Under the hood of transformer networks for trajectory forecasting

Franco, Luca; Leonardo, Placidi,; Giuliari, Francesco; Hasan, Irtiza; Cristani, Marco; Galasso, Fabio

doi:10.1016/j.patcog.2023.109372

Cited by 13 publications

(7 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This approach yielded highly impressive outcomes by independently modeling the pedestrian's historical trajectory without taking into account human-human or environmental interactions. The authors also provided a comprehensive analysis of the distinctions between LSTM and Transformer in their publication (Franco et al 2023). Then, (Yu et al 2020, Su et al 2021, Chen et al 2023, Achaji et al 2022 paired with a CVAE to explain the multi-modality of trajectories and simulate the spatiotemporal interactions between pedestrians, providing an advantage over the Transformer, which employs just historical pedestrian trajectories.…”

Section: Trajectory Prediction Based On Transformermentioning

confidence: 99%

VOSTN: Variational One-shot Transformer Network for Pedestrian Trajectory Prediction

Wang,

Sang,

Chen

et al. 2024

Phys. Scr.

View full text Add to dashboard Cite

The accurate and reliable prediction of pedestrian future trajectories is of crucial significance for ensuring the safe navigation of autonomous driving systems. This paper introduces a novel approach called the Variational One-Shot Transformer Network (VOSTN) for the prediction of future trajectories within the 2D on-board domain. VOSTN presents an innovative method called the One-Shot Generation Block, which aims to generate queries simultaneously in order to predict future trajectories in a parallel manner. The utilization of the parallel prediction effectively addresses the issue of error accumulation resulting from autoregression, improves the efficiency of inference time, and enhances the precision of long-term trajectory prediction. And the cross-attention module investigates the inter-relationship between trajectory and ego-motion. Given the inherent stochasticity of pedestrian movement, we employ the Conditional Variational AutoEncoder to forecast available multimodal trajectories. Experimental results demonstrate that our model effectively exploits the information associated with trajectory and ego-motion, leading to the acquisition of more comprehensive feature representations. Moreover, our model outperforms the performance of existing methods on the two 2D on-board domain datasets. Our deterministic/multimodal prediction models show a reduction in the bounding box center final displacement error by 8% / 9% and 0.7% / 3% on PIE and JAAD, respectively, when compared to the most optimal baseline.

show abstract

Section: Trajectory Prediction Based On Transformermentioning

confidence: 99%

VOSTN: Variational One-shot Transformer Network for Pedestrian Trajectory Prediction

Wang,

Sang,

Chen

et al. 2024

Phys. Scr.

View full text Add to dashboard Cite

show abstract

“…Now, with the success of the transformer model [ 37 ] in natural language processing, the sequence feature extraction tends to shift gradually from the RNN model to the transformer model, which breaks through the limitation that the RNN model cannot be calculated in parallel, and the special attention mechanism can take into account both global and local information. The latest research in crowd feature extraction based on the transformer model [ 38 ] has achieved excellent results. The trajectory prediction task shows that extracting crowd trajectory feature should consider not only the trajectories of a single individual but also the influence of other individuals.…”

Section: Related Workmentioning

confidence: 99%

“…4) ADE and FDE are commonly used in trajectories prediction tasks. [35,36,[38][39][40] Although this article focuses on the generation tasks, which are essentially different from prediction tasks, the methods used in training are similar to prediction tasks, so we use these two indicators to evaluate the accuracy of our model. CD is typically used to assess the spatial similarity between point clouds, which is similar to the spatial similarity of crowd positions.…”

Section: Evaluation Metricsmentioning

confidence: 99%

Stylized Crowd Formation Transformation Through Spatiotemporal Adversarial Learning

Yan,

Huang,

Zhang

et al. 2023

Advanced Intelligent Systems

View full text Add to dashboard Cite

Achieving crowd formation transformations has wide‐ranging applications in fields such as unmanned aerial vehicle formation control, crowd simulation, and large‐scale performances. However, planning trajectories for hundreds of agents is a challenging and tedious task. When modifying crowd formation change schemes, adjustments are typically required based on the style of formation change. Existing methods often involve manual adjustments at each crucial step, leading to a substantial amount of physical labor. Motivated by these challenges, this study introduces a novel generative adversarial network (GAN) for generating crowd formation transformations. The proposed GAN learns specific styles from a series of crowd formation transformation trajectories and can transform a new crowd with an arbitrary number of individuals into the same styles with minimal manual intervention. The model incorporates a space–time transformer module to aggregate spatiotemporal information for learning distinct styles of formation transformation. Furthermore, this article investigates the relationship between the distribution of training data and the length of trajectory sequences, providing insights into the preprocessing of training data.

show abstract

“…BERT utilizes historical power grid load data from a specific time period as input. This data is fed into a fully connected layer Franco et al (2023) to generate forecast outputs. The model is trained using the mean square error as the loss function.…”

Section: Trainingmentioning

confidence: 99%

Intelligent grid load forecasting based on BERT network model in low-carbon economy

Tao

et al. 2023

Front. Energy Res.

View full text Add to dashboard Cite

In recent years, the reduction of high carbon emissions has become a paramount objective for industries worldwide. In response, enterprises and industries are actively pursuing low-carbon transformations. Within this context, power systems have a pivotal role, as they are the primary drivers of national development. Efficient energy scheduling and utilization have therefore become critical concerns. The convergence of smart grid technology and artificial intelligence has propelled transformer load forecasting to the forefront of enterprise power demand management. Traditional forecasting methods relying on regression analysis and support vector machines are ill-equipped to handle the growing complexity and diversity of load forecasting requirements. This paper presents a BERT-based power load forecasting method that leverages natural language processing and image processing techniques to enhance the accuracy and efficiency of transformer load forecasting in smart grids. The proposed approach involves using BERT for data preprocessing, analysis, and feature extraction on long-term historical load data from power grid transformers. Multiple rounds of training and fine-tuning are then conducted on the BERT architecture using the preprocessed training datasets. Finally, the trained BERT model is used to predict the transformer load, and the predicted results are compared with those obtained based on long short-term memory (LSTM) and actual composite values. The experimental results show that compared with LSTM method, the BERT-based model has higher short-term power load prediction accuracy and feature extraction capability. Moreover, the proposed scheme enables high levels of accuracy, thereby providing valuable support for resource management in power dispatching departments and offering theoretical guidance for carbon reduction initiatives.

show abstract

Under the hood of transformer networks for trajectory forecasting

Cited by 13 publications

References 11 publications

VOSTN: Variational One-shot Transformer Network for Pedestrian Trajectory Prediction

VOSTN: Variational One-shot Transformer Network for Pedestrian Trajectory Prediction

Stylized Crowd Formation Transformation Through Spatiotemporal Adversarial Learning

Intelligent grid load forecasting based on BERT network model in low-carbon economy

Contact Info

Product

Resources

About