Effective monitoring of road fixtures is essential for urban safety and functionality. However, traditional inspections are time-consuming, costly, and error prone, while current automated solutions struggle with high initial setup costs, limited flexibility preventing wide adaptation, and reliance on centralized processing that can delay response times. This study introduces an edge AI-based remote road fixture monitoring system which automatically and continuously updates the information of the road digital twin (DT). The main component is a small-sized edge device consisting of a camera, GPS, and IMU sensors designed to be installed in typical cars. The device captures images, detects the fixture, and estimates their location by employing deep learning and feature matching. This information is transmitted to a dedicated cloud server and represented on a user-friendly user interface. Experiments were conducted to test the system’s performance. The results showed that the device could successfully detect the fixture and estimate their global coordinates. Outputs were marked and shown on the road DT, proving the integrated and smooth operation of the whole system. The proposed Edge AI device demonstrated that it could significantly reduce the data size by 80–84% compared to traditional methods. With a satisfactory object detection accuracy of 65%, the system effectively identifies traffic poles, stop signs, and streetlights, integrating these findings into a digital twin for real-time monitoring. The proposed system improves road monitoring by cutting down on maintenance and emergency response times, increasing the ease of data use, and offering a foundation for an overview of urban road fixtures’ current state. However, the system’s reliance on the quality of data collected under varying environmental conditions suggests potential improvements for consistent performance across diverse scenarios.
27 July 2025 in Proceedings of the 37th International Symposium on Automation and Robotics in Construction (isarc)
The real-time estimation of human activity level in indoor environments is crucial for thermal comfort prediction and optimizing HVAC systems, as well as supporting health monitoring and ergonomic assessments. Existing methods typically categorize activities into predefined types and therefore are not efficient for new activities. Furthermore, the transition between activities is not captured in these classification methods. To address these limitations, we propose a novel Activity Intensity Score (AIS) framework that provides a continuous, non-intrusive assessment of human activity intensity. The proposed AIS framework uses kinematic parameters derived from video-based pose estimation to calculate a dynamic intensity score. Key kinematic parameters, such as angular speed, angular acceleration, range of motion, movement frequency, and rotational energy, are extracted from pose landmarks using MediaPipe. These parameters are then normalized and combined to generate a continuous AIS that reflects real-time variations in movement intensity. The proposed approach was validated through controlled experiments with participants performing activities of varying intensities, including low (e.g., sitting), moderate (e.g., walking), and high (e.g., jumping jacks). Results showed that the AIS effectively distinguishes between different activity intensities, with higher AIS values corresponding to more intensive activities. The proposed method addresses the need for a more granular understanding of activity levels beyond simple categorical classification. This research contributes to real-time activity estimation, which has applications in optimizing indoor environmental conditions, health monitoring, and dynamic ergonomic assessments.
27 July 2025 in Proceedings of the 37th International Symposium on Automation and Robotics in Construction (isarc)
3D human joint estimation is essential for enabling effective human-robot interaction in construction automation, facilitating precise monitoring of worker movements to enhance safety, ergonomics, and operational efficiency. Vision-based systems utilizing multi-camera setups offer diverse perspectives to address challenges such as occlusions, projection ambiguities, and sensor noise. However, these systems depend heavily on accurate camera calibration to align views and synchronize measurements. Traditional manual calibration methods are timeconsuming, labor-intensive, and prone to human error, making them unsuitable for dynamic construction sites where frequent camera repositioning is required due to shifting conditions and tasks. This study proposes a novel framework that leverages external marker detection for real-time, automated camera calibration, eliminating the need for manual intervention. This approach significantly reduces setup time, minimizes errors, and ensures reliable performance in rapidly changing environments. Furthermore, the framework integrates an Extended Kalman Filter (EKF) to fuse 2D joint locations from multiple cameras, effectively handling sensor noise and the nonlinear nature of human motion. By combining marker-based calibration with EKF-based fusion, the proposed framework delivers a robust and automated solution for 3D human joint estimation, enhancing safety, efficiency, and adaptability in construction automation applications.
27 July 2025 in Proceedings of the 37th International Symposium on Automation and Robotics in Construction (isarc)
Highway work zones present hazardous conditions where worker safety is at constant risk due to interactions with construction equipment. In such environments, accurately defining hazard zones around equipment is vital for timely detection and prevention of incidents. Traditional static hazard zone models, often relying on simple geometric shapes, fall short as they cannot adapt to changes in equipment speed and movement, leading to either inadequate safety coverage or frequent false alarms. This study proposes a dynamic hazard zone model that leverages a probabilistic approach based on bivariate Gaussian distribution to generate adaptive and realistic hazard zones. Real-time data from GPS and inertial measurement unit (IMU) sensors enable the model to adjust hazard zone dimensions dynamically in response to equipment behavior, improving both accuracy and responsiveness. Simulations conducted in Webots software demonstrated that the proposed model achieved 93% accuracy in hazard detection, significantly outperforming the 60% accuracy of static circular zones tested in the same environment. The findings underscore the potential of dynamic hazard zone modeling to enhance safety in highway work zones, providing a practical and effective approach for reducing risks.