AST-GNN: An attention-based spatio-temporal graph neural network for Interaction-aware pedestrian trajectory prediction

Zhou, Hao; Ren, Dongchun; Xia, Huaxia; Fan, Mingyu; Xu, Yang; Huang, Hai

doi:10.1016/j.neucom.2021.03.024

Cited by 106 publications

(22 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…There are also 46% singletons in the dataset. Metrics. For comparison, we select several prediction methods as baselines including: (1) Social‐LSTM, a modified LSTM model with social pooling layer, (2) Bi‐LSTM [11], a cyclist trajectory prediction method using bidirectional recurrent neural network, (3) Mi‐LSTM [9], an LSTM model using multiple interaction clues, (4) SR‐LSTM [3], an LSTM model utilizing the current states of all participants in the crowd through a message passing mechanism, (5) Social‐GAN [12], a generative adversarial networks based encoder decoder framework for trajectory prediction, (6) STGCNN [5], a unified Graph Convolutional Neural Network for pedestrian trajectory forecasting, (7) AST‐GNN [13], an attention‐based spatio‐temporal graph neural network for pedestrian trajectory prediction, (8) Graph‐TCN [14], a spatio‐temporal interaction modelling for human trajectory prediction.Their performances are evaluated on the non‐vehicle dataset in terms of the Average Displacement Error (ADE) and the Final Displacement Error (FDE) in Tables 1 and 2, respectively [2]. The evaluation is conducted according to a fivefold cross validation, which is widely used in the trajectory prediction.…”

Section: Experiments and Results Analysismentioning

confidence: 99%

Trajectory prediction of cyclist based on spatial‐temporal multi‐graph network in crowded scenarios

Chen

2021

Electronics Letters

View full text Add to dashboard Cite

Cyclist trajectory prediction is an essential task in autonomous driving and surveillance systems. This task is challenging due to that the bicycles go much faster than the pedestrians and a minor prediction error could lead to a severe deviation in the actual path. Existing cyclist trajectory prediction models usually employ the social pooling mechanism to depict the mutual interactions between targets. They ignore that the pooling operation is leaky in information. Moreover, they prefer to use the recurrent architecture to capture the time‐varying features, which is not efficient in computation and parameter learning. To address these issues, a spatial‐temporal multi‐graph module which employs the topology of graphs to represent social interactions and design multi‐kernel functions to depict the social attributes from various aspects is proposed. Instead of the recurrent architecture, a temporal convolution to forecast the future paths is introduced. Experimental results on real‐world datasets demonstrate its superior performance against state‐of‐the‐art baselines. It reduces 9% prediction error when compared to recurrent neural network based models and is more effective in crowded scenarios.

show abstract

Section: Experiments and Results Analysismentioning

confidence: 99%

Trajectory prediction of cyclist based on spatial‐temporal multi‐graph network in crowded scenarios

Chen

2021

Electronics Letters

View full text Add to dashboard Cite

show abstract

“…In addition, many authors have recently been drawn to the advantages of a graph neural network in areas such as traffic flow prediction [14]- [17], parking availability prediction [18], pedestrian trajectories prediction [19], [20], urban vehicle emission prediction [21], wind speed prediction [22], weather prediction [23] and solar irradiance prediction [24]. The air quality domain, among others, has also benefited from these advantages, and various authors have used graph neural networks to forecast air quality.…”

Section: Related Workmentioning

confidence: 99%

Graph Neural Network for Air Quality Prediction: A Case Study in Madrid

2023

View full text Add to dashboard Cite

show abstract

“…Transformers have been introduced to the literature with the promise of tackling the issue of long-term temporal correlation as well as parallelizing the decoding process. Inspired by its distinct attention mechanism, various attention-based techniques have been adopted in [ 1 , 7 ]. However, the multi-headed attention mechanism, originally proposed in the traditional transformer [ 8 ], has not extensively been explored in the highway trajectory prediction problem, mainly due to the problem of accumulative errors resulting from the autoregressive decoding procedure of transformers [ 9 ].…”

Section: Introductionmentioning

confidence: 99%

MALS-Net: A Multi-Head Attention-Based LSTM Sequence-to-Sequence Network for Socio-Temporal Interaction Modelling and Trajectory Prediction

Hasan

Huang

2023

Sensors

View full text Add to dashboard Cite

Predicting the trajectories of surrounding vehicles is an essential task in autonomous driving, especially in a highway setting, where minor deviations in motion can cause serious road accidents. The future trajectory prediction is often not only based on historical trajectories but also on a representation of the interaction between neighbouring vehicles. Current state-of-the-art methods have extensively utilized RNNs, CNNs and GNNs to model this interaction and predict future trajectories, relying on a very popular dataset known as NGSIM, which, however, has been criticized for being noisy and prone to overfitting issues. Moreover, transformers, which gained popularity from their benchmark performance in various NLP tasks, have hardly been explored in this problem, presumably due to the accumulative errors in their autoregressive decoding nature of time-series forecasting. Therefore, we propose MALS-Net, a Multi-Head Attention-based LSTM Sequence-to-Sequence model that makes use of the transformer’s mechanism without suffering from accumulative errors by utilizing an attention-based LSTM encoder-decoder architecture. The proposed model was then evaluated in BLVD, a more practical dataset without the overfitting issue of NGSIM. Compared to other relevant approaches, our model exhibits state-of-the-art performance for both short and long-term prediction.

show abstract

AST-GNN: An attention-based spatio-temporal graph neural network for Interaction-aware pedestrian trajectory prediction

Cited by 106 publications

References 16 publications

Trajectory prediction of cyclist based on spatial‐temporal multi‐graph network in crowded scenarios

Trajectory prediction of cyclist based on spatial‐temporal multi‐graph network in crowded scenarios

Graph Neural Network for Air Quality Prediction: A Case Study in Madrid

MALS-Net: A Multi-Head Attention-Based LSTM Sequence-to-Sequence Network for Socio-Temporal Interaction Modelling and Trajectory Prediction

Contact Info

Product

Resources

About