Freight vehicles approaching signalized intersections require reliable detection and motion estimation to support infrastructure-based Freight Signal Priority (FSP). Accurate and timely perception of vehicle type, position, and speed is essential for enabling effective priority control strategies. This paper presents the design, deployment, and evaluation of an infrastructure-based multi-modal freight vehicle detection system integrating LiDAR and camera sensors. A hybrid sensing architecture is adopted, consisting of an intersection-mounted subsystem and a midblock subsystem, connected via wireless communication for synchronized data transmission. The perception pipeline incorporates both clustering-based and deep learning-based detection methods with Kalman filter tracking to achieve stable real-time performance. LiDAR measurements are registered into geodetic reference frames to support lane-level localization and consistent vehicle tracking. Field evaluations demonstrate that the system can reliably monitor freight vehicle movements at high spatio-temporal resolution. The design and deployment provide practical insights for developing infrastructure-based sensing systems to support FSP applications.
Roadside sensing using Light Detection And Ranging (LiDAR) and cameras has gained increasing attention for transportation applications such as traffic monitoring and cooperative perception. In this work, we focus on freight signal priority (FSP) and leverage infrastructurebased sensing to estimate vehicle time of arrival (ToA) for priority requests.
FSP aims to improve truck mobility at signalized intersections while accounting for their slow acceleration and high fuel cost in stop-and-go conditions. Despite accounting for only ~9% of vehicles and ~17% of VMT, heavy-duty trucks contribute ~39% of life-cycle GHG emissions (Moultak et al., 2017). Eco-FSP approaches have shown 5-10% network-wide fuel savings and up to 26% travel time reduction, with connected-truck systems reporting up to 25.3% fuel and CO₂ reduction (Kari et al., 2014;Park et al., 2019). However, deployments relying on V2I/V2V communication require onboard equipment, limiting real-world adoption.
To overcome these challenges, we develop an infrastructure-based, multi-modal freight vehicle detection system integrating LiDAR and cameras. The system contains two subsystems tailored to site-specific terrain constraints: (1) a long-range LiDAR subsystem with camera installed at the intersection, and (2) a mid-range LiDAR subsystem with camera placed midblock to overcome visibility blockage caused by roadway curvature. Wireless communication transmits midblock detection results to the intersection for downstream FSP operations.
The key contributions of this work are summarized as follows: l A complete design, implementation, and evaluation pipeline for an infrastructure-based, multi-modal freight vehicle detection system is presented. l A dual-subsystem detection architecture is proposed to address the unique challenges of the FSP deployment site, particularly limited visibility caused by roadway curvature. l Data acquisition and processing pipelines are developed, and preliminary detection and tracking results are demonstrated and evaluated. l LiDAR-derived coordinates is registered with geodetic reference frames, enabling lanelevel vehicle localization.
As shown in Figure 1, the FSP system is deployed at the main intersection with pole-mounted sensors providing long-range coverage in two opposing directions. However, roadway curvature in one direction causes partial occlusion, limiting far-distance visibility. To mitigate this, a midrange detection subsystem is installed at the midblock using the existing pole and cabinet to detect freight vehicles earlier. This two-site setup introduces an additional requirement for reliable wireless communication between the midblock and the intersection. Specifically, Figure 1(c) presents a bird’s-eye view illustrating the terrain and subsystem placement. The system consists of two subsystems-one at the main intersection and the other at the midblock-as shown in Figure 2. A LiDAR-camera pair with a long-range solid-state LiDAR (Livox TELE-15) is installed at the intersection to monitor the southbound direction. To address the occlusion caused by horizontal roadway curvatures, a second LiDAR-camera pair with a midrange mechanical LiDAR (Robosense Ruby Plus) is deployed at the midblock to ensure continuous detection of approaching traffic. A wireless communication bridge (Ubiquiti AirFiber 5XHD) enables data transmission between the two subsystems. For the detection system components, a mid-range sensing unit is deployed at the midblock to monitor vehicles approaching from the south. It includes detection sensors, a PoE switch, an edge computer, and a point-to-point wireless transmitter. The edge computer processes LiDAR point clouds locally to obtain vehicle detection results, which are then transmitted to the main intersection via a Precision Time Protocol (PtP) wireless bridge, with time synchronization maintained using PTP (Cho et al., 2009). At the main intersection, the edge computer receives detections from both directions and generates freight priority requests for the traffic controller. As shown in Figure 3, the LiDAR, camera, and AirFiber are connected and powered through a Power-over-Ethernet (PoE) switch. Because the LiDAR does not support native PoE, a PoE splitter is used to merge power and data into a single cable. The camera images and LiDAR point clouds are transmitted through the PoE switch to the edge computer for data acquisition. The AirFiber link is primarily used to send detection results and will not be further discussed in this paper.
Camera and LiDAR data are collected using ROS2 with timestamps synchronized to the system clock. Figure 4 shows example data captured at the main intersection and the midblock, where freight vehicles are clearly visible in both the images and the point clouds.
For GPS calibration, we register the LiDAR point cloud to global coordinates (latitude/longitude). An OXTS AV200 GNSS unit is mounted on a Toyota Corolla, and GPS data is recorded alongside LiD
This content is AI-processed based on open access ArXiv data.