A Survey on Motion Prediction of Pedestrians and Vehicles for Autonomous Driving

Abstract

Autonomous vehicle (AV) industry has evolved rapidly during the past decade. Research and development in each sub-module (perception, state estimation, motion planning etc.) of AVs has seen a boost, both on the hardware (variety of new sensors) and the software sides (state-of-the-art algorithms). With recent advancements in achieving real-time performance using onboard computational hardware on an ego vehicle, one of the major challenges that AV industry faces today is modelling behaviour and predicting future intentions of road users. To make a self-driving car reason and execute the safest motion plan, it should be able to understand its interactions with other road users. Modelling such behaviour is not trivial and involves various factors e.g. demographics, number of traffic participants, environmental conditions, traffic rules, contextual cues etc. This comprehensive review summarizes the related literature. Specifically, we identify and classify motion prediction literature for two road user classes i.e. pedestrians and vehicles. The taxonomy proposed in this review gives a unified generic overview of the pedestrian and vehicle motion prediction literature and is built on three dimensions i.e. motion modelling approach, model output type, and situational awareness from the perspective of an AV.INDEX TERMS Autonomous driving, road vehicles, roads, trajectory prediction, vehicle safety, human intention and behaviour analysis.

DMMGAN: Diverse Multi Motion Prediction of 3D Human Joints using Attention-Based Generative Adverserial Network

“…The MADE uses multi-modal GT future motions by grouping similar past motions. 2 There are 4 other actors without ground truth data…”

“…An important ability of an intelligent system interacting with humans is to estimate plausible human body pose and trajectories in 3D space. With the advancement of artificial intelligence, there are multiple industrial applications for such algorithms in human-robot interactions (HRI) [1], autonomous driving [2] or visual surveillance [3]. Specifically, detailed human 3D body motion prediction plays a crucial role in many robotic applications, such as robot following ahead of a human [4,5] or crowd navigation [6].…”

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

“…It partially overlaps our work, mainly in the presentation of the sensorial part, but it lacks the presentation of the DL-based algorithms associated with each particular type of sensor that we provided in Section 2. Some surveys consider the general framework of motion prediction for pedestrians and vehicles in the context the autonomous driving [127]. Although the authors have provided a brief overview of learning-based models and further propose a taxonomy categorization of DL-oriented methods, they do not refer to datasets, metrics, and experimental results as we have provided in Table 2 and Figure 12.…”

Foresee the Unseen: Sequential Reasoning about Hidden Obstacles for Safe Driving

“…II-A, a prediction is needed of where those obstacles could reach in the future. This prediction can be computed in numerous ways, as shown in [16], [17]. Input: Current field of view, FoV k , previous set, P L k−1 .…”