Prediction Horizon Requirements for Automated Driving: Optimizing Safety, Comfort, and Efficiency

TL;DR

This work investigates how trajectory-prediction horizons influence automated-vehicle performance in urban scenarios with crossing pedestrians. Using a state-of-the-art, risk-based trajectory planner and idealized predictions, it simulates horizons up to $20$ seconds to quantify safety, comfort, and efficiency, and develops a framework to derive minimum required and optimal horizons for different applications. The results show that $1.6$ s suffices to avoid collisions, horizons of $7$–$8$ s optimize efficiency, and horizons up to $15$ s improve passenger comfort, though longer horizons increase computation and can reduce real-time performance. Acknowledging the horizon is application-dependent, the authors propose an overall guideline of $11.8$ s for pedestrian-crossing contexts and outline methods to tailor horizons to specific operational design domains. The work advances horizon design from ad hoc choices to a structured planning-aware specification, with future work on prediction-accuracy requirements and broader scenario generalization.

Abstract

Predicting the movement of other road users is beneficial for improving automated vehicle (AV) performance. However, the relationship between the time horizon associated with these predictions and AV performance remains unclear. Despite the existence of numerous trajectory prediction algorithms, no studies have been conducted on how varying prediction lengths affect AV safety and other vehicle performance metrics, resulting in undefined horizon requirements for prediction methods. Our study addresses this gap by examining the effects of different prediction horizons on AV performance, focusing on safety, comfort, and efficiency. Through multiple experiments using a state-of-the-art, risk-based predictive trajectory planner, we simulated predictions with horizons up to 20 seconds. Based on our simulations, we propose a framework for specifying the minimum required and optimal prediction horizons based on specific AV performance criteria and application needs. Our results indicate that a horizon of 1.6 seconds is required to prevent collisions with crossing pedestrians, horizons of 7-8 seconds yield the best efficiency, and horizons up to 15 seconds improve passenger comfort. We conclude that prediction horizon requirements are application-dependent, and recommend aiming for a prediction horizon of 11.8 seconds as a general guideline for applications involving crossing pedestrians.

Figures (10)

• Figure 1: Typical assessment of trajectory prediction work (left) and our approach (right).
• Figure 2: Methodology to derive prediction horizon requirements.
• Figure 3: Histogram of pedestrian speeds used during this study and corresponding scenario when the vehicle does not react to the pedestrian.
• Figure 4: Example of two situations yielding a different overall optimal horizon. Left: it is possible to choose one of the optimal horizons without negatively affecting any metric. Right: it is not possible to choose an optimal horizon without negatively affecting some metric.
• Figure 5: Example of two situations yielding a different overall required horizon. Left: it is possible to choose a horizon that satisfies the required value of all metrics. Right: it is not possible to choose a horizon that satisfies the required value of all metrics.