Pedestrian Models for Autonomous Driving Part I: Low-Level Models, From Sensing to Tracking

Abstract

Autonomous vehicles (AVs) must share space with pedestrians, both in carriageway cases such as cars at pedestrian crossings and off-carriageway cases such as delivery vehicles navigating through crowds on pedestrianized high-streets. Unlike static obstacles, pedestrians are active agents with complex, interactive motions. Planning AV actions in the presence of pedestrians thus requires modelling of their probable future behaviour as well as detecting and tracking them. This narrative review article is Part I of a pair, together surveying the current technology stack involved in this process, organising recent research into a hierarchical taxonomy ranging from low-level image detection to high-level psychology models, from the perspective of an AV designer. This self-contained Part I covers the lower levels of this stack, from sensing, through detection and recognition, up to tracking of pedestrians. Technologies at these levels are found to be mature and available as foundations for use in high-level systems, such as behaviour modelling, prediction and interaction control.

A Review of Deep Learning-Based Methods for Pedestrian Trajectory Prediction

“…The topic is of high interest, as the number of published research papers proves. This also includes several surveys on PTP published in recent years, e.g., [122,123]. For example, an in-depth overview of human motion trajectory prediction is presented in [124].…”

Pedestrian Detection Using Multispectral Images and a Deep Neural Network

“…where c 1 corresponds to the class label in the ground truth, and c 1 is the predicted class label. The loss function of the employed YOLO model is a summation of Equations ( 1), ( 4), ( 7)- (9).…”

Space Invaders: Pedestrian Proxemic Utility Functions and Trust Zones for Autonomous Vehicle Interactions

“…A comprehensive review on pedestrian models for autonomous driving is proposed in [9,10], ranging from low-level sensing, detection and tracking models [9] to high-level interaction and game theoretic models [10]. In the context of autonomous vehicles, more work has been focused on pedestrian crossing behaviour [53], trajectory prediction [84] and for eHMI (external Human-Machine Interface) [20,29,50,54].…”

“…Autonomous vehicles (AVs) are claimed by many organisations to be close to commercial reality, but their lack of human behaviour understanding is raising concerns. While robotic localisation and navigation in static environments [76] and pedestrian detection [9] are well understood, AVs do not yet have the social abilities of human drivers-who can read the intentions of other road users, predict their future behaviour and then interact with them [10]. Pedestrians, unlike other road users such as cyclists, do not usually follow specific traffic rules, in particular when crossing the road at unsigned crossing points, making them especially difficult to model, predict, and interact with.…”