking and reducing the risk of accidents when reaching the bottleneck area near Viladecans, where congestion builds up, making controlled speed adjustments particularly crucial. By benefiting from real-time data, these recommendations help ensure a safer and more efficient traffic flow along this corridor.
The post Enhancing Road Safety and Efficiency with Speed Recommendations appeared first on Aimsun.