Multimodal Hybrid Pedestrian: A Hybrid Automaton Model of Urban Pedestrian Behavior for Automated Driving Applications

Jayaraman, Suresh Kumaar; Robert, Lionel; Yang, X. Jessie; Tilbury, Dawn M. · 2021 · OpenAlex

DOI: 10.1109/access.2021.3058307

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the challenge of predicting pedestrian trajectories for automated vehicles (AVs) in urban environments, where pedestrian behavior is inherently multimodal and stochastic. Existing methods often fail to capture multiple possible future paths, assume all pedestrians intend to cross, or are limited to short-term predictions. To enable safe AV navigation and motion planning, the authors developed the Multimodal Hybrid Pedestrian (MHP) model, a probabilistic hybrid automaton framework designed to predict long-term, multimodal pedestrian behaviors. The MHP model structures pedestrian behavior into four discrete states: approaching a crosswalk, waiting, crossing, and walking away. Transitions between these states are governed by continuous dynamics modeled as constant velocity with Gaussian noise, and probabilistic decision-making at specific points. The model incorporates two support vector machine (SVM) classifiers: one to predict the probability of a pedestrian’s intent to cross, and another to predict gap acceptance decisions based on vehicle proximity and timing. This allows the system to branch into multiple trajectory tracklets at decision points, capturing the uncertainty of whether a pedestrian will cross, wait, or continue walking. The model assumes pedestrians interact primarily with the closest approaching vehicle and operates within a defined interaction zone relative to the ego-vehicle. The study validates the MHP model against two baselines—a baseline hybrid automaton and a constant velocity model—using both real-world and virtual datasets across various urban scenarios, including midblocks and intersections. The results demonstrate that the MHP model more accurately predicts ground truth trajectories than the baselines. By explicitly modeling crossing intent, the MHP avoids the conservative predictions associated with assuming all pedestrians intend to cross, thereby reducing the risk of the "freezing robot" problem where AVs hesitate to move. Furthermore, the model’s ability to generate multiple probable futures provides richer data for AV motion planning compared to single-trajectory predictions. The significance of this work lies in its contribution to human-centered robotics and AV safety. The MHP model offers a computationally efficient, interpretable alternative to data-intensive deep learning approaches, making it suitable for real-time operation. By capturing the multimodal nature of pedestrian decisions and distinguishing between crossing and non-crossing intents, the model enables AVs to navigate structured urban environments more safely and comfortably. The findings suggest that incorporating probabilistic decision-making into hybrid systems improves prediction accuracy and supports robust motion planning in dynamic, high-risk pedestrian-vehicle interactions.

Key finding

The Multimodal Hybrid Pedestrian model more accurately predicts ground truth pedestrian trajectories than baseline models by incorporating probabilistic crossing intent and gap acceptance into a hybrid automaton framework.

Methodology

modeling

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed. Discovered via openalex_abstract on 2026-05-08 (6 acquisition events logged).

StageOutcomeToolModelPromptAttemptsCompleted
discover success 1 2026-05-05
archive success unpaywall 4 2026-06-02
extract success cached 2 2026-06-10
clean success 1 2026-06-01
chunk success 1 2026-06-01
embed success 1 2026-06-02
enrich success openalex 2 2026-05-08
promote success 3 2026-06-06
summarize success llm qwen3.6-27b-prismaquant summ-v5 3 2026-06-10
tag success vector_similarity 18 2026-06-11
verify success 2 2026-06-10

Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).