Are All Traffic Conflict Studies Equal? How to Ensure Reliability in Road Safety Analysis
With increasing concerns over the accuracy and timeliness of crash data, traffic conflict studies have emerged as a proactive, data-rich alternative to traditional crash-based safety analysis. Traffic conflicts—or near-miss events—occur far more frequently than crashes, offering a wealth of information to assess risk and improve road safety in real time.
However, as the popularity of conflict-based methods grows, so do the concerns over their reliability, accuracy, and standardization. Not all conflict studies are created equal, and failing to address this could result in flawed safety conclusions and ineffective interventions.
Why Traffic Conflict Studies Matter
Unlike crash data that relies on reactive measures, traffic conflict techniques provide instantaneous insights into safety issues—without waiting for accidents to occur. The combination of video analytics, Lidar, AI, and connected vehicle data now enables real-time identification of conflicts. But the question arises: how reliable are these systems?
📌 Proactive safety insights
📌 Higher frequency of usable data
📌 Avoids ethical issues of crash-based learning
⚠️ Important Note: Just because a system detects “conflicts” doesn’t mean it’s measuring true crash risk unless it follows validated methodologies.
Understanding the Core Issues in Conflict Analysis
Issue 1: How Should We Measure Conflicts?
Most systems rely on just one indicator—such as Time to Collision (TTC) or Post-Encroachment Time (PET)—but this is often insufficient.
True conflict assessment must combine three core elements:
Time and space proximity
Evasive action intensity
Collision consequences
Example: A pedestrian’s evasive action (like a sudden stop or sprint) can be a powerful conflict indicator when combined with proximity data.
Best Practice: Use multiple, complementary indicators rather than relying on a single metric.
Issue 2: Which Conflict Threshold Is Correct?
Thresholds (like TTC < 3 seconds) are often arbitrary, varying widely across studies. This inconsistency causes unreliable outcomes.
✅ Recommended Approach:
Use thresholds that show the strongest correlation with actual crash data
Focus on extreme conflict events, not minor interactions
Avoid liberal thresholds that misclassify normal interactions as risky
Issue 3: Accuracy of Conflict Extraction Systems
The quality of conflict data extraction is critical. Unfortunately, many systems cut corners on essential steps:
❌ Ignoring vehicle size → leads to underestimations
❌ Skipping camera calibration → causes distorted distance and speed measurements
❌ Relying on simple tracking in image space → introduces major geometric errors
✅ What You Need:
Real-world coordinate tracking (not just pixel space)
Camera calibration to remove perspective errors
Probabilistic path prediction models (not naïve extrapolation)
📸 Figure 2: Proper camera calibration significantly improves the detection accuracy of traffic events.
Key Takeaways
- Not all conflict detection systems are equal.
- Use multiple indicators to better represent severity and risk.
- Standardize thresholds based on correlation with crash data.
- Ensure precise conflict extraction through proper calibration and modeling.
- Avoid treating traffic conflicts as mere traffic volume indicators.
As traffic conflict analysis becomes central to proactive road safety strategies, accuracy, reliability, and standardization must guide implementation. Ongoing research and technological refinement will ensure these methods reach their full potential—saving lives in the process.
