Source-linked AI summary
Highly Detailed Simulation for Connected Automated Vehicle Cooperative Driving
Andrei Fizulin, Ilia Dolgov, Andrei Karpukhin
TL;DR
End-to-end CAV simulation needs realistic modeling across mobility, environment, communication, and control, but planar abstractions can miss vertical occlusions. The paper presents a CAVISE workflow combining map-based mobility and 3D-aware ray-tracing comparisons with modular AIM integration, finding markedly optimistic connectivity in the flattened baseline.
Problem
Planar V2X propagation models cannot represent elevation-dependent occlusions in multi-level infrastructure, limiting credible CDA and AIM simulation.
Method
The paper couples map-based 3D scene construction, microscopic mobility, identical-configuration Sionna RT comparisons of 3D and flattened 2D geometry, and modular AIM integration.
Results
42% outage in 3D versus 0% in flattened 2D, with an average signal-loss shift on the order of 10 dB in the overpass case.
Takeaways & Limitations
Propagation traces with physically present occlusions provide a basis for evaluating how communication conditions affect downstream CDA and intersection-management studies.
Takeaways & Limitations
The reported results come from a controlled comparison using one investigated overpass scenario and fixed Sionna RT, radio, and solver configurations.
Abstract
from arXiv · showhide
End-to-end simulation of connected and automated vehicles requires consistent fidelity across mobility, environment modeling, V2X radio propagation, and decision-making/control modules. However, 2D representations of complex road infrastructure often fail to capture critical signal propagation dynamics, leading to overly optimistic connectivity assumptions. This paper presents a unified workflow within CAVISE that integrates map-based scene preparation with microscopic mobility simulation. The proposed framework incorporates a comparative analysis of trace-driven propagation in 3D versus a simplified 2D baseline, and a modular interface for integrating Autonomous Intersection Management (AIM) models. Collectively, these capabilities enable high-fidelity Cooperative Driving Automation (CDA) experiments. Leveraging ray tracing using Sionna RT at 5.9 GHz with the same radio and solver configuration for both geometry variants, we show that planar reduction in multi-level road infrastructure can remove physically present occlusions and substantially distort signal-loss dynamics. In a bridge overpass case study, the 2D baseline eliminates an occlusion interval observed in 3D, changing the outage behavior from intermittent to consistently connected and yielding an average signal-loss shift on the order of 10 dB. These propagation-induced biases highlight the inability of planar models to capture vertical occlusions, necessitating 3D-aware communication modeling to ensure the validity of cooperative driving automation and intersection control evaluations.
I. INTRODUCTION
End-to-end CAV simulation must couple realistic mobility, environment, V2X propagation, and control modeling. This paper addresses planar-model limitations through an integrated CAVISE workflow combining 3D-aware propagation analysis with AIM integration.
- Integrated simulation fidelity is essential because subsystem simplifications can bias system-level conclusions and mislead design decisions.
- 2D propagation abstractions cannot represent elevation-dependent occlusion and non-line-of-sight conditions in overpasses, underpasses, bridges, and stacked junctions.
- CDA and AIM evaluations require coupled mobility, propagation, and control modules because burst losses and connectivity holes can alter controller information and qualitative outcomes.
- Prior ray-tracing integration work identified challenges in constructing true 3D environments, leaving a need for controlled evaluation of multi-level infrastructure effects.
- The proposed CAVISE workflow combines map-based scene preparation, microscopic mobility, 3D-versus-2D propagation, and AIM integration for reproducible CDA experiments.
II. RELATED WORK
Related work establishes the need for 3D vehicular propagation modeling and physics-based ray tracing in complex environments. This paper distinguishes its evaluation by isolating geometry flattening while holding propagation and radio settings constant.
- Planar reduction can systematically fail in multi-level road environments, motivating 3D modeling and practical 3D scenario support for V2X studies.
- GPU-accelerated geometry-based ray tracing models occlusion, reflection, and refraction in complex scenes relevant to CDA communication.
- The paper isolates geometry flattening by keeping the propagation engine and radio configuration identical between the 3D and 2D baselines.
III. WORKFLOW AND EXPERIMENT SETUP
The experimental setup generates comparable 3D and flattened 2D propagation traces in Sionna RT, covering the workflow, radio settings, baseline definition, and overpass case study.
- The setup is organized around generating comparable 3D and flattened 2D propagation traces in Sionna RT.
- It defines the pipeline overview, radio settings, flattened baseline, and representative overpass scenario used in the comparison.
- The setup supports analysis of propagation differences without changing the stated experimental dimensions across geometry variants.
A. Workflow Overview
The workflow connects map-based scene preparation, microscopic mobility generation, and Sionna RT propagation evaluation, with 2D or 3D traces available for comparison.
- Workflow Overview: The pipeline constructs 3D scenes from map data, generates vehicle trajectories microscopically, and evaluates propagation through ray tracing.
- Workflow Overview: SUMO provides time-indexed vehicle poses, while Sionna RT updates node positions and computes received power between transmitter–receiver pairs.
- Workflow Overview: The workflow supports trace-based comparison after propagation evaluation in either 2D or 3D geometry.
