DD-MDN: Human Trajectory Forecasting with Diffusion-Based Dual Mixture Density Networks and Uncertainty Self-Calibration

2026-02-11

AI Summary

This work addresses the challenges of inadequate uncertainty modeling, lack of calibration, and performance degradation under short observation horizons in human trajectory prediction. To this end, we propose DD-MDN, an end-to-end probabilistic forecasting model that uniquely integrates a denoising diffusion mechanism with dual mixture density networks. DD-MDN eliminates the need for predefined anchors by leveraging a few-shot diffusion backbone and an anchor-free generation strategy, enabling self-calibrated modeling of occupancy regions and probabilistically ranked trajectory predictions. Evaluated on the ETH/UCY, SDD, inD, and IMPTC datasets, our method demonstrates substantial improvements in prediction accuracy, robustness, and uncertainty reliability—particularly in scenarios with limited observation lengths.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Human Trajectory Forecasting (HTF) predicts future human movements from past trajectories and environmental context, with applications in Autonomous Driving, Smart Surveillance, and Human-Robot Interaction. While prior work has focused on accuracy, social interaction modeling, and diversity, little attention has been paid to uncertainty modeling, calibration, and forecasts from short observation periods, which are crucial for downstream tasks such as path planning and collision avoidance. We propose DD-MDN, an end-to-end probabilistic HTF model that combines high positional accuracy, calibrated uncertainty, and robustness to short observations. Using a few-shot denoising diffusion backbone and a dual mixture density network, our method learns self-calibrated residence areas and probability-ranked anchor paths, from which diverse trajectory hypotheses are derived, without predefined anchors or endpoints. Experiments on the ETH/UCY, SDD, inD, and IMPTC datasets demonstrate state-of-the-art accuracy, robustness at short observation intervals, and reliable uncertainty modeling. The code is available at: https://github.com/kav-institute/ddmdn.

Problem

• Human Trajectory Forecasting
• Uncertainty Calibration
• Short Observation Periods
• Probabilistic Prediction
• Autonomous Driving

Innovation

• Diffusion-based forecasting
• Mixture Density Networks
• Uncertainty self-calibration
• Short-observation robustness
• Anchor-free trajectory prediction