Advanced artificial intelligence-based multi-sensor fusion for environmental perception in autonomous electric vehicles

As autonomous electric vehicles (AEVs) continue to evolve, the demand for robust obstacle detection systems becomes increasingly critical to ensure safety, efficiency, and adaptability in real-world environments. This review presents a comprehensive synthesis of recent advancements in sensor fusion technologies, emphasizing the integration of light detection and ranging(LiDAR), radar, and camera-based vision systems. It highlights the role of deep learning architectures—such as you only look once (YOLO), convolutional neural networks (CNNs), and related neural models—in enhancing object detection, classification, and segmentation. The review categorizes key research themes, including fusion methodologies, real-time processing, edge computing, performance in adverse weather conditions, pedestrian detection, and sensor calibration. Special attention is paid to techniques that merge spatial, velocity, and semantic data to mitigate individual sensor limitations. The paper also discusses hardware-accelerated solutions for low-latency inference and the use of lightweight models for deployment on edge devices. Benchmark datasets, of vehicle-to-everything (V2X) and internet of thing (IoT)-based infrastructure, and calibration challenges are examined for their roles in ensuring accuracy and reliability. Drawing from over 100 referenced studies, this work serves as a foundational resource for researchers and developers aiming to advance artificial intelligence (AI)-based sensor fusion systems in next-generation AEVs.