- Workflow Overview: The workflow uses Sionna RT in Path Solver mode with identical radio parameters and solver settings for the 3D and flattened 2D configurations.
- Workflow Overview: Propagation evaluation uses 5.9 GHz, 20 dBm transmit power, single-element isotropic antennas, reflections, refraction, and a maximum path depth of four interactions.
C. Flattened 2D Baseline
The baseline isolates elevation by comparing original 3D road geometry with a flattened plane while preserving horizontal footprints and object placement. In the overpass scenario, flattening removes an overhead occluding structure present between vehicles on different road levels.
- Scene construction: The 3D variant preserves elevation differences, whereas the 2D variant collapses vertical coordinates into a single plane.Both variants retain horizontal footprints and object placement, with all other settings unchanged.
- Scenario: The representative scene contains an upper bridge and a lower road beneath it, creating different vehicle elevations.This overpass–underpass configuration is used to examine visibility relationships affected by planar reduction.
- Occlusion mechanism: In 3D, the bridge deck can occlude propagation between vehicles on different road levels when a receiver is under the deck.The corresponding overhead obstruction does not remain in the flattened geometry.
IV. PROPАGATION FIDELITY RESULTS
The controlled comparison evaluates signal-loss time series and outage behavior for a selected overpass link in flattened 2D and full 3D geometries. The 3D bridge creates intermittent outage, while flattening removes the occlusion and produces consistently connected behavior with a large signal-loss difference.
- Comparison: The experiment compares propagation fidelity through signal-loss time series and outage behavior for a representative overpass link.Figures 2 and 3 show the flattened 2D baseline and full 3D geometry, respectively.
- 3D result: In 3D, the bridge introduces an occlusion interval that produces increased signal loss and intermittent outage.The occlusion is visible in the full 3D geometry's signal-loss trace.
- Quantitative result: 42% (3D) to 0% (2D) is the outage-rate change caused by flattening in this case.The 2D trace is consistently connected after the occlusion disappears.
- Quantitative result: An average signal-loss shift on the order of 10 dB accompanies the change from full 3D to flattened 2D geometry.The radio and solver settings are not varied in the reported comparison.
B. Implications for Cooperative Applications
The propagation traces are intended to provide trace-consistent communication conditions for downstream CDA experiments. Geometry-induced distortions can bias cooperative driving and intersection-management evaluations when planar simplifications remove burst losses or transient occlusions.
- Application implications: Propagation-induced distortions can translate into application-level bias in cooperative driving and intersection-management evaluations.The paper describes these distortions as relevant to downstream CDA experiments.
- Application implications: Planar simplifications may make cooperative driving and intersection management appear more stable by removing burst losses and transient occlusions.This consequence follows from the stated propagation-induced distortions.
- Workflow role: The workflow supplies trace-consistent communication conditions for downstream CDA experiments.The traces are generated within the end-to-end workflow for use by downstream models.
- Reproducibility: The comparison is reproducible through fixed scene, mobility, coordinate-mapping, and Sionna RT configuration details.The paper lists the map definition, Blender assembly, SUMO settings and seed, coordinate mapping, and ray-tracing configuration.
V. EXPERIMENT LIMITATIONS
The reported findings come from a controlled full-3D versus flattened-2D comparison using identical ray-tracing, solver, and radio settings. The downstream architecture supports CDA and AIM studies, but the propagation results are bounded by one representative overpass configuration.
- Experimental control: The experiment changes only scene geometry flattening while keeping the ray-tracing engine, solver settings, and radio parameters identical.The propagation setup uses Sionna RT with the reported configuration for both variants.
- Scope boundary: The reported signal-loss dynamics and outage behavior correspond specifically to the selected bridge link and representative overpass scenario.This defines the scope of the reported propagation results.
- Downstream integration: Propagation traces serve as the communication layer for downstream CDA studies, including AIM-driven intersection decision-making.The architecture supports controlled experimentation and repeatable model comparison across scenarios and seeds.
- AIM scope: An AIM model includes both machine learning–based and classical algorithmic models.This terminology applies to the paper's modular AIM integration.
A. Architecture Overview
CAVISE uses a decoupled AIM integration architecture that connects OpenCDA with interchangeable AIM modules through a central manager.
- Architecture Overview: An independent AIM module interface supports integration with OpenCDA and external systems such as ROS.
- Architecture Overview: The AIM Manager orchestrates communication between OpenCDA and multiple AIM modules.
- Architecture Overview: The architecture provides plug-and-play evaluation of different decision-making and prioritization algorithms without modifying OpenCDA.
C. AIM Module
CAVISE represents AIM modules as independent, configurable components and demonstrates their execution within the simulation environment.
- C. AIM Module: An AIM Module packages a model implementation with configuration files and optional training components or model weights.
- C. AIM Module: The modular structure supports rapid prototyping, comparative evaluation, and systematic validation under identical simulation conditions.
- C. AIM Module: The AIM scenario demonstration shows the CAVISE simulation environment during execution of an intersection-control scenario.
- C. AIM Module: 42% outage in 3D versus 0% in the flattened baseline illustrates how geometry simplification can produce deceptively optimistic connectivity.