InSPE: Rapid Evaluation of Heterogeneous Multi-Modal Infrastructure Sensor Placement

TL;DR

InSPE introduces a rapid evaluation framework for infrastructure sensor placement at intelligent intersections, defined by three surrogate metrics—perception coverage $C$, occlusion $O$, and information gain $IG$—and validated via a CARLA-based data generator that yields the Infra-Set dataset. The method models sensing with ray-cast camera and LiDAR frustums, computes voxel-level coverage and occlusion, and quantifies information gain from waypoint-driven occupancy statistics, subsequently fusing these signals into a final perception score. Infra-Set provides a large-scale, open benchmark to study heterogeneous multi-modal infrastructure sensing, and extensive experiments show LiDAR-enhanced configurations substantially improve detection performance, with surrogate metrics correlating to 3D perception outcomes. Overall, InSPE enables fast, scalable sensor-placement optimization for infrastructure-to-infrastructure perception in diverse intersection geometries and conditions, supporting cost-effective, robust intelligent intersection design.

Abstract

Infrastructure sensing is vital for traffic monitoring at safety hotspots (e.g., intersections) and serves as the backbone of cooperative perception in autonomous driving. While vehicle sensing has been extensively studied, infrastructure sensing has received little attention, especially given the unique challenges of diverse intersection geometries, complex occlusions, varying traffic conditions, and ambient environments like lighting and weather. To address these issues and ensure cost-effective sensor placement, we propose Heterogeneous Multi-Modal Infrastructure Sensor Placement Evaluation (InSPE), a perception surrogate metric set that rapidly assesses perception effectiveness across diverse infrastructure and environmental scenarios with combinations of multi-modal sensors. InSPE systematically evaluates perception capabilities by integrating three carefully designed metrics, i.e., sensor coverage, perception occlusion, and information gain. To support large-scale evaluation, we develop a data generation tool within the CARLA simulator and also introduce Infra-Set, a dataset covering diverse intersection types and environmental conditions. Benchmarking experiments with state-of-the-art perception algorithms demonstrate that InSPE enables efficient and scalable sensor placement analysis, providing a robust solution for optimizing intelligent intersection infrastructure.

Figures (9)

• Figure 1: Illustration figure of three types of sensor placements at three example intersections. (a) features sensors concentrated near the center of the intersection, whereas (b) and (c) employ a more dispersed placement throughout the intersection. The camera arrangement in (a) is similar to that of the V2XSet xu2022v2xvit dataset, (b) resembles those in the DAIR-V2X yu2022dairv2x and RCooper hao2024rcooper datasets, and (c) is akin to the V2X-Real xiang2024v2xreal dataset. FOV direction is the field of view direction of the camera.
• Figure 2: Sensor Placement Evaluation Framework. HM Perception framework refers to heterogeneous multi-model perception framework.
• Figure 3: Camera View Frustum Model
• Figure 4: Mechanical and Solid State LiDAR Model
• Figure 6: The relationship between perception surrogate metrics and multi-class mAP under different sensor placements.