Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Integrating occlusion awareness in urban motion prediction for enhanced autonomous vehicle navigation

Vinicius Trentin, Juan Medina-Lee, Antonio Artuñedo, Jorge Villagra

TL;DR

This work addresses safety challenges from occlusions in urban motion prediction for autonomous vehicles by extending the MultIAMP framework with an occlusion-aware layer. The authors integrate hidden vehicles and virtual corridors, occlusion tracking, and relations to map layout within a Dynamic Bayesian Network and Markov chain-based prediction pipeline to produce a probabilistic motion grid $\mathcal{M}$ that informs a state-of-the-art motion planner. Through highway and intersection simulations, occlusion-aware predictions yield planner performance close to omniscient scenarios, while outperforming occlusion-unaware baselines in safety and comfort metrics. The study demonstrates a practical, scalable approach to robust navigation under partial observability, with potential impact for safer urban autonomous driving in the presence of occlusions.

Abstract

Motion prediction is a key factor towards the full deployment of autonomous vehicles. It is fundamental in order to ensure safety while navigating through highly interactive and complex scenarios. Lack of visibility due to an obstructed view or sensor range poses a great safety issue for autonomous vehicles. The inclusion of occlusion in interaction-aware approaches is not very well explored in the literature. In this work, the MultIAMP framework, which produces multimodal probabilistic outputs from the integration of a Dynamic Bayesian Network and Markov chains, is extended to tackle occlusions. The framework is evaluated with a state-of-the-art motion planner in two realistic use cases.

Integrating occlusion awareness in urban motion prediction for enhanced autonomous vehicle navigation

TL;DR

This work addresses safety challenges from occlusions in urban motion prediction for autonomous vehicles by extending the MultIAMP framework with an occlusion-aware layer. The authors integrate hidden vehicles and virtual corridors, occlusion tracking, and relations to map layout within a Dynamic Bayesian Network and Markov chain-based prediction pipeline to produce a probabilistic motion grid that informs a state-of-the-art motion planner. Through highway and intersection simulations, occlusion-aware predictions yield planner performance close to omniscient scenarios, while outperforming occlusion-unaware baselines in safety and comfort metrics. The study demonstrates a practical, scalable approach to robust navigation under partial observability, with potential impact for safer urban autonomous driving in the presence of occlusions.

Abstract

Motion prediction is a key factor towards the full deployment of autonomous vehicles. It is fundamental in order to ensure safety while navigating through highly interactive and complex scenarios. Lack of visibility due to an obstructed view or sensor range poses a great safety issue for autonomous vehicles. The inclusion of occlusion in interaction-aware approaches is not very well explored in the literature. In this work, the MultIAMP framework, which produces multimodal probabilistic outputs from the integration of a Dynamic Bayesian Network and Markov chains, is extended to tackle occlusions. The framework is evaluated with a state-of-the-art motion planner in two realistic use cases.
Paper Structure (13 sections, 10 figures, 1 table)

This paper contains 13 sections, 10 figures, 1 table.

Table of Contents

  1. Introduction
  2. Multimodal Interaction-Aware Motion Prediction
  3. Framework adaptation
  4. Search and classify
  5. Hidden vehicles and virtual corridors
  6. Track and simplify occlusions
  7. Find relations between corridors and layout
  8. Acceleration profile
  9. Generate predictions
  10. Experiments
  11. Highway scenario
  12. Intersection scenario
  13. Concluding remarks

Figures (10)

  • Figure 1: Flowchart of the occlusion layer in MultIAMP.
  • Figure 2: Example of an occluded situation.
  • Figure 3: Evolution of an occlusion at (a) 0 s, (b) 1.0 s, (c) 2.0 s and (d) 3.0 s. Search represents (in red) the occlusion found in the current time step and update shows the occlusion intersected with the prediction of the previous time step (in yellow). The intersection between the two areas is displayed in orange.
  • Figure 4: Example of a (a) bounded occlusion and (b) the evolution of the initial velocity distribution.
  • Figure 5: Highway scenario layout. The area in cyan is the EV's FOV. Vehicles outside the FOV are not perceived by the EV.
  • ...and 5 more figures