Pedestrian-Collision Avoidance Strategy Using Deep Neural Network

As self-driving vehicles continue to gain traction worldwide, the demand for robust safety systems, particularly concerning pedestrian's safety, has become increasingly critical. This paper introduces a pedestrian collision avoidance strategy that focuses on detecting pedestrians and estimating their distance from the vehicle. The key contributions of this approach include: (1) the detection of multiple pedestrians using an onboard vehicle camera, achieved through the training of a neural network; (2) the estimation of pedestrian distance by integrating Lidar point cloud data onto the camera's 2D imagery; and (3) the implementation of a responsive control system that overrides the vehicle's default controller to stop the vehicle when pedestrians are detected within a dangerous proximity.