Accurately perceiving and predicting the behavior of vulnerable road users like pedestrians is a significant challenge for autonomous ego-vehicles in urban environments, primarily due to occlusions that hinder accurate multipedestrian tracking. While V2X communications offer a promising solution by enabling the exchange of measurements and contextual information between other vehicles and infrastructure, the effective fusion and processing of this diverse data remain a critical challenge. This research presents a robust framework for multi-pedestrian tracking that fuses sensor data from multiple V2X units using set-membership estimation. The approach employs a vertexbased representation for bidimensional pedestrian positional state estimation, ensuring computational efficiency while maintaining safety. The framework is further enhanced by integrating data association techniques through a threshold-based Hungarian algorithm, enabling reliable measurement fusion from multiple V2X units. Evaluations on simulations and public benchmarks demonstrate that the proposed method outperforms traditional Kalman filterbased approaches and a Set-Membership Estimator, offering a more reliable solution for enhanced ego-vehicle awareness in complex urban scenarios.
Davide Nascivera (Wed,) studied this question.