Handling Occlusions in Automated Driving Using a Multiaccess Edge Computing Server-Based Environment Model From Infrastructure Sensors

Buchholz, Michael; Müller, Johannes; Herrmann, Martin; Strohbeck, Jan; Völz, Benjamin; Maier, Matthias; Paczia, Jonas; Stein, Oliver; Rehborn, Hubert; Henn, Rüdiger-Walter · 2021 · OpenAlex-citations

DOI: 10.1109/mits.2021.3089743

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 sensor occlusions in automated driving, particularly in urban environments where buildings or other traffic participants block the field of view (FOV) of on-board sensors. Such occlusions force connected automated vehicles (CAVs) to adopt conservative driving behaviors, such as unnecessary braking and re-acceleration, which reduces energy efficiency and passenger comfort. To mitigate this, the authors propose a system that extends the CAV’s FOV by leveraging infrastructure sensors and Multi-Access Edge Computing (MEC). The motivation is to provide CAVs with predictive information about occluded areas via Vehicle-to-Anything (V2X) communication, enabling smoother and more efficient motion planning. The study implements a proof-of-concept system at a T-junction in Ulm, Germany, where a building occludes the view of the priority road for vehicles merging from a side road. Infrastructure sensors, including video cameras and lidars mounted on lampposts, detect objects and send processed data to an MEC server. The server fuses these detections using a centralized Labeled Multi-Bernoulli (LMB) Multi-Object Tracker to create a comprehensive environment model, which includes object states and short-term trajectory predictions. This model is transmitted to the CAVs. The authors evaluate two methods for integrating this external data into the CAV’s motion planning: a track-to-track (T2T) fusion approach that combines external and on-board data for hierarchical planning, and a fusion-free approach that uses external data for predictive synchronization while relying on on-board sensors for reactive control. Both methods include reliability estimation schemes to ensure safety. The results are derived from real-world field trials with prototype CAVs in mixed traffic. The study demonstrates that the MEC-based infrastructure support allows CAVs to perceive traffic in occluded areas, thereby avoiding the stop-and-go behavior typical of baseline scenarios without external information. The authors evaluate the impact of these approaches on maneuver time and energy consumption. By utilizing the extended FOV and predictive models, the CAVs were able to plan trajectories more smoothly, merging into traffic gaps without unnecessary deceleration. The findings confirm that the system reduces energy consumption and improves traffic efficiency compared to driving without infrastructure support. The significance of this work lies in its demonstration of a feasible, low-latency architecture for cooperative automated driving that offloads computationally intensive fusion and prediction tasks to edge servers. This approach reduces the computational burden on vehicles while enhancing safety and efficiency in complex urban scenarios. The paper highlights the potential of MEC-based environment models to enable energy-optimizing control strategies, offering a scalable solution for handling occlusions in automated driving systems.

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.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

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

• situational awareness