Multimodal Interaction-Aware Trajectory Prediction in Crowded Space

Authors: Xiaodan Shi, Xiaowei Shao, Zipei Fan, Renhe Jiang, Haoran Zhang, Zhiling Guo, Guangming Wu, Wei Yuan, Ryosuke Shibasaki11982-11989

AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments over several trajectory prediction benchmarks demonstrate that our method is able to forecast various plausible futures in complex scenarios and achieves state-of-the-art performance.
Researcher Affiliation Academia 1Center for Spatial Information Science, the University of Tokyo 2Earth Observation Data Integration and Fusion Research Initiative, the University of Tokyo 3Information Technology Center, the University of Tokyo
Pseudocode No The paper describes the model architecture and equations but does not include any clearly labeled pseudocode or algorithm blocks.
Open Source Code No The paper does not contain any explicit statement about releasing source code or a link to a code repository.
Open Datasets Yes In this section, the proposed model is evaluated on two publicly available datasets: UCY(Lerner, Chrysanthou, and Lischinski 2007) and ETH(Pellegrini et al. 2009).
Dataset Splits No The paper mentions training and testing splits (leave-one-out approach) but does not explicitly describe a separate validation dataset split with details like percentages or sample counts for hyperparameter tuning or early stopping criteria.
Hardware Specification Yes The experiments are implemented using Pytorch under Ubuntu 16.04 LTS with a GTX 1080 GPU.
Software Dependencies No The paper mentions "Pytorch" and "Ubuntu 16.04 LTS" but does not provide version numbers for all key software components (e.g., PyTorch version is missing), which would be necessary for full reproducibility.
Experiment Setup Yes The size of hidden states of all LSTMs is set to 128. All the embedding layers are composed of a fully connected layer with size 128 and Re LU nonlinearty activation function. The batch size is set to 8 and all the methods are trained for 200 epochs. The optimizer RMSprop is used to train the proposed model with learning rate 0.001. We clip the gradients of LSTM with a maximum threshold of 10 to stabilize the training process. The model outputs GMMs with five components.