Abstract: Dump trucks, cranes, excavators, and loaders are among the heavy load vehicles used on construction sites that have significantly increased in number as a result of the quick growth of infrastructure. These devices greatly increase production, but they also pose serious safety issues because they are hard to spot, site circumstances change, people grow tired, equipment malfunctions, and they are near employees. In order to enhance the safety, dependability, and accident avoidance of high-load construction trucks, this article proposes a comprehensive safety framework that integrates Artificial Intelligence (AI) and Embedded C++. Using cameras, radar, lidar, inertial sensors, and load cells, the suggested system aggregates data from several sensors. Additionally, it uses real-time edge AI models to identify dangers, monitor blind spots, avoid overloads, and detect hazards. A real-time operating system (RTOS) and a safety-critical embedded software architecture designed in modern C++ (C++17/20) guarantee that the system can handle faults, respond consistently, and connect safely to car actuators. For extended use, the framework incorporates fail-safe modes, extra watchdogs, and safe over-the-air updates. Controlled field testing and simulations are used to generate an exact experimental design. Time-to-intervention, false alarm rate, detection accuracy, system delay, and system availability are examples of performance measures. The suggested approach shows how AI-assisted active safety systems can be implemented on resource-constrained embedded platforms while meeting dependability and real-time requirements. The next generation of smart construction vehicles will have a workable, scalable, and industry-ready solution thanks to this effort.

Keywords: fault-tolerant systems, sensor fusion, embedded C++ systems, high load trucks, construction safety, predictive risk assessment, autonomous safety intervention, edge artificial intelligence, real-time operating systems (RTOS), and IoT-enabled construction equipment.

Downloads: | DOI: 10.17148/IJIREEICE.2025.131222