Application of Machine Learning Techniques for Predicting Potential Vehicle-to-Pedestrian Collisions in Virtual Reality Scenarios

Abstract

The definition of pedestrian behavior when crossing the street and facing potential collision situations is crucial for the design of new Autonomous Emergency Braking systems (AEB) in commercial vehicles. To this end, this article proposes the generation of classification models through the deployment of machine learning techniques that can predict whether there will be a collision depending on the type of reaction, the lane where it occurs, the visual acuity the level of attention, and consider the most relevant factors that determine the cognitive and movement characteristics of pedestrians. Thereby, the inclusion of this type of model in the decision-making algorithm of the AEB system allows for modulating its response. For this purpose, relevant information on pedestrian behavior is obtained through experiments made in an ad-hoc, Virtual Reality (VR) environment, using a portable backpack system in three urban scenarios with different characteristics. Database generation, feature selection, and k-fold cross-validation generate the inputs to the supervised learning models. A subsequent analysis of the accuracy, optimization, error measurement, variable importance, and classification capability is conducted. The tree-based models provide more balanced results for the performance metrics (with higher accuracy for the single decision tree case) and are more easily interpretable and adaptable to the algorithm. From them it is deduced the high importance of the reaction type and the relative position where it occurs, coinciding with the high significance of these factors in the analyzed collisions.

Effectiveness of the Autonomous Braking and Evasive Steering System OPREVU-AES in Real World Vehicle-to-pedestrian collisions

“…This angle is defined as "Minimum Angle of Attention, MAA", and takes into account the maximum distance the pedestrian must travel to a point of the crossing where the collision is possible and the minimum distance the piloted vehicle would need to brake completely from cruising speed. The MAA calculations and the geometric definition in the zenithal planes of the streets are shown in detail in the previous paper [27]. For Machupichu, the MAA is αmin=36.9° and for Hermanos García Noblejas, αmin=43.3°.…”

“…During the data processing phase [27], and after applying an approximation of the minimum redundancy maximum relevance (mRMR) technique for feature selection, four relevant variables were identified in pedestrian behavior and related to the output variable (Avoidance: "0", Collision: "1"): Reaction type (accelerate, stop and step backwards, no reaction), Reaction zone (before hit lane, within hit lane, no speed change), the Percentage of Attention Time (PAT), and the average error (%) made in the distance estimation test in the experimental session in VR. Due to the difficulty of computing this last variable for a system that requires a fast data processing speed, and that would demand access to individual pedestrian information through V2P (Vehicle-to-Pedestrian) technology, the mean error in distance calculation is discarded, and the following variable is chosen as it guarantees the lowest correlation with the rest of the explanatory variables and the highest correlation with the response variables: Street type (reduced visibility, visibility).…”

Understanding Pedestrian Cognition Workload in Traffic Environments Using Virtual Reality and Electroencephalography

“…Virtual reality (VR) presents a compelling alternative to address these challenges in traffic safety research. Authors have already successfully developed and tested a VR simulator to study VRU reactions when facing potential collision situations [7,8]. VR enables the creation of realistic pedestrian simulators, allowing for the meticulous design of environments reflecting real-world conditions.…”

Effectiveness of the Autonomous Braking and Evasive Steering System OPREVU-AES in Simulated Vehicle-to-Pedestrian Collisions

“…The PAT is calculated as the percentage of time used by the pedestrian to look with a head rotation angle that is less than the MAA. The MAA and PAT calculations are shown in detail in the paper in [18].…”

“…The OPREVU project aims to optimize systems for identifying vulnerable users by characterizing their behavior in potential pedestrian collision situations in urban environments using VR techniques. One of the main outputs was the development of machine learning techniques to predict collisions based on pedestrians' kinematics, attention level, and visual perception [18]. The incorporation of an adaptation of this predictive model into a commercial AEB system is proposed in this paper, in addition to the integration of the AES system.…”