PyTorch-based implementation of label-aware graph representation for multi-class trajectory prediction

Abstract

Trajectory Prediction under diverse patterns has attracted increasing attention in multiple real-world applications ranging from urban traffic analysis to human motion understanding, among which graph convolution network (GCN) is frequently adopted with its superior ability in modeling the complex trajectory interactions among multiple humans. In this work, we propose a python package by enhancing GCN with class label information of the trajectory, such that we can explicitly model not only human trajectories but also that of other road users such as vehicles. This is done by integrating a label-embedded graph with the existing graph structure in the standard graph convolution layer. The flexibility and the portability of the package also allow researchers to employ it under more general multi-class sequential prediction tasks.

Pedestrian Trajectory Prediction in Pedestrian-Vehicle Mixed Environments: A Systematic Review

“…A variant of GNNs called Graph Convolutional Networks (GCN) has also been used for trajectory prediction of pedestrians in a mixed environment of multi-class agents [51]- [53]. In these models, the feature of each node and the structure of the graph's connections in form of an adjacency matrix are inputted to a multi-layer convolutional network for extraction of common patterns in the trajectories.…”

“…Some have used Graph Convolutional Network (GCN) to directly operate on the graph [45], [51]- [53]. In these models, an adjacency matrix is constructed which stores the connected edges in the graph and specifies weights for them that are proportional to the inverse of the relative speed [52] or relative distances [45], [53] of the interacting agents.…”

“…Among the papers that implement the structure of a graph neural network (GNN) for encoding interaction, there are a few papers that account for the difference between interactions of homogeneous agents versus interactions of heterogeneous agents [32], [45], [47]- [49], [51]- [53].…”

KGCN‐LSTM: A graph convolutional network considering knowledge fusion of point of interest for vehicle trajectory prediction

“…Road network interface and road similarity are also considered in the representation learning method by Sun [26], and a comprehensive representation is provided for the input of trajectory prediction. Moreover, the Word2vec and Node2vec methods, combined with the CNN model [27], and a corpus of vehicle preference features [28], as well as graph embedding layers [29] are proposed for traffic problems. Based on the complexity and diversity of trajectory data, noise trajectory [30], different characteristics of vehicle motion [31], and vehicle distribution features [32] are considered in proposing trajectory representation models.…”

Multiclass-SGCN: Sparse Graph-Based Trajectory Prediction with Agent Class Embedding

“…Intuitively speaking, even if two agents have a similar velocity, human instincts would force us to pay more attention to the movements of the larger agents, such as considering car over bicycle. To address this issue, Semantics-STGCNN [10,11] considered class labels for multi-class trajectory prediction by embedding agent-label features into the velocity representations [12], ensuring that the upcoming GCN [13] aggregates both features. Nevertheless, Semantics-STGCNN still suffers from the superfluous interactions problem as it uses a densely connected graph.…”