posterior sampling prediction

Joint Pedestrian Trajectory Prediction through Posterior Sampling

Haotian Lin , Yixiao Wang , Mingxiao Huo, Chensheng Peng, Zhiyuan Liu , Masayoshi Tomizuka

Tsinghua University, UC Berkeley, Carnegie Mellon University

ArXiv

Can guided full trajectory diffuser support more abilities beyond the normal prediction method?

Abstract

Joint pedestrian trajectory prediction has long grappled with the inherent unpredictability of human behaviors. Recent investigations employing variants of conditional diffusion models in trajectory prediction have exhibited notable success. Nevertheless, the heavy dependence on accurate historical data results in their vulnerability to noise disturbances and data incompleteness. To improve the robustness and reliability, we introduce the Guided Full Trajectory Diffuser (GFTD), a novel diffusion model framework that captures the joint full (historical and future) trajectory distribution. By learning from the full trajectory, GFTD can recover the noisy and missing data, hence improving the robustness. In addition, GFTD can adapt to data imperfections without additional training requirements, leveraging posterior sampling for reliable prediction and controllable generation. Our approach not only simplifies the prediction process but also enhances generalizability in scenarios with noise and incomplete inputs. Through rigorous experimental evaluation, GFTD exhibits superior performance in both trajectory prediction and controllable generation.

Framework Introduction

To improve the robustness and reliability, we introduce the Guided Full Trajectory Diffuser (GFTD), a novel diffusion model framework that captures the joint full (historical and future) trajectory distribution. By learning from the full trajectory, GFTD can recover the noisy and missing data, hence improving the robustness.

Experiment Results

Controllable Generation: Our prediction visualization shows a reliable result with the goal point guidence. The goal point guidence shows a meaningful result in our work.

Prediction with clean trajectory data: Although flexibility is the main advantage of our method, we still shows competitive results in some traditional prediction benchmarks, with the state of art methods.

Incomplete Data Prediction: Our framework exhibits competitive performance under incomplete data settings. Even when 75% of history observation is missing, the average JADE/JFDE performance only degenerated for 21.01%/16.62%.

BibTex

@article{lin2024joint,

title={Joint Pedestrian Trajectory Prediction through Posterior Sampling},

author={Lin, Haotian and Wang, Yixiao and Huo, Mingxiao and Peng, Chensheng and Liu, Zhiyuan and Tomizuka, Masayoshi},

journal={arXiv preprint arXiv:2404.00237},

year={2024}

}