RA-GAT: Repulsion and Attraction Graph Attention for Trajectory Prediction

Ding, Zhezhang; Yao, Ziyang; Zhao, Huijing

doi:10.1109/itsc48978.2021.9564907

Cited by 9 publications

(5 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Recursive neural networks have proven to be efficient in solving sequence prediction tasks [19][20][21][22]. Furthermore, some works have improved the performance of trajectory prediction tasks in complex interactive scenarios by incorporating attention mechanisms [23][24][25][26][27][28][29].…”

Section: A Trajectory Predictionmentioning

confidence: 99%

A Novel Lane-Change Decision-Making With Long-Time Trajectory Prediction for Autonomous Vehicle

Wang,

Hu,

Wei

et al. 2023

IEEE Access

View full text Add to dashboard Cite

In the process of autonomous vehicle lane changing, a reliable decision-making system is crucial for driving safety and comfort. However, traditional decision-making systems have short-term characteristics, which makes them susceptible to real-time inference from surrounding vehicles. Usually, system sacrifices driving comfort to ensure the safety of the lane change. Balancing driving safety and comfort has always been a research challenge. Long-term trajectory prediction can provide accurate future trajectories of target vehicles, providing reliable long-term information to compensate for the short-term variability of decision systems. This paper proposes a novel decision-making model with long-term trajectory prediction for lane-changing. First, we constructed a long-term trajectory prediction model to predict the trajectories of surrounding vehicles. Besides, we built a lane change decision-making model based on fuzzy inferencing, considering the predicted trajectories to infer the relative relationship between other vehicles and the self-driving car. The establishment of the fuzzy rule library considered the vehicle speed, acceleration, system delay time, driver delay time and the distance between vehicles. Finally, we created a dataset for training and testing the trajectory prediction model, and we built 4 cases simulation environments, for two or three vehicles on a straight road or curved road, respectively, to test the decisionmaking model. Experimental results show that our proposed model can ensure driving safety and improve driving comfort.

show abstract

Section: A Trajectory Predictionmentioning

confidence: 99%

A Novel Lane-Change Decision-Making With Long-Time Trajectory Prediction for Autonomous Vehicle

Wang,

Hu,

Wei

et al. 2023

IEEE Access

View full text Add to dashboard Cite

show abstract

“…It has a spatial GNN and an additional temporal GNN. In [9], the Repulsion and Attraction Graph Attention (RA-GAT) model for trajectory prediction is presented. The model is based on two stacked Graph Attention Networks [27], that address either free space or vehicle state information through distinct graph definition.…”

Section: Related Workmentioning

confidence: 99%

“…Simultaneously, the model most probably fails to capture the minority class of lane change predictions. This problem was investigated in more depth by Ding et al [9]: As analyzed, about 96.37 % of the scenarios contained in NGSIM dataset are keep lane scenarios. While the performance of their introduced trajectory prediction model on the overall dataset is superior to other baseline models, it performs poorly on the lane change scenarios.…”

Section: Experimental Evaluationmentioning

confidence: 99%

“…Furthermore, the global description of the datasets is transformed Architecture Where needed, the grid definitions of the baseline models were adopted, so that the same traffic scenario is provided to each model. Results marked with * 1 are taken from [19] and * 2 from [9], since their problem definition is equal to the one of this work.…”

Section: Experimental Evaluationmentioning

confidence: 99%

“…The GFTNN is compared with state-of-the-art baselines including conventional recurrent models (CS-LSTM [7], MHA-LSTM(+f) [19]) and novel graph-based approaches (Two-Channel GNN [20], RA-GAT [9]). For evaluation the widely used metrics Average Displacment Error (ADE)…”

Section: B Evaluationmentioning

confidence: 99%

See 2 more Smart Citations

A Multidimensional Graph Fourier Transformation Neural Network for Vehicle Trajectory Prediction

Neumeier

Tollkuhn²,

Botsch

et al. 2022

2022 IEEE 25th International Conference on Intelligent Transportation Systems (ITSC)

View full text Add to dashboard Cite

This work provides a comprehensive derivation of the parameter gradients for GATv2 [4], a widely used implementation of Graph Attention Networks (GATs). GATs have proven to be powerful frameworks for processing graph-structured data and, hence, have been used in a range of applications. However, the achieved performance by these attempts has been found to be inconsistent across different datasets and the reasons for this remains an open research question. As the gradient flow provides valuable insights into the training dynamics of statistically learning models, this work obtains the gradients for the trainable model parameters of GATv2. The gradient derivations supplement the efforts of [2], where potential pitfalls of GATv2 are investigated.

show abstract

Transformer based composite network for autonomous driving trajectory prediction on multi-lane highways

Sharma,

Sahoo,

Puhan

2024

Appl Intell

View full text Add to dashboard Cite

RA-GAT: Repulsion and Attraction Graph Attention for Trajectory Prediction

Cited by 9 publications

References 16 publications

A Novel Lane-Change Decision-Making With Long-Time Trajectory Prediction for Autonomous Vehicle

A Novel Lane-Change Decision-Making With Long-Time Trajectory Prediction for Autonomous Vehicle

A Multidimensional Graph Fourier Transformation Neural Network for Vehicle Trajectory Prediction

Transformer based composite network for autonomous driving trajectory prediction on multi-lane highways

Contact Info

Product

Resources

About